Intel x86-64. **Updated 2/17** It IS AMD64 compatible...out next quarter

[sandorski]

Alkaline seems to be closest to my opinion.

I'm sure Intel has been developing and x86-64 chip, just in case. They probably would have let it die, but with AMD's success they now will go with it. So in effect, the recent announcement is somewhat of a knee-jerk reaction, but they were prepared for the possibility of x86-64 being a success.

As for Intel bringing Itanium or some other x86-64 incompatible 64bit processor to the Desktop, not going to happen. Itanium(or other non-x86) on the desktop to compete with Athlon 64 would simply be suicide, the wailing of consumers buying the Itanium after their software begins running like sh1t would send hordes to Athlon 64. Intel's mystique would be gone.

Raynor offers an interesting suggestion though, a x86-64/Itanium hybrid. This might be a good move if an OS(Windows) could allow switching between the 2 on the fly without Users noticing it, if the Itanium performs much better than x86-64, I suspect that Software developers would code for the Itanium much more agressively. I don't know if such a cpu could be made at a reasonable cost though and/or if it could switch from one mode to the other.

As for Intel introducing a non-100% compatible version of AMD's x86-64, unless Microsoft is currently adding support for such a thing, it also would be a blow to Intel if people ran into compatibility issues. No one(very few anyway) has ever forgiven Cyrix for their incompatibilities, nothing pisses people off more than running into problems that shouldn't exist.

Intel's desktop 64bit cpu will be x86-64 compatible according to AMD's design.

 

[NesuD]

Originally posted by: FishTankX

Originally posted by: arod
I guarantee you that if the doom3/HL2 benchies really show a 35% performance increase as rumored/demonstrated AMD64 will be flying off the shelf. If Intel doesnt have something comparable nobody will want to buy an intel machine except those that buy from best buy and the likes but any sertious gamer will buy 64bit after doom3/HL2.

Most games have about a 35% decrease from running 64 bit..

Umm those are 32 bit games being run on a beta 64 bit OS with beta 64 bit drivers. I wouldn't put stick my neck out just yet with that claim.

 

[acejj26]

Here's my take....tear me to shreds if you like but here goes:

K8 has been public knowledge since 1999. It has been widely known that AMD was going to introduce 64 bit extensions to the IA-32 instruction set. It has also been known that the 64 bit instructions add no more than 10% to the die size. Since Intel has cross licensing agreements with AMD, they had full access to what is now known as AMD64 and they have been working with it, maybe jointly with AMD, maybe not. However, they had invested so much money in IA-64 that they didn't want anyone to know this and so they kept it quiet. So they let AMD introduce it to test the waters. If AMD were to tank before A64 really took off (which many, many people speculated last year), no one ever had to know about Intel's plans for AMD64, and Intel could continue to push Itanium. If AMD64 proves popular, which it now seems is happening, Intel can release it around the same time Microsoft releases Windows XP 64 and they don't lose too much face/market share to AMD.

However, it remains to be seen if there are any kind of real benefits in Intel's AMD64 architecture. AMD dramatically increased the number of registers in their processor when it is being used in full 64 bit mode, which can offer substantial benefits. Did Intel do this too? Or does their processor merely support the new instruction set (kind of like how AMD has implemented SSE/SSE2 but it doesn't do it as well as Intel)?

My personal opinion (and who knows about this) - Intel's backing of AMD64 will be a MAJOR boon for AMD. Both Intel and AMD on the same page will really push software development and I think AMD's implementation of AMD64 will be superior to Intel's (again, think AMD with SSE/2). While Intel may have a Pentium 4 with 64 bit instructions, AMD will have a processor with 64 bit instructions, an integrated memory controller, and a far superior bus. AMD will then be able to tout both 32 bit and 64 bit superiority (well, 32 bit may be questionable depending on the benchmark, but I think AMD will dominate 64 bit benches).

