1,205 Posts served

4,986 Conversations started

64-bit and Intel

By Steve Pitzel (Intel) (16 posts) on September 20, 2006 at 7:06 pm

A few days ago an OEM asked whether or not Intel had 64-bit... It was kind of funny, kind of sad - and, of course, the OEM should know we do...but, then again, I have to admit we've done a pretty lousy job to date of letting folks know about our 64-bit offerings - while AMD has even included 64 proudly in some of their product names. Okay, if you're into Roman numerals, I guess the 64 isn't exactly hidden in Viiv, but that's another story... You probably know about Itanium, which has been a 64-bit architecture since its inception. But the truth is our Xeon line has supported 64-bit for over a year and everything out there you'll find with a "D," as in Pentium? D, or Celeron? D (yes, even Celeron? D supports 64-bit), and all the new Intel? Core2 Duo processors as well. Intel? 64 is alive and well. And...your 32-bit OS and apps will run just fine on them too :).

Categories: Intel® Software Network 2.0, Multicore

Comments (154) 

By denise.latscha@intel.com on September 21st, 2006 at 3:01 pm
Actually our Xeon processors have supported 64 bits for over two years now, and our desktop processors starting shipping with 64-bit computing support over a year ago (around the time the supporting Windows OS released). By the end of 2005 the vast majority of our desktop processor shipments supported Intel® 64.

As you say, all processors based on the new Intel® Core (TM) architecture support 64-bit computing as well, including mobile processors.

Note that Itanium, wihch is a 64-bit architecture targeted for the most demanding compute environments, released in 2001.

By jim.mccartney@crypkey.com on September 25th, 2006 at 9:46 pm
been on your site waaay too long, still can't find out for cetatin if T2700 is a 64 bit processor. Even this article is vague - "supports 64 bit" - what the hell does that mean? Emulates 64 bit? Doesn't crash if it sees 64 bit code?
Either it is a 64 bit processor or not, and it if it is it should say so. Can you say "Celeron® D _IS_ a 64 bit processor"? The fact that you don't is making people run, not walk, to AMD, where the is no obfuscation of this issue.
You would do well to put a list of processors that _ARE_ 64 bit on your front Web page, and just put an end to it - the confusion is uncharacteristically unprofessional.

By Steve Lionel on September 25th, 2006 at 11:41 pm
The T2700 is a "Core Duo" processor and does not supprt 64-bit. The T7200 is a "Core 2 Duo" and does support it. These processors can run in either 32 or 64 bit mode depending on which OS you boot. So if you boot Windows XP on a T7200, that's 32 bit and it won't execute 64-bit code. If you run XP Professional x64, then that's 64-bit and you can run both 64 and 32-bit code. With Linux, you have to have a distro's "x86_64" variant to run 64-bits. It's no different with AMD - some of their processors support 64-bit, some don't, and those that do can run either way.

It would be just as confusing to say that the processor is "64-bit" for someone who wants to run a 32-bit OS.

By rp_tomj@hotmail.com on September 27th, 2006 at 11:28 pm
i still cannot understand any of this. Here is the OEM order code for the T7200: LF80537GF0414M. As you can see there is no "D" in it. I can find no spec that tells whether a processor runs 64 bit o/s anywhere. How can a mere mortal decode this web site?

By bmozmuzik@yahoo.com on September 28th, 2006 at 11:31 am
HERE"S MY TAKE ON IT EVERYONE! About 4 months ago I was looking into buying a laptop. Now, i am a very thorough buyer and investigate things. The new core duo was just being released. All the sales people at the stores (like Best Buy etc) all told me that this processor was 64 bit. And they explained it had to be because the Pentium D was also. But, I couldn't seem to verify this on this website. I fully agree that Intel needs to make this fact clear on this website, because in my opinion 64 bit is the future, and people buying computers these days want to think about the future especially when a paradigm shift to a completely different bit level like this is being made. Now, finally, PC's will have graphics capabilities like Mac's. Which brings me to why I can conclude Intel is making it difficult to distinguish why the first Core Duo was not 64 bit. After researching the first Core Duo intensely for about 2 weeks I finally found out that it was only 32 bit and was released in a rush for the new Mac line of Intel computers. But Intel was always planning on releasing the Core 2 Duo later when it was ready. I know this email sounds rushed, but you see the research I found suggested that intel made a mistake by not meeting a certain deadline for the Mac computer contracts it had and had to release the 32 bit version right away. The worst part of all this is that I called the Intel cus support line multiple times and most the reps were not even aware that the first version was only 32 bit! I spoke to maybe 2 that were aware and I eventually found the whole story on Computer review forums. But, anyway, that is the past and it now seems that the new Core 2 Duo is here and it is 64 bit? Right? Not that this website would say!! Good thing Steve Pitzel wrote this little comment otherwise who would know right? Ridiculous.

By fractalzone@gmail.com on October 1st, 2006 at 1:55 am
What currently shipping Intel desktop and laptop CPUs are 64-bit *AND* dual core?

That is a major question in my mind, for two reasons:

1) I sell computers at retail and recommend them as a consultant.

2) I earn the bulk of my living as a software developer and sometimes need to write low-level code...x64 (actually AMD64) is what I prefer, given the generally kludgey architecture of the whole x86/x64 series. Performance matters to me. Marketing hype does not.

I am really looking for a dual-core, 64-bit chip uProc. AMD's Athlon X2 has been my choice for my next machine, but I'm open to suggestion. Bandwidth to memory is important. What I need is a decent technical, side-by side comparison of all of Intel's and AMD's CPUs. It is suspicious that Intel isn't making those sorts of specifications readily available. Maybe AMD is making the more advanced uProcs for the money?

Mr. Wizard
(former Intel stockholder)

By bmozmuzik@yahoo.com on October 1st, 2006 at 10:29 am
I recently started looking at this website again and still find it ABSOLUTELY vague, misleading, and suspicious!!! Check this out. I wrote tech support with the following question:

I feel ill when I browse this web site. What are you guys hiding from us? My head is spinning. i am looking for a laptop, and didn't buy one months ago because after 2 weeks of mind spinning research and inept sales people in stores and even your own cus support reps, found that the core
duo was not 64 bit. You know the only thing that is clearly 64 bit on your web site is your server processors. I just spent over 2 hours spinning on your web site to see if the new core 2 duo is 64 bit and I still don't really know. Is it just enableing 64 bit- What is the deal? What processors are REALLY and FULLY TRULY 64 bit that you guys offer? Why is it that AMD seems to have no problem expressing this to their customers.

Here is the response from tech support:

Thank you for contacting Intel® Technical Support.

I understand you want to know if the new Intel® Coreâ"ž¢2 Duo Desktop Processor supports Intel®
Extended Memory 64 Technology.

This processor has Intel® Extended Memory 64 Technology, meaning, it will run 64 bit applications.

For further information on the features and characteristics for the Intel® Coreâ"ž¢2 Duo Desktop Processor, please check the following website: http://indigo.intel.com/compare_cpu/showchart.aspx?mmID=8843.....ture=en-US

Additionally, for further information on the Intel® Extended Memory 64 Technology, you can visit the following website:
http://www.intel.com/technology/64bitextensions/

For your information, the only real 64 bit processor is the Intel® Itanium® Processor.

Did you catch the last part. Itanium is currently the only REAL 64 bit processor. So, the game of mystery goes on. This brings more questions to my mind. 1, are AMD REAL 64 bit processors? Intel's website says that the advantage of 64 bit comes in being able to access more than 4 Gigs of cache memory. AMD's website says nothing about this. 2, Should we all be buying Itanium processors for our own consumer computers? and 3, What about quad core? AMD is introducing this shortly. This is craziness. Wouldn't it be nice if this info was just cut and clear. Who is the key master to this drama of hidden facts? Are we all just supposed to prod the tech support people of various companies to find out what they are really trying to sell us?

By thatdummie@aol.com on October 1st, 2006 at 5:39 pm
Pitzel says:
(yes, even Celeron® D supports 64-bit), and all the new Intel® Core™ 2 Duo processors as well.

What's in a name? (thatdummie) ok I read this whole page and it seems to me 64 bit is alive and well in somethings with 2 duo core, but not available in "any" laptops. Did I miss something or do I need to postpone my purchase of a laptop for a while longer?

By davehewlett@msn.com on October 2nd, 2006 at 9:34 pm
URL of all the Intel processors and what they support. Note Viiv, Pentium D, Core 2 Duo all support the 64 bit need. Note that the other supporting hardware and bios are needed. As said before the core duo is a 32 it CPU only. It seems the 64 bit answer was always there but not easy to find.

http://www.intel.com/products/processor_number/proc_info_table.pdf

Try this list.

By husseindilawer@hotmail.com on October 3rd, 2006 at 3:12 am
I have had the same problem with intel, now i just build computers with AMD processors because they sell better and outperform any comparable intel chip. Intel has been poor with marketing their processors correctly, where as on the other hand AMD flaunts their 64 bit chips with pride.

By bmozmuzik@yahoo.com on October 3rd, 2006 at 10:59 pm
OK everybody, it seems there are still comments being made as to what is 64 bit. My post a few days ago should clarify exactly what is reality here. Various chips from Intle "support 64 bit" For the desktop, P-D, Cel-D, and both server processors, Xeon, and Itanium. For laptops ONLY Core 2 Duo. Core Duo 1 never supported it and I explained this about a week ago, it was a premature manufactr for a deadline set by Macintosh. So all these processors support 64 bit. But ONLY ITANIUM is a true 64 bit architecture. So, in reality this is the only true 64 bit processor!!!!! I unfortuneately don't know if the AMD processors are true 64 bit architecture either, but I would venture to say they are. I recently sent another message to Intel tech support to clarify this and they chose not to respond.

All I know is that I will probably never buy an Intel chip again, and will move to AMD (HP is a great supporter of their chips). Intel is way too deceptive and concealing of VITAL information about their chips. They are not giving us the real deal on the website, through their tech support, or the sales people out in the stores that sell their products. This is not the type of company I feel comfortable doing bizness with!!!!

By jrumble@comcast.net on October 4th, 2006 at 6:05 am
I have a pent D and it sure wont run 64 bit windows..
a gateway 835gm to be exact.. the only computer i ever bought in one piece..

By andy_jp@hotmail.com on October 6th, 2006 at 2:02 pm
I'm no processor expert, but here's my understanding of what's going on:

Both AMD and Intel use very similar technologies in making their processors "64-bit". However, these are simply extensions to the x86 architecture, meaning that these processors are not true 64-bit chips (I'm talking about any Intel processors with "EMT64" or AMD processors with "AMD64").

Unlike AMD, Intel seem reluctant to market these chips as 64-bit, which may make Intel the more honest of the two companies. However, I could be completely wrong...

By andy_jp@hotmail.com on October 6th, 2006 at 6:52 pm
Sorry, "EM64T". It's been bugging me all day...

By jzmcfire@hotmail.com on October 6th, 2006 at 9:12 pm
Why doesn't Steve Pitzel, who began this, or anyone from Intel who can speak with authority jump in and clarify: is the Core 2 duo capable of accessing more than 4GB of memory? If not, why not? And if not what exactly does "supports 64-bit" mean?

Failure to jump in and clarify would support intimations made here of a kind of dishonesty more often seen in used car sales.

By jzmcfire@hotmail.com on October 6th, 2006 at 9:22 pm
Oh ... 1 terabyte:

"Intel® 64 architecture (formerly known as Intel® Extended Memory 64 Technology, or Intel® EM64T) enables 64-bit computing on server, workstation, desktop and mobile platforms when combined with supporting software.¹ Intel 64 architecture improves performance by allowing systems to address more than 4 gigabytes (GB) of both virtual and physical memory. Today, all Intel® processors for server and workstation platforms support 64-bit computing. And with the introduction of Intel® Core™ 2 Duo processors in the second half of 2006, most Intel desktop and mobile processors are also 64-bit capable. Intel 64 provides support for:

* 64-bit flat virtual address space
* 64-bit pointers
* 64-bit wide general purpose registers
* 64-bit integer support
* Up to one terabyte (TB) of platform address space
"
at http://www.intel.com/technology/intel64/index.htm

Sorry for doubting, Intel gents.

By anhedral@gmail.com on October 6th, 2006 at 10:19 pm
Hi There,

I had my nice shiny new dell inspiron 9400 delivered in February of this year. I purchased it on the premise the Duo was 64 bit, so I had an upgrade path available to me (especially as I was switching from Desktop to Laptop).

So I look in to uprgading to Windows 64bit, and I find that the Duo is in fact only 32 bit, and that if I wanted 64-bit, I should have purchased a laptop with a Duo "2" inside it, which was released shortly after???

I feel cheated and upset.

By Razer336@hotmail.com on October 7th, 2006 at 9:40 pm
I wish Intel would get its act together and put 64 on all of their 64 bit processors, as you cannot put a patent on a number ( as Intel found out with their 286, 386 & 486 processors) they would not have any comeback from AMD. I have recently bought a Core 2 Duo and find it very fast and stable and it runs multiple application with no problems (an area where my AMD processor would often freeze) it would just be nice to know if I purchased a 64 bit OS it would work and I wouldn't be wasting my money.

By todd_gibson@hotmail.com on October 10th, 2006 at 3:42 pm
I agree, why has Intel not responded to this thread? I see answers, thoughts, and facts from many people that are outside of Intel. I just purchased a MacBook Pro which the salesperson told me that it was a 64-bit processor (T2700).

By denise.latscha@intel.com on October 10th, 2006 at 6:54 pm
To answer some of the questions that have surfaced from my confusing answer above as well as on the other blog thread on this site:

1. Folks outside our companies use x86 to refer to both AMD64 and Intel® 64 architectures (the latter was earlier named Intel® EM64T and we have recently renamed it to reduce confusion - which hopefully it will do eventually).
2. Intel 64 architecture supports both 64bit and 32bit computing. You can install either a 32bit or 64bit Operating System on the hardware, and if you install a 64bit OS you can run either 32bit or 64bit apps on it (your drivers will need to be all 64bit - see your Operating System vendor for details).
3. The Intel® Core(TM) Duo processors support only 32bit computing. Intel® Core(TM) 2 Duo processors support both 64bit and 32bit computing. The former released early this year and the latter recently on laptop platforms.
4. Yes to Razer336 above - you should be able to install a 64bit OS on your system and it would work. Again you need to consult your platform or OS vendor for details on how to get it.
5. We had 64bit support in the desktop hardware well before the Intel Core 2 Duo processor was released - the vast majority of Q4'05 desktop processor shipments by Intel supported 64bit computing (we released desktop shipments for 64bits the quarter before the Windows OS starting supporting it).
6. Mobile support for 64bit computing started with the new Intel® Core(TM) 2 Duo microarchitecture, which released recently.
7. AMD64 and Intel 64 architectures are not identical but are similar enough to allow the same Operating Systems and applications to run on them with few changes. If one is a "real" 64bit processor, the other is as well. The Intel Itanium processor also is a 64bit processor, but with a very different architecture targeting high end servers.
8. Not sure what to say about your sales guy, Todd. That processor supports only 32bit computing.

By Steve Pitzel on October 10th, 2006 at 7:35 pm
Thanks, Denise! And thanks everybody for the questions, the answers, and even the baiting and bashing! Keep it comin' ! It's the only way we get better. As I said, we didn't do a great job of letting folks know about Intel and 64bit (outside of Itanium). Truthfully, even though we liked the idea of 64 bit and lots of memory, we didn't exactly have people tearing down the walls to go 64bit at first -- except for my side of the world in Digital Content Creation -- even there, adoption was slow at first. Our friends at Cakewalk really got the ball rolling there -- Cakewalk ported their entire Sonar app to 64bit in about a month. Then they helped us move the entire audio ecosystem of plug-in vendors over to 64bit. Hat's off to those guys!
BTW, jzmcfire deserves a laurel, and hardy handshake ;) for including our 64bit info link http://www.intel.com/intel64/index.htm (in case you missed it). Thanks, again!
- Pitz

By denise.latscha@intel.com on October 10th, 2006 at 8:15 pm
Here is more info from the Intel website that was recently updated:

http://www.intel.com/technology/intel64/index.htm

enjoy!

By tim@franksinc.com on October 11th, 2006 at 7:41 pm
It took me 20 minutes to find this thread concerning the support of x64 in Intel processors. I am so glad I am not the only one befuddled by Intel's lack of ability in communicating this seemingly simply bit of vital information concerning their products.

By todd_gibson@hotmail.com on October 11th, 2006 at 9:10 pm
Thanks Denise for the response. I called Apple's Technical Support yesterday and they told me that the Intel Core Duo was a 64bit platform. I asked them for a link but they could not find any data and then referred me to an Apple Store. I talked to 3 separate laptop sales reps at Frys and they told me the same thing - 64bit. Now I have purchased a laptop on the basis that it was a 64bit platform and now realize that I have been duped. Being a significant decision maker with 7+ figure budgets, it is good to see that the Intel market/products is as confusing for the OEM as it is for the sales reps. Maybe I should pursue AMD products since they are quite clear on their 64bit support.

By mailrsm@gmail.com on October 12th, 2006 at 2:03 pm
The Intel Itanium is also "a 64bit processor". Can I use that statement with a CoreDuo2? Not "support" 64bit architecture, but is. The Intel CoreDuo 2 IS "a 64bit processor"?

By ctanner@cox.net on October 13th, 2006 at 12:52 am
I think one thing that was not mentioned that probably should be and may clear up the comment about the Itanium processor being the only true 64bit processor.

Itanium was never designed to be backward compatible with support for 32bit instructions. Because of this, they could effectively wipe the slate clean and develop a 64bit processor architecture that truly only handles 64bit instructions. I think that is the only reason for stating Itanium as a true 64bit processor vs. the other 64bit Intel offerings which were designed to support both 64bit and 32bit instructions (a hybrid).

It is not clear to me if AMD has a similar processor as the Itanium in their inventory?

I personally think this is not that difficult to figure out but I do agree that a distinct common market name change would have been preferred for these hybrids 64bit processors by Intel.

By r-neveds@excite.com on October 14th, 2006 at 4:22 pm
I would like to buy a top of the line Dell with a 64 bit chip/system. They are limited to a max RAM of 4 Gigs. This is not support of 64 bit.

By markchm@msn.com on October 14th, 2006 at 10:14 pm
It looks to me as if both AMD and Intel are hiding behind the "can run a 64bit OS" story. (of course, they can run commodore 64 progs in emulation, too! LOL) EM64T/64Bit Architecture, or whatever euphemism you prefer - none have 64bit wide bus/data/IO structure. They just rely on translations via the system BIOS to allow the development of 64 bit operating systems to function on a semi-crippled chip until people can afford what everyone is referring to as REAL 64 bit throughout this thread.

By ed_valentine@hotmail.com on October 15th, 2006 at 10:47 pm
I am still confused... I understand that Merom supports 64 bit processing. However, Business Week quotes "But because of limitations in supporting chips, which were designed for older processors, it is limited to 32-bit operation" (http://www.businessweek.com/technology/content/oct2006/tc200.....technology).

So is the combination of Merom with the current Intel® 945GM chipset still 64bit capable? Will they be able to take advantage of the new software that maximizes the 64 bit capabilities?

By Steve Pitzel on October 18th, 2006 at 11:18 pm
Hi Ed,
Thanks for the catch on the Businessweek article - Merom itself does support 64bit and will run a 64bit OS - laptops built to hold more than 4GB of memory are still somewhat hard to find. That brings up another issue r-neveds and Todd Gibson brought up earlier about Dell and Apple. The truth is although we make the processors and chipsets, we don't build the entire system. We can't speak for our OEMs who take those components and create their own configurations. it's always best to go diirectly to them for info on their product lines.
- Pitz

By sguin_programmer@hotmail.com on October 21st, 2006 at 6:51 pm
I see on many vendor sites the tern 64EM technology and the terms you use like "supports 64-bit". Given Intel's record on marketing _schemes_, frankly it worries me. Remember the 386sx? It was "32-bit capable". More digging on the subject revealed that it was 32-bit addressing in 16-bit data chunks.

So let's get to my main question. Looking at the Core 2 Duo, the Pentium D, and the Celeron D, are they 64-bit addressing *and* 64-bit data? If so, why not just *say* that?

Once I find out for sure I'm rushing to my favorite vendor to purchase one. But my first PC was a 386sx... so fool me once, shame on me, fool me twice -- and it's AMD. ;)

By modem1123@hotmail.com on October 22nd, 2006 at 9:04 pm
Yes, reading thru these posts has not answered the question whether Intel Core 2 Duo is a 64 bit processor. I went to Best Buy and Circuit City and all the reps there told me that Intel's Core 2 Duo is a 32 bit processor. They said Intel is coming out w/ a 64 bit processor in scheduled in December. They also said that AMD 64 is a 64 bit processor. Yes, Core 2 Duo will support 64 bit apps but it is NOT a 64 bit processor. I don't know why Intel's reps can't answer a simple question.

By prezweezy@gmail.com on October 23rd, 2006 at 7:02 pm
Alright I'm getting tired of people getting mad about this when all of the answers are right there above. Let's recap. Both the Athlon x64 and the Core x64 technologies are, by and large, the same. The calls themselves are a bit different, but in reality the bus structure, the support abilities, and the main architecture are close enough that they support the same apps and OS's. Neither is an "x64 bit processor" in the truest sense. The Itanium is an x64 bit processor. It is only an x64 bit processor. It's also outlandishly expensive for the average consumer, because it's not made for the average consumer. In order to have an "x64 bit processor" it would be required that the OS, the apps, the drivers, 100% of the computer, be 64 bit. If that's what you want, you can run XP Pro 64, and I think there is an open office x64 distro as well. But if you want to use any software that isn't x64, about 95% of the software out there, then you're just SOL. So if that's what you want, tell Intel and I'm sure they would be happy to put one out for you. However, if you actually want to use your computer, you'll need x86 support as well, which is 32 bit. Therefor, you have a non-x64 bit proc. To clear up the rest, Pentium D, Celeron D, and Pentium D EE all have EM64T. Core Duo, ie the x4000 series chips, are x86 based, no x64 bit. The Core 2 Duo, x6000 series, ARE 64 bit. The Merom x5000 series is the same as the x4000 desktops, and are 32 bit. While the x2000 series does not support 64 bit, the x5000 and x7000 do. Just to stress it one more time, for all of you who have had the wool pulled over your eyes by AMD, they don't have a "true" x64 processor either. And the reason that Intel can't answer a "simple" question, is that if they did, they would have all kinds of legal issues to deal with. And you have the idiots who like to sue for every little thing to thank for that. Also, in case I missed something here, and you're not sure if a processor supports 64 bit, just go to newegg and look it up. It says right at the top "64 bit support: yes or no."