 

[NFS4]

Originally posted by: NesuD

Originally posted by: FishTankX

Originally posted by: arod
I guarantee you that if the doom3/HL2 benchies really show a 35% performance increase as rumored/demonstrated AMD64 will be flying off the shelf. If Intel doesnt have something comparable nobody will want to buy an intel machine except those that buy from best buy and the likes but any sertious gamer will buy 64bit after doom3/HL2.

Most games have about a 35% decrease from running 64 bit..

Umm those are 32 bit games being run on a beta 64 bit OS with beta 64 bit drivers. I wouldn't put stick my neck out just yet with that claim.

I'd agree with that assessment. They were using BAD beta drivers on a beta OS.

 

[Acanthus]

Originally posted by: Hottie
My guess is it is a last min thing,
1) if Intel have a 64bit down the pipe, they won't said we don't need it until 06(that is 3yrs alway)
2) if Intel have a 64bit down the pipe, MS won't said they will only 1 version of 64bit xp for AMD.

They said 64bit for end users wasnt needed until '06. They had 64bit for servers long before AMD... This just further builds on that.

 

[Wingznut]

Originally posted by: dullard
Also I was under the impression that Microsoft said publicly that they will NOT make an OS using a different form of x86-64.

I have never seen anything from MS stating that. The closest thing we have is that Reuters article which reports that a "market analyst" says that MS said that... And "market analysts" opinions must be taken with a grain of salt.

If the situation did arise, I have a very hard time believing that MS would choose to shut out the manufacturer that holds 80%+ of the desktop cpu market. MS will go where the money is, as they have done for years. Going by the past record of how well MS and Intel have worked together, there is no reason to think they wouldn't continue to work together.

 

[buleyb]

Originally posted by: Wingznut

Originally posted by: dullard
Also I was under the impression that Microsoft said publicly that they will NOT make an OS using a different form of x86-64.

I have never seen anything from MS stating that. The closest thing we have is that Reuters article which reports that a "market analyst" says that MS said that... And "market analysts" opinions must be taken with a grain of salt.

If the situation did arise, I have a very hard time believing that MS would choose to shut out the manufacturer that holds 80%+ of the desktop cpu market. MS will go where the money is, as they have done for years. Going by the past record of how well MS and Intel have worked together, there is no reason to think they wouldn't continue to work together.

Even if this was true, it doesn't make sense for Intel to sink more money into another arch. If they go IA64, they've got loads of research and experience. If they go x86-64, they will compete head to head with AMD again, and they've proven that they can do that.

 

[thorin]

IIRC Intel already has licensing for AMD x86-64.

Originally posted by: OddTSi
Yup, Intel saw the performance improvements from the A64 about 6 months ago and they decided to build an x86-64 processor themselves. This processor should be ready for demonstration in a week or two.

They been working on it for much more then 6 months, a processor's design time is closer to 2 years, changing an instruction architecture isn't exactly something you do on a whim.

Thorin

 

[AIWGuru]

Intel's President Paul Otellini has admitted that Intel will be demoing an x86-64 chip "really soon" and the intel developers forum is "really soon."
Where HAVE you people been?
Ed Stroligo - 1/30/04


--------------------------------------------------------------------------------


There's been a raft of articles all saying that Intel is going to say something about x86-64 Real Soon Now, and the "real soon" will be at the Intel Development Forum in a little more than two weeks.

And people believe the answer will be "Yes."

If you listen to the interview with Intel's President Paul Otellini, he confirms this in a rather roundabout way.

When asked about it, he spends most of his time saying why it isn't really needed now, then concludes by saying that despite all this, Intel will have it once there is an OS and apps for it. However, if you listen closely, he is referring to x86-64 rather than Itanium when he talks about 64-bit computing on the desktop.

The way you can tell is that he says that there is no 64-bit operating system from Microsoft, outside of a beta Windows kernel, available yet. He isn't talking about Itanium because it already has a 64-bit Windows OS. That leaves x86-64 (and his description fits Windows for x86-64 perfectly).