By bret.toll@intel.com on October 23rd, 2006 at 9:14 pm
I'm an architect working on Intel® 64 architecture so let's see if I can clear up a few things.

Markchm, the external data bus has been 64-bits wide since at least the mid 90's, but the external data bus width is not what is meant by a "64-bit processor" or by Intel® 64 architecture. A 64-bit processor is defined as one whose basic, internal arithmetic/logic unit (ALU) is 64-bits wide so it can process 64-bit wide data as a fundamental arithmetic size. Intel® 64 architecture does this and the Merom processor family architecture supports this native instruction width as did earlier Intel® Xeon® and Pentium® processor generations (with Intel® 64 architecture) on the desktop, workstation and server.

I believe another point of confusion here is virtual addressing vs. physical addressing. Virtual addresses are what applications use to refer to memory locations. The OS manages the translation of these virtual addresses into physical addresses (which correspond to the physical memory in the system) and this provides great flexibility in application management. While the Intel® 64 processors can support significantly greater than 4G of RAM, not all platforms will support greater than 4G of RAM.

By martin.p.kemp@gmail.com on October 24th, 2006 at 4:18 pm
"Here's what you need to use Windows XP Professional x64 Edition

Important: Windows XP Professional x64 Edition cannot be successfully installed on x86 (32-bit) systems or 64-bit Intel Itanium"“based systems. 32-bit device drivers are not supported on Windows XP Professional x64 Edition.

"¢ Computer with a supported processor: AMD Athlon 64, AMD Opteron, Intel Xeon with Intel EM64T support, Intel Pentium 4 with Intel EM64T support "

from Microsoft Website - why doesn't the only "true" 64bit processor support Win XP 64?

By modem1123@hotmail.com on October 25th, 2006 at 1:24 am
Thank you, Bret from Intel, for putting everything is precise terms. I will now go w/ the Intel Core 2 Duo instead of the AMD x2. I didn't have a preference but just wanted to find the answer. From reviews that I've seen, the C2D seems to do what I need.

By marksch on October 25th, 2006 at 8:23 pm
Wow this is certainly an interesting debate. As the "demogod" from Softimage (thanks Steve) who has done considerable testing of x64 running on a variety of processors I think I can add a little bit information for you all.

First, the formerly named EM64T processors are, as far as im concerned 64 bit. I dont care if the CPU is a "real" 64 bit part or not, what I care about is that I can install XP 64, drop 8 or 16 gigs of RAM in the machine and actually use it with one app. While at first Intel didnt make a lot of noise about which procs did or did not include EM64T I was pleasantly surprised when I found my P4 (Prescott rev E core) was 64 bit- months after I bought it. So as any good nerd would do I dual booted to Win64 and went "Ok that works, now what" I only had 2 gigs of RAM in the machine so my experiment quickly turned back to using the x32 partition... that is until...

So about a year and a half ago I was invited to go to WinHEC in Seattle and show off Softimage XSI's brand new 64bit native version...and Intel was kind enough to send me a bad-ass 840EE with 8 gigs in it. x64 was installed, XSI 64 bit was installed and holy cow if i didnt suddenly have a 3D scene with 500,000,000 triangles in it occupying over 6 gigs of total RAM footprint. 64 bit had arrived.

Since then every non-mobile Intel CPU has been 64 bit enabled to best of my knowledge. I think the confusion here is really in the notebook line up. But you have to ask yourself, if your notebook can only support 2 or 4 gigs of RAM maximum (chipset and physical limitations) why would you ever need a 64 bit OS? Its highly doubtful that any "killer-app" in the foreseeable future is going to be 64 bit only unless it uses huge amounts of RAM, in which case your notebook wouldnt cut the mustard anyway. I just think running x64 on a notebook isnt practical today for 99.9% of the users out there.

On the workstation side though it is entirely a different story, especially in 3D content creation space (film and games). We routinely hit the RAM barrier both creating the scenes; high poly count, large textures and also during render time when millions of "rays" are firing around the scene ray-tracing. So for our industry 64 bit has been remarkable, and I think Intel has done a great job in educating our market of the advantages. At all the industry events I have been to recently they have done a great job explaining which parts are 64 bit and which are not and why you need or don't need 64 bit. Again, I think all the confusion lies with notebooks.

Microsoft and Intel produced this video with me showing off the differences between Softimage's 64 and 32 bit versions. If you want to see a crazy 3D scene check it out...
http://www.microsoft.com/windowsxp/64bit/videos/default.mspx
click on the second link for "revolutionizing 3D Animation." You can also check out http://www.softimage.com for more information of the software and its native 64 bit version.

cheers!

By dfumento at gmail.com on October 26th, 2006 at 5:12 am
I am completely baffled as I cannot figure out if the newest IBM/Lenovo Thinkpad notebooks X60 and T60 using the Core 2 Duo (Merom) will be able to run Vista-64 or not? Will there be some restriction because of the Napa platform/chipset or is there some sort of Napa-64 platform/chipset that enables the Core 2 Duo (Merom) to run in 64-bit so that it can run Vista-64.

There is this business week article which implies that the answer might be no, but perhaps they are not talking about the Napa-64 platform/chipset but using the Napa-32 platform/chipset?
http://www.businessweek.com/technology/content/oct2006/tc200.....technology

Because 64-bit has many more registers and especially for double precision float point as used in programs such as Matlab there are some benchmarks that run much faster in 64-bit mode versus 32-bit mode regardless as to the amount of RAM.

There are a lot of questions about this in the Thinkpad forum ( forum.thinkpads.com ).
Incidentally, Novell SLED Linux works on Thinkpads and the reason why I read is because Intel asked IBM/Lenovo to put a Linux on the Thinkpads so for chip designers who use chip design software that runs under Linux.

By David on October 27th, 2006 at 3:04 pm
I was bothered by the Business Week article in a previous post to this forum and wrote the author. This is what he wrote back:

You and Intel are both right. There was an editing error in that item that I missed; it should have said that the 945 chipset prevents Merom from taking full advantage of its 64-bit nature. Specifically, Merom can handle 64-bit instructions and has a 64-bit data path. The 945 chipset, however, can only handle 32-bit addressing, meaning that the Core 2 Duo/945 pairing is still limited to 4 GB of RAM. This will be fixed in Santa Rosa.

Not long ago, I though the 4 GB ceiling didn't mean much on anything but servers but I underestimated the memory appetite of apps, especially for image and sound processing. Apple is currently selling 3 GB Macbook Pros and will surely go above 4 GB max when Santa Rosa is available.

Steve Wildstrom Technology & You columnist
BusinessWeek
1200 G St NW
Suite 1100
Washington, DC 20005
http://www.businessweek.com/technology/wildstrom.htm

By Dan Pendergrass on October 27th, 2006 at 9:44 pm
I'm recompiling a xen kernel right now and did a bunch of hardware probing on a D820 Latitude with a Cuor 2 Duo T72000. It reported a 36 bit address register with 48 bits of virtual address. The "system" consume over half a gig when you go to 4 gigs of ram. My guess is it is using the Ram for the EMT64 addressing space (Yes 36+ 48 = 84 not 64 but they probably have error checking bits in there. They can run 64 bit OS's but they are Virtual 64 bit processors.

By modem1123 on October 28th, 2006 at 12:06 pm
Just a bit off topic but when can we expect the Kentsington quad to come out on the market and for notebooks? Is this the same arch as the core 2 duo? I am about to get the core 2 duo but thought I may wait for the quad if it's not too far in the future. Thanks.

By David on October 29th, 2006 at 1:36 am
In laptops for all practical purposes, people will be running with at most 2 GB RAM (which is what Dell recommends for Vista according to http://www.tomshardware.com ) because there are only two RAM slots and 2 GB RAM sticks (only available from Transcend) go in the roughly $400 prices range as compared to about $100 for 1 GB RAMs. Most people will elect to spend $200 for RAM and not $500 for 3 GB or $800 for 4 GB.

By mmjjrr on November 1st, 2006 at 4:22 pm
My concern is with desktop and small server systems. For me, memory addressability is critical, and I want to use systems of 4G real memory and beyond. The new processors (D and Core 2 Duo) appear to promise memory addressing of 8G and beyond. But when I put a D processor in an Intel D945GTP motherboard, I only am able to see 3.2G of memory (to the OS) when 4G is installed. The documentation indicates this to be the case where reserved memory and PCI IO mapping takes up the top 770M of the 4G space. This is actually worse than previous non PCIe Intel motherboards (like the D865PERL) where only about 300M is reserved, leaving about 3.7G for the OS.

A new Intel motherboard made for the Core 2 Duo processors (DG965RY) is advertised to accept 8G of memory. So I was hoping that when 4G was installed, we would see the total 4G available to the OS. But when I look at the technical document on the board (Order Number: D56011-001US), page 42 has a memory adress map diagram that shows exactly the same limitations below 4G as the previous (D865PERL) board. The upper 4G is not labeled as accessible to the OS. I believe a DG965RY motherboard populated with 8G of memory would only provide OS addressibility to 3.2G, which does not appear to be a practical offering for 64 bit processing. I am hoping that this impression is incorrect.

If this truly is a limitation of the DG965RY, what alternative desktop (low cost) motherboard does Intel offer that provides access to the additional addressability of the D and Core 2 Duo processors? I think this is one area where AMD and motherboard makers for the AMD have an advantage. Addressing full 4G and more memory space with AMD boards appears to be at least doable (with memory hole configurations) and encouraged (with 32 memory slot opteron servers).

By Bryan Stauffer on November 15th, 2006 at 2:15 pm
Intel has a lot of potential politicians since there are so many people who are so good at not answering a question. I confess I only had the energy to read about 2/3 of this thread after seeing the same patterns repeated and re-repeated. Only Itanium is true 64-bit, I got that. I learned C/S a dozen years ago and back then it was about CPU register size. I assume the Itanium has 64-bit registers. I also assume, since software is required to make a Core Duo processor run in '64-bit mode', that the Core Duo has 32-bit registers but since it has two cores, can operate in the mysterious '64-bit mode'. Is that about right? If not, for God's sake, will someone from Intel give a straight answer! You can't ALL be president one day! :-)

By David Schwartz on November 16th, 2006 at 8:05 am
It is erroneous to say you cannot run 64-bit code on a 32-bit operating system. It is not easy, but it can certainly be done. For example, Vmware has no problem running 64-bit code on regular 32-bit Windows XP.

Also, to reply to Bryan, the Core 2 Duo's have two cores, but that has nothing to do with whether they're running in 32-bit or 64-bit mode. Having two cores has nothing to do with things like the processor register size.