IDF: Peering Into The Future

Intel usually doesn't introduce production processors at IDFs. Rather, they show off future processors at IDFs, usually CPUs that will see the light of day a year or a bit more later.

That would mean Tejas, not Prescott.

It doesn't look like Windows for AMD64 is going to be ready any time soon, which is another indicator against an immediate x86-64 introduction.

Reason For Delay

If Intel does announce that some form of x86-64 will be included in Tejas, that really would deliver the last blow for anyone outside of the severely hurting to buy a Prescott system.

A 2005 Tejas system would give you x86-64 AND (by that time) affordable (and faster) DDR-II RAM AND a BTX form factor AND (probably) a better (read more overclockable) CPU due to improved electrochemistry.

You'd get none of these with Prescott.

Prefer AMD? Well, if Intel jumps in, that will definitely mean software for x86-64, but by the time it becomes available, you can also get a 90nm Hammer to go along with it. Strikes me as being silly to buy a Hammer for 64-bits today just to sing "Someday, my prince will come."

"I Made The Gun That Shot Me"

Should Intel decide to say, "Me too," there will be those who will claim that this will be some great victory for AMD.

Frankly, that's the argument of a loser. It's like Saddam Hussein saying, "Look at what I did, I made the Americans invade me twice!" That's hardly a victory chant, at least not to sane people.

It doesn't kill Itanium (which finally began to sell last quarter). Itanium will continue to grow and develop on the high end, and once the fabrication processes allow it to fit the desktop, Intel will probably introduce it on the desktop three or so years from now. We'll see what happens then.

What Intel's introduction of x86-64 will do is kneecap any prospects of AMD breaking out of its second-class citizenship by removing the main potential reason for the bulk of Intel users to shift.

This doesn't mean AMD is going to die; it just precludes the possibility of great success. A moral victory doesn't mean much when the morally defeated ends up with the bulk of the sales. It's sort of like the Howard Dean advocates saying that their opponents stole their candidate's message. It's largely true, and entirely irrelevant. Neither the voting booth nor cash register counts moral wins.

 

[AIWGuru]

"Intel spokesman Robert Manetta says the Santa Clara, Calif.-based chipmaker has a working prototype of a 64-bit x86 design that it could bring to market "when customers request it."
SOURCE
This is Intelese for "we're going to end up doing this."

 

[Ardan]

After talking to a family member many times that used to be working on code for the IA64, I doubt the Itanium is going to be making this huge coup of both markets anytime soon. That is a good point, there has been 64-bit workstations and servers LONG before Intel or AMD had them. They had said that they did work on an Itanium system but gave up on it because they were taking way too long to develop the Itanium and they felt it was wasting money to try and wait for Intel to finish it. He said they have no plans to use the Itanium, ever, and that the systems they are upgrading for the US Navy's submarines are using Solaris+Sparc machines for some, and Linux+Pentium4 Xeon for the others. I've also heard from others that they, too, felt that Intel was dragging their feet on crucial parts of it. I feel that the hard facts that they took so long to develop an expensive architecture goes to show that it is crazy to think they'll do it all over again..it makes no sense. I agree that the most probable future is an x86-64 chip from them, unless they are scheming to attempt to make everyone code all over again for something ELSE. Don't tear my thoughts to shreds please, especially considering the sources have worked in this area for roughly 20-30 years and I'd take their thoughts over someone that hasn't been alive that long .

 

[AIWGuru]

"ZDNet: Any thoughts on whether Intel will go the hybrid 32/64-bit route as AMD has with Opteron?

Marengi: Intel is there. I'm not saying that they have the technology ready or that I've seen it. But for them to put that technology into the marketplace would take a nanosecond. "
SOURCE: Senior VP Dell

 

[AIWGuru]

People have assumed that it has to be either x86-64 or IA64 for Intel. That's not true at all. It can be both.

Intel has the inherent right to x86-64; x86 is their property, and AMD only can use that due to a deal made a long time back.

AMD, on the other hand, has no current right to IA-64; it only has IP rights to x86.

There's nothing to preclude Intel providing three ways to skin the cat. If you're doing a IA64/IA32 hybrid, it would be a trifling addition to make it into a IA64/IA32/x86-64 total package and make it swing all ways.

Would MS support such a thing? Sure, why wouldn't they? They have Windows for IA64 already. They're building Windows for AMD64. All they'd have to do is get the two to play well together, which would be only a little harder than getting 32-bit XP and IA64 to do so.

It would be a lot easier for Intel to provide a "universal" chip that just happens to support x86-64 rather than swallow x86-64 by itself.

Nor would Intel have to provide IA64 immediately with a modular chip, either. They could provide x86-64 today and provide IA-64 later.

If Intel did that, all the seemingly contradictory stories floating around stop being contradictory. You could buy an Intel chip with x86-64 to begin with, then upgrade to IA-64 if you chose later.

Something AMD can't offer.

I'm not saying Intel is going to do this, but if the statements in the article are true, they certainly make it possible for Intel to do this.

 

[buleyb]

Originally posted by: AIWGuru
The way you can tell is that he says that there is no 64-bit operating system from Microsoft, outside of a beta Windows kernel, available yet. He isn't talking about Itanium because it already has a 64-bit Windows OS. That leaves x86-64 (and his description fits Windows for x86-64 perfectly).

Well, although there is Windows for Itanium, it isn't a desktop OS, its the server build. I'm not saying this all doesn't make sense, but he was referring to the desktop market, and he makes it clear all the time that the two are very different, regardless of how similar they may seem to you and I.

 

[Regs]

I thought it was because of AMDs shorter pipeline and onboard memory controller that made it so spiffy?

At least on 32 bit apps as of now.

 

[Chadder007]

Originally posted by: NFS4

Originally posted by: DAPUNISHER
There are some pretty bold assertions in that article, but if he's correct then AMD will have something that will make it worth spending the $$$ on a TV advertising blitz to help garner mind share of the average consumer. I recall some of the Intel employees who post here saying that AMD and Intel always cross-license everything so they should be good to go on using AMD's 64bit tech right?

Yeah, AMD uses SSE/SSE2 so Intel should be cool with AMD64 AFAIK

AMD64 > SSE crap
But I would like to see HyperThreading come over to AMD ....

 

[FishTankX]

Originally posted by: Chadder007

Originally posted by: NFS4

Originally posted by: DAPUNISHER
There are some pretty bold assertions in that article, but if he's correct then AMD will have something that will make it worth spending the $$$ on a TV advertising blitz to help garner mind share of the average consumer. I recall some of the Intel employees who post here saying that AMD and Intel always cross-license everything so they should be good to go on using AMD's 64bit tech right?

Yeah, AMD uses SSE/SSE2 so Intel should be cool with AMD64 AFAIK

AMD64 > SSE crap
But I would like to see HyperThreading come over to AMD .... [/]

Unlikely. Athlon64 CPU's would benefit very little because their efficency is already through the roof. On the other hand, the P4 has it's double pumped double ALU's that are idle so much of the time, as well as SSE2 units and FPUs that are often waiting for a cache reload on a thread.

 

[RaynorWolfcastle]

Originally posted by: Chadder007

Originally posted by: NFS4

Originally posted by: DAPUNISHER
There are some pretty bold assertions in that article, but if he's correct then AMD will have something that will make it worth spending the $$$ on a TV advertising blitz to help garner mind share of the average consumer. I recall some of the Intel employees who post here saying that AMD and Intel always cross-license everything so they should be good to go on using AMD's 64bit tech right?

Yeah, AMD uses SSE/SSE2 so Intel should be cool with AMD64 AFAIK

AMD64 > SSE crap
But I would like to see HyperThreading come over to AMD ....

SSE/SSE2 are included in AMD64 instructions (as well as 3DNow and MMX for that matter), think before you make a statement like that next time. SSE/SSE2/MMX/3DNow are the SIMD instructions in the AMD64 instruction set, so "AMD64 > SSE crap" is like saying "Power tools > power drill crap".