As for populating over 4GB of memory or getting every last drop when you do populate 4GB, there are chipset, BIOS, and OS limitations. Most likely, on any modern mobo that claims to support *more* than 4GB, it would be an OS limitation you are hitting. There are still many mobos that top out at 4GB (according to docs) but can't use all the memory due to chipset limitations. I know of no motherboard that claims to support *more* than 4GB that can't get all of 4GB due to chipset or BIOS limitations.

By David Schwartz on November 16th, 2006 at 8:14 am
Markchm wrote:
It looks to me as if both AMD and Intel are hiding behind the "can run a 64bit OS" story. (of course, they can run commodore 64 progs in emulation, too! LOL) EM64T/64Bit Architecture, or whatever euphemism you prefer - none have 64bit wide bus/data/IO structure. They just rely on translations via the system BIOS to allow the development of 64 bit operating systems to function on a semi-crippled chip until people can afford what everyone is referring to as REAL 64 bit throughout this thread.

-

What the heck are "translations via the system BIOS"?

The Core 2 Duo is a chip that can run 32-bit code or 64-bit code. It can run 64-bit code because it has 64-bit instructions, 64-bit internal registers, and so on. Most internal busses have been wider than 64-bits for ages because cache lines have exceeded 32-bits since before the first Pentiums.

No emulation is used or needed for a Core 2 Duo (or any other EM64T CPU) to run 64-bit code. What would they be emulating excatly?

It is theoretically possible to run 64-bit code on a 32-bit processor by using full emulation (including things like registers and instructions), but it would be extremely slow and nobody does that.

Some marketing types do try to claim that these are not "real" 64-bit CPUs because they still hope to sell us CPUs like Itanium. If they can outperform x86-64 CPUs, we'll buy them, if not we won't.

What more do you want than a CPU that can execute 64-bit instructions operating on 64-bit registers as fast as it can execute 32-bit instructions operating on 32-bit registers? You want the 32-bit mode to be slowed down somehow?

Again, *NO* translations or emulations are involved in running in 64-bit mode. They simply switch to 64-bit mode and basically the operations that used to operate on 32-bit registers now operate on 64-bit registers. (The registeres, of course, have to be 64-bits to begin with -- the upper halves are simply unused in 32-bit mode.)

By Bryan Stauffer on November 20th, 2006 at 4:49 pm
Thanks David for the last paragraph of the last post. I do not pretend to be an expert anywhere near the level of the folks at Intel or yourself. That the upper halves of the 64-bit registers are unused in 32-bit mode makes perfect sense to me, finally. Now I'll be able to sleep at night. Thanks!

By An Intel Guy on November 21st, 2006 at 6:20 am
So - I am someone from Intel. A chipset validation guy to be complete. I'm not exactly sure what I am allowed to say and what Im not allowed to say. All I can say is that core 2 duo 'supports' 64-bit with special 64-bit extension registers. there is no mysterious dual-core dual-32 bit mode or anything like that. It will support your 8GB of memory - no problem. Pentium-D can as well. So - EAX is still 32-bits. Im sure on an AMD - EAX is also 32-bits. However - there are special registers to enable the EM64T or the AMD64 mode to allow it to run in 64-bit mode with software support. The Itanium IS THE only 64bit processor. Im sure Intel wouldn't want to lie to the public saying it has 64-bit CPU's when they really aren't true 64-bit. Who says you need 64-bit anyways? Just get a quad core Kentsfield and overclock the crap out of it :) (by the way - you can access 8GB of memory with only 36 bits, you don't even need 64!)

By araoz on November 21st, 2006 at 8:22 pm
Hi. Recently a customer ask me for a 64bit PC. When I ask him what had in mind, an Intel dual core o something else, he look at me confused and told me once again that he was looking for a 64 bit processor, AKA an AMD CPU, not an Intel one.
I confess that I was unable to convince him, since Intel doesn't shows in a clean and direct way if is or is not supporting 64 bit computing. That's the point.

By EGAVGA.BGI on November 24th, 2006 at 5:13 pm
I'm looking to buy a new PC. I've been postponing it for the last 2 years because I don't want to get a 32-bit PC. I'm waiting for the 64-bit line of PCs and the 64-bit OSes, which are now available, to be released.

My question is are there any "64-bit only" processors or any "64-bit processors that supports 32-bit"? I don't want the processors to be the other way round. Extended 64 doesn't make much sense to me.

By Tech.Master on December 4th, 2006 at 11:39 pm
Intel has not been as clear about 64-bit technoogy as it could have been (from end user perspective - the tech docs are very clear), but you will not understand what it is or what it does or which processor to buy unless you are a computer technology expert and have many years of experience.

Computer sales people at electronics stores ARE NOT and never will be computer experts. Do not ask them technical questions they will get it wrong 90% of the time. You have to ask a real expert or do the hard research yourself. Do not expect simple answers anymore with so many different processor models on the market. Those of you who think Intel is hiding something are just too lazy to do the research.

FYI a REAL 64-bit CPU can only run 64-bit instructions. Most current CPU on the market can run BOTH 32-bit OS and 64-bit OS, therefore they "support" 64-bit but are not

By Tech.Master on December 4th, 2006 at 11:42 pm
(continued) ... only for 64-bit use.

And take it from a real tech expert. Buy Core 2 Duo processors. They kick ass!

By EGAVGA.BGI on December 5th, 2006 at 6:05 pm
Tech.Master, but is Core 2 Duo 64-bit only with a support for 32-bit? From what I have read and heard, it's a 32-bit processor that supports 64-bit instructions. So, not REAL 64-bit.

I'm quite happy with my Pentium 4 PC even now which is about 4 1/2 years old. I do want to buy a new Desktop or a Laptop though, which is why I'm looking for a major upgrade that can have over 4GB of RAM and supports 64-bit instruction execution natively.

By Senior Intel Architecture Designer on December 5th, 2006 at 11:13 pm
I would just like to say that my brother and I just had the same debate about whether Core 2 Duos were true 64bits, and if AMD was truer somehow. And after reading all the way through this thread not only am I still without an answer, I am the most confused I have ever been in my life. I want my last 30 minutes back.

By Richard on December 6th, 2006 at 10:15 am
I know assumptions make an ***** etc etc and are potentially risky. But based on the above trail I think some need to be made.

Intel make some great kit, it works really well and powers large chunks of the world's PC's at the least and probably fridges, cars,lawnmowers and space shuttles for all I know. The point is they obviously have many very intelligent and dedicated scientists, engineers and production experts to deliver these products to the marketplace.

I also think it is safe to ASSUME that most if not all of these people are very good at describing in plain language what they want, need and can or cannot deliver to get Intel's products to market.

So why can't anyone from Intel describe in plain language which of their processors do or don't operate in native 64bit or do or do not operate in 64bit emulation, in conjunction with X or Y operating system, connected to A or B motherboard or other peripherals? (I am sure they don't have vague disingenuous conversations about this topic within Intel, I am sure they know absolutely.)

It's probably for marketing, legal or maybe partner/vendor related reasons. Until the right time obfuscation rules. You don't want to say one thing now then have to do something different in 3-6-9 months time, we could end up as confused then as we are now.

So I'm with Tech.Master I'm buying a Core 2 Duo it's a shitload faster than my 3 year old notebook. Worst case, (as long as you have money to keep shelling out of course) in 12 or 18 months time, when the next new all singing all dancing native 64bit intel processor lands in the marketplace, if it adresses my requirements and it's shitloads faster than my Core 2 Duo, I'll probably buy it.

Signed. A Marketing Victim

By Tech to English Translator on December 7th, 2006 at 12:19 am
Ok, hopefully I won't make a major misstatement here.

The Intel64 and AMD64 chips both use compatible microcode to run the "64 bit extensions" required to run Windows XP x64 and Windows Vista64 operating systems. In fact, if memory serves, Intel has licensed the code from AMD, just as AMD has licensed much of the 32 bit code from Intel.

Yes, Virtually all of the current Intel Desktop processors and the new Core 2 Duo processors will run the Windows desktop "64" OSs.

You must check with the computer manufacturer to see how much memory a specific model will support. Many factors other than the processor come into play. Right now, very few mobile computers support more than 2GB.

My advice. Don't wait. "64 bit" is here, and it is useful. And it's not expensive. How much is your time worth? I replace my laptop at least once per year, not only to avoid problems from wear and tear, (inevitable), but also to take advantage of more memory, faster processors, and larger hard disk drives (I do a lot of digital photography). I spent less than $1300 on my most expensive laptop of the last 2 years, and less than $1000 on my current one (maxxed out at 2GB ram, btw). And I'm running Vista64 on that.

There are a variety of technical, legal, historical, and marketing reasons that Intel hasn't been very forthcoming with information, and it's hurt their market share. (Can you say Not Invented Here?) I think they recognize that, and the renaming of the extensions to Intel64 is a step in the right direction. But Intel, there is a lot of catching up to do!

By Steve Pitzel on December 8th, 2006 at 4:41 pm
Time to recap a bit:

Except for the Itanium, which is 64bit through and through, Intel and AMD have remarkably similar 64bit implementations - this is why the same 64bit and 32bit software will run on both Intel and AMD hardware. What makes this implementation really great is that it allows you to run a 64bit OS, and still run your legacy 32bit software as needed.

CoreDuo will not run a 64bit OS.
Pentium D, Core2Duo and everything released after the Pentium D (except CoreDuo) can run both 32 and 64bit OS's.
If you're looking for "TRUE" 64 bit without a 32 bit legacy, then Itanium is the processor for you - but you probably wouldn't want to buy one to play Quake 4 at home.
- Pitz

By Steve Pitzel on December 8th, 2006 at 5:15 pm
Ahh Marketing! The 64bit debate. Let's go back a couple years to June 2, 2004. The headline read, AMD64 Goes Down To The Crossroads With Eric Clapton At Guitar Festival In Dallas. The story which hit the music trades as well as the computer trades, (and was even mentioned in the Wall Street Journal) touted AMD64 technology and hinted this 64bit technology was used to record Eric Clapton live at the Crossroads concert. It was seen as a huge 64bit breakthrough for AMD - and was the flashpoint for much of the 32bit/64bit AMD-Intel debate that followed.
What most folks (and the resulting media hype) missed was that the OS used was 32bit, as was the software application used to record Eric Clapton, Steinberg's Nuendo 2.0.

By Ross on December 12th, 2006 at 7:29 pm
So where can I take my Toshiba P100 to get upgraded to a Core 2 Duo from a Core Duo, which I understood was future proof for at least two years-for free? I am running high end architectural Building Information Modelling software -Autodesk Revit Building v9.1 to be precise. One would expect that the next release would be 64 bit in such a processor and memory hungry app.

By David Schwartz on December 13th, 2006 at 2:57 pm
Wow, what a weird distortion on a set of facts.

"So - I am someone from Intel. A chipset validation guy to be complete. I'm not exactly sure what I am allowed to say and what Im not allowed to say. All I can say is that core 2 duo 'supports' 64-bit with special 64-bit extension registers."

Sure, all the CPU registers are extended to 64-bits and there's a switch to set it to 64-bit mode. The only difference between that and a "real" 64-CPU is that you have to flip a switch. The core 2 is a 64-bit CPU just as much as it is a 32-bit CPU.

"There is no mysterious dual-core dual-32 bit mode or anything like that. It will support your 8GB of memory - no problem. Pentium-D can as well. So - EAX is still 32-bits. Im sure on an AMD - EAX is also 32-bits. However - there are special registers to enable the EM64T or the AMD64 mode to allow it to run in 64-bit mode with software support."

Yes, there are other ways to use more than 4GB of memory without having 64-bits. They are inelegant and have some performance penalties, but they exist. This has nothing to do with whether the Core2 is a "real" 64-bit CPU. Hacks in *other* CPUs aren't relevant.

And yes, there are registers to enable 64-bit mode on the Core2. If a CPU isn't a real 64-bit CPU because it can also operate in 32-bit mode and needs a switch flipped to be a 64-bit CPU, then the Core 2 isn't a "real" 64-bit CPU. But that's a crazy definition.

"The Itanium IS THE only 64bit processor. Im sure Intel wouldn't want to lie to the public saying it has 64-bit CPU's when they really aren't true 64-bit. Who says you need 64-bit anyways? Just get a quad core Kentsfield and overclock the crap out of it :) (by the way - you can access 8GB of memory with only 36 bits, you don't even need 64!)

What is "true 64-bit"? If a CPU has 64-bit registers and can add, subtract, multiply, and divide them, uses 64-bit pointers, and does so with no penalty over smaller word sizes, then it's a 64-bit CPU.

I don't know what snake oil you've been buying.

Take a look at this page:
http://www.intel.com/technology/intel64/index.htm

I bet that you will be unable to define what "true 64-bit" means without resorting to technically-irrelevant nonsense.

I find the "without a 32-bit legacy" comment really, really strange. How does support for 32-bits hurt me?

You can try to argue that there's some performance penalty associated with also supporting 32-bit mode. That may even be true. But that's hardly a justification for purchasing a 64-bit CPU that performs worse. Maybe you'll feel better that the performance is slow because the design isn't as sophisticated rather than because of supporting a legacy mode.

Simply put, if you are looking for a 64-bit CPU, pick the one that provides the best price and performance for what you need. But don't buy the snake oil that there are "real" 64-bit CPUs and EM64T isn't "real" 64-bits.

If they claim the backward compatability hurts performance, let them prove it with benchmarks.

By abhijit on December 13th, 2006 at 10:38 pm
When will intel start saying "64 - bit" and not "Supports 64 bit". I think if some one really opens the processor and finds its not a true 64 bit processor, he will kick intel ass. And intel dont want this to happen, so they are strictly saying "Supports 64 bit". Even a chimp can say that "Supports 64" dosent mean "64". It just means that it will run 64 bit OS and Apps. No one from intel says that "it is 64 bit". By making some registers 64 bit, it does not make the processor 64 bit. The whole processor architecture will have to be changed. And the processor will cost a lot ! It can also be designed for backward compatibility for 32 bit applications. A true 64 bit processor can access 16 exabytes where 1 exabyte = 1,024 petabytes = 1,024 ¥ 1,024 terabytes = 1,024 ¥ 1,024 ¥ 1,024 gigabytes. i think EMT64T uses 40 bit addressing not 64. They have just tweeked 32 bit technology to make it run 64 bit applications. Applications designed for x64 run a bit faster because some registers are wider than 32bit processors. We will have to wait for some time before we get 64 bit processors on out desktops. Till then Keep running stupid x64 OS on the EMT64T processors. Listen carefully to what intel says, and you will get you answers.

By abhijit on December 13th, 2006 at 11:00 pm
Does intel do 64 bit ? This is the topic. It means "can intel processors run 64 bit applications". It is no where close to " is EMT64 true 64 bit architecture". I have even heard some application developer say that the extra registers added to 32 bit architecture are not exactly 64 bit. they are smaller. Intel is just saying "supports 64 bit". Even when people are forcing them to say that it is a 64 bit. Intel wants to play safe. AMD is planning to say (or has planned to say) "well it runs 64 bit OS and Applications"

By David Schwartz on December 14th, 2006 at 7:02 am
abhijit:

I am saying they are true 64-bit. If you want to continue to insist that they are not, please tell me what they lack.

I agree that making some registers 64-bit does not make the processor 64-bit. SSE has provided 128-bit registers for a long time and nobody argues that every SSE CPU is a 128-bit CPU. Making *everything* 64-bit makes the CPU a 64-bit CPU.

What is not 64-bit on an EMT64 CPU?

"They have just tweeked 32 bit technology to make it run 64 bit applications."

A 64-bit CPU is one that can run 64-bit applications without a speed penalty. 64-bit applications are ones designed to run on a 64-bit CPU.

"i think EMT64T uses 40 bit addressing not 64."

They support up to 48-bit virtual addressing and 40-bit physical addressing. That's enough to address 256,000 gigabytes of memory mappings per process and 1,024 GB of physical RAM. I believe no current software actually uses that many bits.

http://en.wikipedia.org/wiki/X86-64

I think there is one fair point that the virtual address translation doesn't support 64-bit addresses yet. However, 48-bits seems to be sufficient for the forseeable life of these CPUs. If they supported 64-bit address translation, it's hard to imagine anyone would use it.

"I have even heard some application developer say that the extra registers added to 32 bit architecture are not exactly 64 bit. they are smaller."

That's meaningless nonsense. If they were not 64-bits, they couldn't be used by 64-bit applications. How many bits are you claiming they are? 48-bits? 51-bits?

"All general-purpose registers (GPRs) are expanded from 32 bits to 64 bits, and all arithmetic and logical operations, memory-to-register and register-to-memory operations, etc., are now directly supported for 64-bit integers. Pushes and pops on the stack are always in eight-byte strides, and pointers are eight bytes wide."

There is nothing a "true 64-bit" CPU could or would have that these CPUs don't have that could possibly be of any conceivable use for the forseeable life of these processors.

If you think there is, state precisely what you think it is.

"well it runs 64 bit OS and Applications"

That's what, and all, 64-bit CPUs do -- they perform 64-bit operations on 64-bit registers with no speed penalty.

By David Schwartz on December 14th, 2006 at 7:08 am
No current Itanium processor supports 50-bit physical addresses. So if 64-bit physical address capability is required to be a "true 64-bit CPU", then no Itanium CPU qualifies.

By David Schwartz on December 14th, 2006 at 7:17 am
Sorry, that should be:

No current Itanium processor supports _64_-bit physical addresses.

By abhijit on December 14th, 2006 at 8:16 am
The purpose of going 64 bit is to increase the performance by 2 times, or at least improve it to some acceptable limit. Running 64 bit applications on current 32 bit processors by modifying a few registers is not the way to do it. If you just want to "show off" that you have a 64 bit CPU, you can use intel 64 and AMD64 CPUs. If you are a technical person, and know exactly why we need to move to 64 bit processing, these processors are a nightmare for you. Moving to 64-bit processors has huge advantages, which current intel 64 technology is not capabe of.
Home users currently dont require huge amounts of RAM, so i dont mind if address register is 40 or 48 bit wide ot a 1024 bit wide.
You can run a 64 bit application on a 8 bit microprocessor, with some modifications. But no one does that.
There is a difference between human and an elephant. To convert a human to an elephant so that he can eat as much as an elephant is not a good idea. You can chopp of the humans arms and legs and put elephents legs there, and say that you have created a elephant, is not a good idea!
If you respect 64 bit technology, and truly want to benefit from it, please dont accept the modified 32 bit processors as 64 bit processors.
The processor manufacturers just want to make money out of it. They just dont care. This is not the first time Processor manufacturers are fooling us. Fooling people is going on from ages.
I believe the Intel 64 gives some improvements over current 32 bit CPUs, but they should not be termed as "64 bit processors". Intel has not yet said that these are 64 bit processors. They just say "supports 64 bit".

By abhijit on December 14th, 2006 at 2:19 pm
Just have a look at the benchmark results of x86 versus x64, on same hardware.

http://www.assemble.tripod.com/emt64t.htm
OR
http://www.hardware.tripod.com/emt64t.htm

Conclusion: FAKE 64 bit !

By Steve Pitzel on December 14th, 2006 at 9:58 pm
abhijit, I hear what you're saying - I look at it in a slightly different way though. For me, the true value of an architecture is how it helps me with a particular usage. My background is Digital Content Creation - mainly animation. There was a time when we were pushing the folks at Alias (now part of Autodesk) pretty hard to come up with a Native 64bit version of Maya for the Itanium. It was a very hard sell, and rightfully so - not because Maya wouldn't have screamed on the Itanium (provided we had graphics support...another issue), but because an entire ecosystem of software would have needed to make that same move before the film|FX industry would have adopted an Itanium port of Maya into their production pipelines. Core2Duo may not give you all the performance of a platform built 64bit from the ground up - but it does give you an extremely powerful platform in terms of usability for a very wide range of software - and it's dang fast too :)
- Pitz

By David Schwartz on December 15th, 2006 at 11:24 am
abhijit:

I don't know if you're uninformed or simply lying, but what you're saying is technical double-speak.

"If you are a technical person, and know exactly why we need to move to 64 bit processing, these processors are a nightmare for you. Moving to 64-bit processors has huge advantages, which current intel 64 technology is not capabe of."

Why are they a nightmare? What are they not capable of?

Do you have any idea what you're talking about or are you just repeating something you don't understand? (That was likely being repeated by someone else who didn't understand it.)

There is nothing wrong with making a 64-bit CPU by taking a 32-bit CPU and changing everything that was 32-bits into 64-bits. How hard is that to understand?

"If you respect 64 bit technology, and truly want to benefit from it, please dont accept the modified 32 bit processors as 64 bit processors."

You are assuming what you want to prove by describing what I'm claiming are 64-bit processors as "modified 32-bit processors". What are they missing to be 64-bit processors? Be precise and specific or stop spouting FUD.

By abhijit on December 15th, 2006 at 3:42 pm
David Schwartz :

I think i am very clear about what i am saying. How come MADspitzel understands what i am saying ?
David, did you visit the links i have provided earlier ?
They show a 16.8 % improvement over 32 bit.
Intel has not changed the whole processor from 32 bit to 64 bit, it only has changed some registers. These registers are only Fraction of the CPU, it is not even part of the actual processing modules, they just hold Data. The new registers which are wider than older processors makes the processor access more data from the memory, and send more data to the processing core. Every thing else is same.
You should not decide the width of a processor with what it can process. If the processor is fully 64 bit wide, then only you should call it 64 bit. An 8 bit processor can process 64 bit code with modifications and so no one will call it a 64 bit processor.

INTEL NEVER SAID THAT THESE ARE 64 BIT PROCESSORS, THEY JUST SAY THEY ARE "64 BIT CAPABLE"

By Steve Pitzel on December 15th, 2006 at 5:58 pm
Ross! Hey, before I jump back in with David and abhijit, I just wanted to get back to you about the Toshiba P100. I've upgraded a whole bunch of desktop and workstation machines - but I have yet to crack open a laptop - that still scares the heck out of me (who knows what's in there? Could be the same white, gooey stuff you'd find inside a cockroach) I would definitely go back to Toshiba - I looked at the Toshiba website and didn't see a listing for the chipset on the P100 - it may be somewhere in the docs that came with it. If you know which one it is, look at the page below - it lists chipsets and from there you can link to the processors they support:
http://www.intel.com/products/laptop/chipsets/index.htm?iid=laptop_body+chips
- Pitz

By David Schwartz on December 15th, 2006 at 11:01 pm
"David, did you visit the links i have provided earlier ?
They show a 16.8 % improvement over 32 bit."