Also, Hyperthreading is Simultaneous Multi-Threading, a technology that existed before Intel gave it a snazzy name; AMD doesn't need Intel licensing to use SMT, they're free to put it in when/how they like.

 

[AIWGuru]

Originally posted by: RaynorWolfcastle

Originally posted by: Chadder007

Originally posted by: NFS4

Originally posted by: DAPUNISHER
There are some pretty bold assertions in that article, but if he's correct then AMD will have something that will make it worth spending the $$$ on a TV advertising blitz to help garner mind share of the average consumer. I recall some of the Intel employees who post here saying that AMD and Intel always cross-license everything so they should be good to go on using AMD's 64bit tech right?

Yeah, AMD uses SSE/SSE2 so Intel should be cool with AMD64 AFAIK

AMD64 > SSE crap
But I would like to see HyperThreading come over to AMD ....

SSE/SSE2 are included in AMD64 instructions (as well as 3DNow and MMX for that matter), think before you make a statement like that next time. SSE/SSE2/MMX/3DNow are the SIMD instructions in the AMD64 instruction set, so "AMD64 > SSE crap" is like saying "Power tools > power drill crap".

Also, Hyperthreading is Simultaneous Multi-Threading, a technology that existed before Intel gave it a snazzy name; AMD doesn't need Intel licensing to use SMT, they're free to put it in when/how they like.

No, actually, he's completely right. SSE2 is nothing more than a SIMD instruction set. AMD64 is an extension to existing registers and several more registers of the extended legth with new instructions. SSE2 is most certainly not part of AMD64. SSE2 is simply a unit tacked somewhere into the die. Not part of AMD 64.
AMD 64 is an extension of X86. SSE, MMX, and SSE2 are not. Not at all. Big friggin' difference. If they were, then X86 would have been 128bit with the release of SSE. But it's not. Big difference. Sorry.

 

[AIWGuru]

I should put this more clearly:
Double the width and number of your GENERAL PURPOSE registers and extending your GENERAL PURPOSE instruction set and it's ability to address memory is a better method of increasing CPU performance than simply adding more SIMD instructions which have to be specifically optomized for (and even then optomized well) and only apply to large sets of data which benefit from the processing of a repeated instruction over and over on it. This is nothing compared to a general purpose boost of this calibre which affects general code.

 

[RaynorWolfcastle]

Originally posted by: AIWGuru

No, actually, he's completely right. SSE2 is nothing more than a SIMD instruction set. AMD64 is an extension to existing registers and several more registers of the extended legth with new instructions. SSE2 is most certainly not part of AMD64. SSE2 is simply a unit tacked somewhere into the die. Not part of AMD 64.
AMD 64 is an extension of X86. SSE, MMX, and SSE2 are not. Not at all. Big friggin' difference. If they were, then X86 would have been 128bit with the release of SSE. But it's not. Big difference. Sorry.

You obviously fail to make the difference between an instruction set and the hardware it runs on; they are related but not the same. Take Transmeta's CPUs for example, their native instruction set is none of the common ones, yet they run x86 just fine. There is a layer of abstraction between an instruction set and the instructions that are actually run by the CPU (heck neither Intel nor AMD actually run x86 at the lowest level in their CPUs nowadays. There is an abstraction layer that issues micro-ops for the core to execute. Implementing SSE2 doesn't necessarily mean that there is an "SSE2 [...] unit tacked somewhere into the die" (I think I remember reading that x87 FP and SSE/SSE2 share some resources on the P4, I'll see if I can dig it up).

Also, I don't see where I disagreed that adding more named registers is a good thing to do. That being said, SIMD instructions definitely still have their place in a modern instruction set as they are extremely useful for many applications. Applying FFTs is a notable example of this since a large portion of media processing relies on it. I was just pointing out that comparing general purpose instructions to SIMD instructions is comparing apples and oranges since they have different design goals (especially saying one kind is better than the other).