Yes, I did. What does the performance improvement have to do with anything? A 64-bit CPU could be much slower than a 32-bit CPU. Performance has nothing to do with whether a processor is really 64-bits or not.

Frankly, a 16.8% percent performance improvement for a jump from 32-bits to 64-bits is much more than I would have expected. Very few things don't fit in 32-bits, so the cases where 64-bits helps you much are not that common. (Encryption and image processing come to mind.)

"Intel has not changed the whole processor from 32 bit to 64 bit, it only has changed some registers. These registers are only Fraction of the CPU, it is not even part of the actual processing modules, they just hold Data. The new registers which are wider than older processors makes the processor access more data from the memory, and send more data to the processing core. Every thing else is same."

This is more FUD. Either state specifically what it is that is not 64-bits or stop claiming that somewhere something is still 32-bits. When we talk about a 32-bit CPU or a 64-bit CPU, mostly what we are talking about is the register size.

That is true though. Most of the processor was already 64-bits or wider. Intel did this because that was an efficient way to make the CPUs faster. Pretty much the only thing that was still 32-bits was the registers and the normaly arithmetic units (the FPU/SSE units were already 64-bits or wider).

"You should not decide the width of a processor with what it can process. If the processor is fully 64 bit wide, then only you should call it 64 bit. An 8 bit processor can process 64 bit code with modifications and so no one will call it a 64 bit processor."

If the core 2 is not fully 64-bit wide, then there must be something that's not 64-bits wide. Well, what is it? (Other than memory, which isn't 64-bits on the Itanium either.)

"INTEL NEVER SAID THAT THESE ARE 64 BIT PROCESSORS, THEY JUST SAY THEY ARE "64 BIT CAPABLE"

This isn't an argument about what Intel said, it's an argument about what the truth actually is. Intel could also say they are 16-bit capable, because they still support V8086 mode.

You are claiming these CPUs are not true 64-bit CPUs. So I'm asking you what is not 64-bits that you think has to be 64-bits for a CPU to be true 64-bits. You say that there are "some things" that aren't. Well, what things?

I am saying the only thing that isn't 64-bits is pointers to memory. The only thing that's 64-bits on the Itanium that isn't on the Core 2 is virtual memory. Physical memory isn't 64-bits on the Itanium.

By David Schwartz on December 15th, 2006 at 11:12 pm
I think Abhijit's core misunderstanding is that he really thinks that, all other things being equal, a 64-bit CPU has to be twice as fast as a 32-bit one. In truth, all things being equal, you would not expect a 64-bit CPU to be much faster than 32-bit CPU in applications other than encryption, image processing, or manipulation of very large data sets. And even then, many of those applications are memory-limited.

Of course, the idea that you can tell the "real" bit width of a CPU from a benchmark is absurd on its face.

"If it was a true 64 bit processor, with a true 64 bit OS, performance would have been approximately doubled !"

That's absolute complete hokum. It's even funnier because the benchmarks he's using mostly just show that the same CPU does the same number of 32-bit operations in 32-bit mode as it does 64-bit operations in 64-bit mode. Well, duh.

The move from 8-bits to 16-bits made a huge performance difference because almost everything a CPU does numbers more than 256. The move from 16-bits to 32-bits made a large performance difference because many things a CPU does number more than 65,536. However, the move from 32-bits to 64-bits won't make that much difference because not too many things number more than 4 billion.

You will not find anyone who has any idea what they're talking about claiming that they expect 64-bit CPUs to double performance. For one thing, memory won't be any faster. Code that makes decisions won't make them any faster.

By abhijit on December 16th, 2006 at 12:06 pm
David, a processor has a lots of things in it, the only things which changed in Intel 64 (EMT64T) are:
64-bit flat virtual address space
64-bit pointers
64-bit wide general purpose registers
64-bit integer support
Registers just hold instructions, results.
These extra registers are a fraction of the whole processor.

We moved from 8 bit uP to 16 bit uP, because there were lots of improvements, every thing either doubled or more than doubled, or at lease improved significantly!

We could have used the same 8 bit processor to process 16 bit code. But why did we go through all the trouble ?

Then we moved to 32 bit, do you remember the 386SX ? intel is doing it again. The difference between 286 and 386SX is similar to the difference between 32 bit an "Intel 64". Just that 386SX gave a much more performance improvement over 286 than intel 64 is giving over 32 bit.
386 could at least be called a 32 bit processor, intel dosent even call Intel 64 a 64 - bit processor !

To run 64 bit applications on 32 bit processor you will need to do a lots of work. If its SSE instruction, there wont be any problem. But if its not an SSE instrn then it will execute slower than two 32 bit instructions. Thats what shows up in the benchmarks !

Why the hell should we use the same 32 bit processor to execute 64 bit instructions ? Some Applications may run faster thanks to the extra registers provided, but some applications will suffer.
Why should we move to 64 bit, if its providing no significant improvement over 32 bit processing? We can obtain that improvement by just using using a processor at a bit higher clock.

Eating like an elephant does not make a human elephant! He will surely be different than normal human, but he will still be a human, he wont get transformed into an elephant.

You can not add parts to a human to transform him into an elephant, and you can not cut an elephant to create a human.

David, nice explanation for "move from 32 bit to 64 bit wont make that much difference". Please read some books on Microprocessors.

By David Schwartz on December 16th, 2006 at 10:33 pm
abhijit:

There is no way this could be honest misunderstanding. You are a liar, plain and simple. You have yet to point to anything that is missing from the core 2 for it to be "real 64-bits" despite numerous responses after I've asked you to point to such a thing.

"To run 64 bit applications on 32 bit processor you will need to do a lots of work. If its SSE instruction, there wont be any problem. But if its not an SSE instrn then it will execute slower than two 32 bit instructions. Thats what shows up in the benchmarks !"

Then why do the benchmarks show any performance improvement at all? By your reasoning, the benchmarks should be slower in 64-bit mode because two 32-bit operations are required to "fake" each 64-bit operation.

In reality, the benchmarks show the precise opposite -- in 64-bit mode, the CPU performs 64-bit operations at about the same rate that it performs 32-bit operations in 32-bit mode. In other words, the Core 2 is just as much a 64-bit CPU as it is a 32-bit CPU. Some improvement is seen because 64-bit operations operate on twice as much data and sometimes that helps and sometimes it doesn't.

You claim the Core 2 is not "real 64-bits". Yet you have not specified anything *SPECIFIC* to show that it's not. What's not 64-bits? You have no idea.

The benchmarks show the opposite of what you claim. They show the CPU performs 64-bit operations at the same speed as 32-bit operations. In that case, it's just as much 64-bit as 32-bit.

As for your brilliant rebuttal of my explanation for why the move from 32-bits to 64-bits won't make that much of a difference, umm, actually there wasn't one. If you think there's some kind of error in it, please explain what the error is. Otherwise, it just looks like more lying or perhaps some kind of odd faith that there must be a mistake somewhere because what you've been told be people who have no idea what they're talking about can't possibly be wrong.

As for your history lesson, it's irrelevant for two reasons. First, it conflates improvements from increases in native bit width with improvements from other changes. Obviously, a CPU from a later generation is going to be faster than a CPU from the a previous generation even at the same bit-width. The '386SX processes 16-bit code faster than the '286 does, so it's not surprising it should process 32-bit code faster.

Second, as I explained, many things CPUs deal with number more than 65,536. So being able to handle 32-bits natively speeds up a large number of processing tasks. 64-bits only helps when you manipulate things that number more than 4 billion, which isn't that much. So our previous experience with performance improvements after changes in native bit width aren't likely to be seen again.

Sorry, but that's the truth.

By abhijit on December 17th, 2006 at 1:48 pm
David, This is the last time i am talking to you.
There are a lots of units in a Processor, like ALU, FPU, AGU, etc. etc. Register are just for holding Data. To run 64 bit instructions on a 32 bit processor, processors should be modified in some way, thats what intel has done. It gives very little or no performance inprovement.

Consider a 16 bit processor:
SUPPOSE, it takes 1 Clock cycle to execute a 16 bit instruction, it will take 2 or even more to execure 32 bit instructions, and 4 or more for a 64 bit instruction.

Now consider same technology used to make a 32 bit processor:
Now the ALU, FPU, AGU, Registers are 2 times wider, so it takes 1 Clock cycle to execute a 32 bit instruction, it will take 2 or even more to execure 64 bit instructions, and 4 or more for a 128 bit instruction.

Now consider same technology used to make a 64 bit processor:
Now te ALU, FPU, AGU, Registers are 4 times wider than 16 bit, so it takes 1 Clock cycle to execute a 64 bit instruction, it will take 2 or even more to execute 128 bit instruction.

Note: Considering above senerio :
A 64 bit processor will take 1 cycle to execure a 1 bit OR 2 bit OR 4 bit OR 8 bit OR 16 bit OR 32bit OR a 64 bit instruction.
(Coz there is nothing less than 1 clock cycle)

A 32 bit processor will take 1 cycle to execure a 1 bit OR 2 bit OR 4 bit OR 8 bit OR 16 bit OR 32bit instruction. AND 2 OR MORE CYCLES FOR A 64 BIT INSTRUCTION ! THIS IS WHAT HAPPENS IN CURRENT "INTEL 64" PROCESSORS !

I think I am clear now, am i ?

If you still think i am not clear, go climb a very tall building and jump down.

If you still are alive, again climb the building and jump again!

Keep repeating it till u are alive !

(just kidding, dont take it seriously)........

By abhijit on December 17th, 2006 at 2:07 pm
David, i am not here to fight you. I just want people to know what actually is happening.
I know there are 2 ALUs in the P4s 16 bit each, which run 2 times faster than Core clock. so 2 * 16 =32: 32 * 2 = 64.
But do any thing you want, they cant execute a 64 bit instructon as fast as a single 64 bit ALU.
I know why Intel has used 2 - 16 bit ALUs instead of one 32 bit ALU.(At lease i think i know). But they could have done some improvements in a 64 bit CPU.

I dont know any thing about the Core 2 duo. Where can i get a detailed architecture of Core 2 Duo ? I can not claim them "not 64 bit" but Intel says they have not made any thing wider. They only talk about the extra registers.

But for Pentium 4 and Pentium D, i am 100 % sure they are still 32 bit processors with "64 bit support" !

By EGAVGA.BGI on December 17th, 2006 at 4:17 pm
What I understood from this argument is that abhijit wants to prove that the processors are not native 64-bit. The guy is not mentioning a critical part called the "pipelines".

David, can you tell us whether the pipelines in Core 2 Duo have been changed to 64-bit to carry in 64-bit instructions at once? Or are they still 32-bits wide.

By abhijit on December 17th, 2006 at 7:34 pm
Core 2 duo have 33% wider pipes, w.r.t AMD.
Actually, after pentium 3 - Core 2 duo should have been launched. These P4s messed up every thing.

By EGAVGA.BGI on December 17th, 2006 at 8:17 pm
I'm not familiar with AMD and with much of the newer Intel architectures (P4 onwards). Things dramatically changed I guess. But, anyway, 33% sounds like an odd number to have for a width which carries instructions, don't you think? Which again goes back to "how can it execute a 64-bit instruction at once?

By abhijit on December 18th, 2006 at 8:50 am
EGAVGA.BGI, thats what i am taking about. Current processors have to split 64 bit instructions. P4 and Pentium D has to split them into 4. Not a very nice way of executing a 64 bit instruction.
Intel and AMD are not revealing their processor architecture details, i dont know why. They just say 40% faster. They dont even specify 40% faster than what ? We just have to guess, its 40% faster than current P4s (per core).
But Intel is doing a good job, without it we would not have cheap desktops. Only rich people would be buying Computers from companies like IBM.
Sometimes Processor manufacturers get carried away and release processors with Netburst technology, but then they also release processors like Core 2.

Intel calls current processors - IA 32e. They also call it Intel 64. The also used to call them EM64T (Extended memory 64 Technology) Actually it should be called EM40T, if you Love intel very much you can call it EM48T! EM64T is way too much.

By Steve Pitzel on December 18th, 2006 at 4:21 pm
Hah! Abhijit - I loved the last comments - both about the P4 and about the cheap processors :). It's amazing how "non-linear" the path of processor design has been here. For some things the P4 was great... but it was definitely a departure. I don't see it as sinister.

Like a lot of high-tech companies, Intel is full of humans with great ideas and passion. A lot of battles are fought, won, and lost... but a lot of learning comes with the battles. Hyper-threading is a good example. Initially it was great for some things... but for real-time processes like audio recording it was best turned off in the BIOS. But then the geniuses at Cakewalk made HT work for their Sonar Digital Audio Workstation package. Then they went one better - they took what they learned from threading their package for HT and applied that to the Core2 technology. With two processors... and 64bit (and I love the debate on 64bit btw... as long as we keep it civil!) Sonar rocks!!! I'm working on a track now that choked my P4 3.2GH. It runs with no problem whatsover using the 32bit Sonar on a Core2Duo (this is a mid-level, E6600 running at 2.4GHz), and when I run that track using the native 64bit Sonar application and XP64Pro -- the Core2 barely knows I'm running anything at all -- the track screams and I barely eek out more than 12% out of either core.

You mentioned "cheap" processors. I came from an industry... Film|FX, where we routinely spent 40-50K for a good workstation. I don't even want to mention what an Inferno system used to cost. Literally, as an animation student, you would pay $75 an hour just to sit in front of an SGI Solid Impact to use a 3D package with no instructor. You sold your soul to learn 3D. Okay, well I did anyway.

Even without the greatest workstation graphics card in the world (I do have friends at both nVidia and ATI...so, hopefully they'll still speak to me at me at Siggraph next year), packages like Softimage|XSI and Autodesk's Maya (man, it's still hard for me not to say Alias) run extremely well on a machine you can build for 1K or less.

My friends at Avid put it well a few years back calling it the "Democratization of 3D."

Now -- the trick is going to be this (and I depend on folks like you, Abhijit and EGAVGA.BGI and Dave Stewart and a lot of the other passionate folks out there who love the technology and want to make sure it keeps moving forward): the trick is -- how do we keep corporations with stockholders focused on making technology not only cheaper (ah yes, stockholders love the idea of a $100 PC for the screaming masses)... but better?

How do we keep the admittedly small number of folks like content creators for film, music and games who will never have enough power and speed happy when there are a billion folks out there who'd just like to log onto Myspace once in awhile, send an email, or run the next version of Word?

Please... keep us moving forward. Keep us honest!

- Pitz

By abhijit on December 18th, 2006 at 9:05 pm
I see P4 as an act of desperation. Intel some how wanted to beat the Athlon. Or at least make the people think that intel is still ahead, and in the mean while build a uP which could really kick some amd ***. And now they have done it. Now if AMD comes out with a uP with higher processing power, then just increase the clock speed, or number of cores. I think AMD is finding it difficult to increase the clock speed of their uPs

I still dont know how AMD got the technology of beating Intel.

I remember that P4 was realesed when i was studying. i had these 2 friends they were best friends (with each other). They wanted to buy a computer. One was into gaming and other just checked mail and used messenger. I told him to buy a Celeron 1.2Ghz, he said i can afford P4, i told him that P4 is a useless processor. So he bought the celeron. The other guy i had recomended Athlon or P3. But he went for P4 (1.4 or 1.5 Ghz). The guy with P4 started teasing the guy with celeron. The guy with celeron started saying " because of you i bought the celeron" and every time he saw me, he gave me this bad look. so i finally decided to show them the truth. I downloaded a few benchmark softwares and showed them the results. The face of guy with P4 was like "I am the unhappiest human on earth" "Intel has ripped me". He bought More RAM, and then a powerful Graphic card, again he bought more RAM. He has a collection of RAM now ! Thanks to P4.

I dont know when P4 HT was released, but it has been a long time, and i am still not using even a single multithreaded application. Because, every thing i use still is sequential, uses only 1 processor. Why dont they write multi threaded programs ? Is it so difficult ? i have learnt some programming languages, languages like Java have multithreading since ages. But not even a single Java program which i use are multithreaded.
I know there are Multithreaded apps out there, but i just dont need them.

And why would i use more than 2 demanding applications at the same time ? Suppose i am playing a game, i would like the game to use both the processors. I wont play 2 games, so that both processors are used. What will people with 4 cores do ?
1st, start a DivX encoder.
2nd, Start a Game
3rd, start some thing else
4th, start anything, coz 1 processor is still empty!

I bought a Pentium D, now i am upgrading to Core 2, that means i just wasted 1 core of my pentium D.it was never used !
I hope some multithreaded apps and games come up before i get old and stop playing games.

By Steve Pitzel on December 18th, 2006 at 9:40 pm
P4 architecture was definitely about clock speed. Initially we were hearing talk of possible frequencies in the 10GHz range. The way I understand it (and I'm an animator, not an engineer da** it!) is that the long pipeline of the P4 also meant more instructions per cycle - but that also meant a heavy reliance on intelligent branch prediction - if the prediction was wrong, you had a very long pipeline to flush. As long as those predictions were right - you had great performance. That meant a lot of programming work.

With multiple cores and multithreading - you once again really have to work on the software programming side. You've probably read some of Dr. Clay Breshear's Blogs - multithreading is not easy. And - it's also not for every software application out there. Some really do benefit though.

You're right to say that in many cases folks will only use one CPU intensive program at a time. Some of those applications - the ones I use every day for animation and music creation - can really take advantage of multiple cores though. Animation software in particular can use it to take care of things like physics, Inverse Kinematics, surface deformation, texture mapping, rendering, audio and particle generation on separate threads - all of these things are meant to happen in real time - but normally most of them would need to be "turned off" to use any one of them that way. Threading those apps for multiple cores improves their performance dramatically.

If you apply the same idea to many of the popular games - you find lots of opportunities to use multiple threads and multiple cores - since games essentially do everything the animation packages do and they really have to do it in real time.

The same goes for music production. You record tracks in real time, and, if you're a singer like I am, you really like to hear some signal processing like reverb and delay happening at the same time - and if you're recording on a computer you also need analog to digital conversion - and you can't afford latency.

The usages are out there - but, like I mentioned before, multithreading is not for every application, and it ain't easy. But it can be extremely worthwhile in many cases.

- Pitz

By David Schwartz on December 18th, 2006 at 10:35 pm
Just use some common sense. If Abhijit was right, most code would be slower in 64-bit than in 32-bit mode. Simple operations that don't particularly care about bit width would take twice as long in 64-bit mode as the CPU had to do two 32-bit operations. The benchmarks *clearly* show that this is not the case.

As for the 16-bit ALUs, that's an absurd argument. If having 16-bit ALUs doesn't mean the CPU isn't a "real 32-bit CPU", how can it mean it's not a real 64-bit CPU? Intel chose 16-bits as the ALU width because a 16-bit ALU runs faster than a 32-bit or 64-bit ALU (for reasonably comparable chip real estate).

http://www.cs.virginia.edu/~skadron/cs451/itanium/Itanium.ppt
Hmm why are the EM64T CPUs listed under the 64-bit CPUs?

By abhijit on December 19th, 2006 at 9:03 am
David, we are at war now !
see, i am talking about work done per cycle. An Intel 64 Can process 2 * 32 bit instructions at the same speed as one 64 bit instruction. Not exactly at the same speed, but similar speeds.
One 64 bit instruction can do the same work as 2 * 32 bit instructions, Although this may not the case for every 64 bit instruction.
But if we want to reach the next generation of processors, it should be able to execute one 64 bit instruction in the same time as a 32 bit processor of earlier generation executes one 32 bit instruction. This is not the case now.
The 64 bit wide registers will improve the performance.
And if you look at the benchmarks carefully, the reason why 64 bit mode scored more was due to error in calculating the dot product.
In some cases the 32bit mode scores more than 64 bit mode.
As the technology improves, faster and wider ALUs can always be built.
Right now we are using the same processor which was designed to execute 32 bit instructions.
Example: You have a 512Kbps internet connection, some one gives you a 1Mbps connection at the same cost. You are happy that u will get 2 times more speed. You also have to upgrade ur old hardware (like we have to buy win XP x64 pro). But when you connect, it connects at 1Mbps, but the throughput is still the same. So whats the use ?

By abhijit on December 20th, 2006 at 7:39 pm
Ok, i abort ! be it 32 bit or 64 bit processor. I just want my things done.
Whats the use ot these discussions ?
Win XP home costs less than Win XP x64. And there are no 64 bit apps for general public.
Dualcore processors have been in the market for a while now, and still there are no multithreaded apps for general people.
I think instead of making dualcore uPs, intel should have expanded the width of the CPU, (like bus width, more L1, 32 bit ALUs, or more ALUs.. etc..) that would have helped all apps.
And now they have quadcore uPs. i know Intel has the technology but i dont know why they come up with such ideas.

By David Schwartz on December 21st, 2006 at 1:12 am
Abhijit said, "An Intel 64 Can process 2 * 32 bit instructions at the same speed as one 64 bit instruction. Not exactly at the same speed, but similar speeds."

But that's simply not true. If that were true, most benchmarks would show a significant reduction in speed in 64-bit mode. Operations that didn't require 64-bits would take twice as long, and the vast majority of operations don't require 64-bits.

"And if you look at the benchmarks carefully, the reason why 64 bit mode scored more was due to error in calculating the dot product.
In some cases the 32bit mode scores more than 64 bit mode. "

Which is exactly what you'd expect for a product that is just as much a 32-bit CPU as a 64-bit CPU. Some benchmarks show an improvement because you can calculate on twice as much data at a time. Some benchmarks show a reduction because you have to move twice as much data through the same memory interface, internal busses, and so on.

Suppose someone took a 64-bit CPU and added a feature that let it run 32-bit code. In some cases, the 32-bit code would run faster because less data would have to move around the same internal pathways. Would that somehow make the chip less "real" as a 64-bit CPU?

Basically, a CPU's bitness is the width of its general-purpose registers and the size operands its normal arithmetic operations operate on -- no more, no less.

As for there being no multithreaded applications for general use, that's a whole separate argument. But it's based on the false assumption that multiple cores are only useful if applications are multithreaded. That's mistaken for a large variety or reasons.

First, the OS is concurrent. Having two cores means that you can process a network packet or the completion of some disk I/O without interrupting the application. Think about a game that needs to write some data to the disk -- the actual disk I/O code can run in the other core.

Second, core system libraries are concurrent. Having two cores means that one core can be running Direct3D code without interrupting the application.

Third, people sometimes run more than one application at a time. Even if they really aren't meaning to, systems do lots of things "in the background" which, with only a single core, means interrupting the application. A lot of gaming stalls and stutters are due to just this.

When MMX and SSE first came out, heck when 32-bit processors first came out, everyone complained that there were no programs that took advantage of them. And that's a good argument not to pay too much extra for that technology short term. There's no point in paying more for a 64-bit CPU if it's obsolete by the time you'd want to put a 64-bit OS on it.

But that's most definitely not an argument against developing the technology. We will absolutely have to have it one day soon. We're hitting a wall with making single cores faster.

By abhijit on December 21st, 2006 at 10:25 am
David do u work for Intel ?

By abhijit on December 21st, 2006 at 12:52 pm
ok David,
i have a Pentium D 2.66. Once win XP put my HDD on PIO mode. So now every thing had to be done by the processor. I started Limewire, now 1 core was busy doing I/O and running Limweire. I thought now i just cant sit there and watch limewire for an hour. So i started Windows media player, and tried to play a movie. Guess what ? I just could not! So i Opened my task manager and could see 1 processor was being used, not 100% but like 80% Average. The second processor was empty, like "i am having sunbath dont disturb me now". Using the task manager i forced the Mediaplayer to use the 2nd Processor. Guess what, i still could not play the movie.

By media player could not play the movie i mean every thing was out of sync Sound was going ahead of the graphics.

Dont tell me that it was due to the Hard disk or RAM, over 600MB physicsl RAM was empty, and PIO mode limits the HDD to 16Mbps, and i was doing nothing to use all 16Mbps.

I was watching a movie of 700Mb and duration was 1hr 30 mins.
1hr 30 mins = 60 + 30 mins = 90mins = 5400 seconds
so the transfer rates would be = 700MB / 5400 secs = 0.12962962962962962962962962962963MBps
Limewire was downloading at 400Kbps = 50KBps = 0.05 Mbps
So the total amount of HDD i was using is = 0.179629629629629629629629629629 MBps AVERAGE.
If you consider loading the programs in the memory, it was already done.
And since sound was playing smoothly, it was already buffered in the memory along with the video.

While selling intel Pentium 4 and Pentium D i make it a point to tell them every thing that actually happens, so that they dont create problems after sale. Not a problem for mail checkers, but people who are technically strong notice small problems and ask why it is happening even when they bought multi processor(multi-core) computers.
I remember one girl who actually asked me to run benchmark softwares on her computer to prove that dual channel RAM gives some improvement over single channel RAM. Thank god that she had bought a single-core processor.

You know that i never sold a single P4 HT 3 Ghz ? I just coud not prove that the processor was faster than 2.8Ghz without HT which was selling almost at half the cost! I directly went for Pentium D. I skipped the HT generation of P4.

By Someone,No-one on December 23rd, 2006 at 8:21 pm
To both Mr. Abhijeet & Mr. david ;
The main topic has turned into a personal fight between you two.I think you both are half true.The only "pure 64-bit cpu" in the world is the ITANIUM.Both Intel and AMD uses an extra amount(eight) of 64-bit registers,64-bit flat virtual address space;64-bit pointers,64-bit wide general purpose registers, 64-bit integer support ;but not a real 64-bit processor core;only some 64-bit extensions over a normal x86 core,which can support 64-bit but not a true 64-bit.If Intel doesn't tells all lies they are bound to respect this truth that the x86 architecture, on which all 32-bit Intel and AMD CPUs are based already provides for 64-bit floating point registers (actually 80 bits internally).But the main difference between the Intel and AMD is that,AMD redesigned a processor core which Intel made only by "supercharging" the x86 architecture exsisting for several years.Even they both wotk in the same way.you both can visit http://www.pcstats.com for any detailed info on so-called "64-bit processor".
As for the matter of multicores,Intel haven't designed any such architecture that could run different apps in different cores;all are done via a single core while others are idle.Thats why AMD is going to hit the market with their second greatest innovation "Anti-Hyper Threading" ,which will divide and share workload of a single threaded application between different cores.
One request further,if anyone finds any errors or dis-consistensy in my opinion,please notify me with the correction.Thanks to everybody;specially MADspitzel for the topic.

By abhijit on December 26th, 2006 at 2:06 pm
Thank you Someone,No-one. Anti-Hyperthreading is what people really need. It can use both the processors for running one app. Its a really bad news for intel.

By David Schwartz on December 26th, 2006 at 7:41 pm
Abhijit wrote:

"i have a Pentium D 2.66. Once win XP put my HDD on PIO mode. So now every thing had to be done by the processor. I started Limewire, now 1 core was busy doing I/O and running Limweire. I thought now i just cant sit there and watch limewire for an hour. So i started Windows media player, and tried to play a movie. Guess what ? I just could not! So i Opened my task manager and could see 1 processor was being used, not 100% but like 80% Average. The second processor was empty, like "i am having sunbath dont disturb me now". Using the task manager i forced the Mediaplayer to use the 2nd Processor. Guess what, i still could not play the movie."

Wow, you cleverly proved that if you misconfigure your hardware, your computer won't work. Great job!

Someone,No-one wrote:

"but not a real 64-bit processor core;only some 64-bit extensions over a normal x86 core"

Are you Abhijit's alter-ego? This is the same nonsense argument. You are claiming that something, somewhere is not 64-bits, that only "some" things are 64-bits. Well, what's not 64-bits?

"The only "pure 64-bit cpu" in the world is the ITANIUM"

Nope, only some things are 64-bits. For example, the Itanium doesn't support 64-bit physical addresses. Notice how at least I can point to something *specific* that's not 64-bits?

And guess what, the Itanium is not a "pure 64-bit CPU" see. The Itaniums have what Intel calls "IA-32 H/W" which is hardware that executes 32-bit operations in 32-bit ALUs.

Not that I'm knocking Itanium for having a feature. Features are good. I'm just showing that the idea of a "pure 64-bit CPU" is dumb. How can having 32-bit capability *too* be bad?!

By David Schwartz on December 26th, 2006 at 9:47 pm
"As for the matter of multicores,Intel haven't designed any such architecture that could run different apps in different cores;all are done via a single core while others are idle."

Uhh, what?! Of course you can run different applications in the different cores. You can also run two threads from the same application in two different cores. (If not, what do you think the advantage of having multiple cores is?)

You are confusing the ability to run a single thread in multiple cores (which does not exist yet) with the ability to run different threads (whether from the same application or not, whether in user-space or kernel-space) in multiple cores (which is the whole purpose of having multiple cores).

"Thats why AMD is going to hit the market with their second greatest innovation "Anti-Hyper Threading" ,which will divide and share workload of a single threaded application between different cores."

I am assuming you mean "single-threaded" rather than "single threaded". That is, an application with but a single thread rather than a single application that is threaded. Threaded applications can have their workload shared by different cores already, that's one of the main reasons you make threaded applications.

A single-threaded application can only run in one core at a time. That doesn't mean the other core can't do things (like processing network and disk I/O as well as some graphics tasks) that might make the application run more smoothly on a dual-core processor.

I think dividing a single thread over multiple cores will turn out to be a losing proposition for a variety of technical reasons. But maybe AMD will think of some clever thing that I don't know about. I'll have to take a wait-and-see approach to commenting on it.

My primary rationale is that it seems to me that if there were a way to make each core faster that you could afford to do and that was practical to do in a particular product, then you should do that in each core. The reason for multiple cores is that it's not practical to make one core any faster. (If taking resources from core B would make core A faster, just make those resources part of core A in the first place.)

Perhaps being able to take one core's cache resources when that core is only minimally used might make sense. As I said, I'll have to wait and see what AMD comes up with.

By abhijit on December 26th, 2006 at 11:22 pm
What did i misconfigure ? Its an intel processor, with intel motherboard, and a 200GB SATA HDD. There was no misconfiguration what so ever ! Just proves that you are dumb.

I think Someone,No-one is talking about Apps that run on 1 core only, and not Multiple apps running on different cores, do u get that David ?
I say that a 8 bit microprocessor can execute 64 bit apps with some modifications to it, so does it become a 64 bit microprocessor ? Just answer this question.
Answer in YES or NO.

IF you say YES, all microprocessors are "infinite bit Microprocessors", coz they can execute infinite bit instructions in infinite amount of time (with some modifications). I hope your brain is good enough to understand what i am saying.
If you say NO, then Intel 64 is not a 64 bit processor.

So whats ur answer ? And please dont repeat your reasoning behind it.

And to Someone,No-one. I will need ur help.

By abhijit on December 26th, 2006 at 11:30 pm
David, whats ur educational qualification ?

By David Schwartz on December 27th, 2006 at 1:21 am
Abhijit:

"I say that a 8 bit microprocessor can execute 64 bit apps with some modifications to it, so does it become a 64 bit microprocessor ? Just answer this question.
Answer in YES or NO."

If those modifications include extending the size of the general-purpose registers to 64-bits and also include support for 64-bit instructions, then yes. When we talk about a "64-bit processor" we mean one with 64-bit general purpose registers that performs operations on 64-bit data.

If you mean something else, then for the love of god, explain *WHAT* you mean.

"IF you say YES, all microprocessors are "infinite bit Microprocessors", coz they can execute infinite bit instructions in infinite amount of time (with some modifications). I hope your brain is good enough to understand what i am saying.
If you say NO, then Intel 64 is not a 64 bit processor. "

You keep using the term "some modifications" with no explanation of what the scope of the modifications are. With sufficient changes, a Xeon becomes an Itanium. If the "modifications" include 64-bit registers and operations that manipulate 64-bits of data, then it's a 64-bit CPU. That's what the definition of a 64-bit CPU is.

You keep saying these processors are not a "real" 64-bit processor, but you refuse to do two things:

1) Define what you mean by a "64-bit processor". So this makes your claim both meaningless and impossible to refute. You aren't saying anything meaningful at all.

2) State what is missing for it to be a 64-bit processor. So there is no way I can argue that what you claim is missing isn't actually missing or doesn't affect the bitness of the processor.

"David, whats ur educational qualification ?"

If you are asking if I'm as qualified as you are to keep repeating the same meaningless drivel while refusing to answer any concrete question, no, I'm not.

This is a simple concrete question -- what is missing? Why are these not real 64-bit processors? Yes, they are 32-bit CPUs with "some changes", but a car is an elephant with "some changes". What changes are missing? What wasn't changed? The fact is that you have *no* idea.

By David Schwartz on December 27th, 2006 at 2:01 am
"What did i misconfigure ? Its an intel processor, with intel motherboard, and a 200GB SATA HDD. There was no misconfiguration what so ever ! Just proves that you are dumb."

What do you think your experiment shows? If there wasn'