Finally, if you don't believe me that AMD64 instructions include SSE and SSE2, look at this document here: Text
Heck I'll even even quote the important parts for you:

Preface xiii
The AMD64 instruction set is divided into five subsets:
General-purpose instructions
System instructions
128-bit media instructions
64-bit media instructions
x87 floating-point instructions

128-bit media instructions
Instructions that use the 128-bit XMM registers. These are a
combination of the SSE and SSE2 instruction sets.

 

[AIWGuru]

Originally posted by: RaynorWolfcastle

Originally posted by: AIWGuru

No, actually, he's completely right. SSE2 is nothing more than a SIMD instruction set. AMD64 is an extension to existing registers and several more registers of the extended legth with new instructions. SSE2 is most certainly not part of AMD64. SSE2 is simply a unit tacked somewhere into the die. Not part of AMD 64.
AMD 64 is an extension of X86. SSE, MMX, and SSE2 are not. Not at all. Big friggin' difference. If they were, then X86 would have been 128bit with the release of SSE. But it's not. Big difference. Sorry.

You obviously fail to make the difference between an instruction set and the hardware it runs on; they are related but not the same. Take Transmeta's CPUs for example, their native instruction set is none of the common ones, yet they run x86 just fine. There is a layer of abstraction between an instruction set and the instructions that are actually run by the CPU (heck neither Intel nor AMD actually run x86 at the lowest level in their CPUs nowadays. There is an abstraction layer that issues micro-ops for the core to execute. Implementing SSE2 doesn't necessarily mean that there is an "SSE2 [...] unit tacked somewhere into the die" (I think I remember reading that x87 FP and SSE/SSE2 share some resources on the P4, I'll see if I can dig it up).

Also, I don't see where I disagreed that adding more named registers is a good thing to do. That being said, SIMD instructions definitely still have their place in a modern instruction set as they are extremely useful for many applications. Applying FFTs is a notable example of this since a large portion of media processing relies on it. I was just pointing out that comparing general purpose instructions to SIMD instructions is comparing apples and oranges since they have different design goals (especially saying one kind is better than the other).

Finally, if you don't believe me that AMD64 instructions include SSE and SSE2, look at this document here: Text
Heck I'll even even quote the important parts for you:

Preface xiii
The AMD64 instruction set is divided into five subsets:
General-purpose instructions
System instructions
128-bit media instructions
64-bit media instructions
x87 floating-point instructions

128-bit media instructions
Instructions that use the 128-bit XMM registers. These are a
combination of the SSE and SSE2 instruction sets.

Ah yes, you set up a straw man by making it look like I said that SSE2 doens't have any uses and then knocked it down.
Good job :gift:

Also, in the case of the info you're providing, they're using AMD64 as a marketing term. It would be better to say x86-64. You can disable the SSE2 unit and still run X86-64 code. It's not part of the X86set any more than it is of the 32bit set.

Anyway, the point is that you were contradicting someone who said that X86-64 offers a better performance increase to CPUs than SSE2. He's right. You were wrong. It will provide an increase in more scenarios and more types of data and usually a larger increase.

 

[RaynorWolfcastle]

Ah yes, you set up a straw man by making it look like I said that SSE2 doens't have any uses and then knocked it down.
Good job :gift:

Also, in the case of the info you're providing, they're using AMD64 as a marketing term. It would be better to say x86-64. You can disable the SSE2 unit and still run X86-64 code. It's not part of the X86set any more than it is of the 32bit set.

Anyway, the point is that you were contradicting someone who said that X86-64 offers a better performance increase to CPUs than SSE2. He's right. You were wrong. It will provide an increase in more scenarios and more types of data and usually a larger increase.

This is my last post regarding the subject since you're obviously not reading my posts and I'm not here to convince you.

A) When I say the AMD64 instruction set, I am actually using the correct term. Text

AMD64 replaces terms such as ?Hammer? (now ?AMD64 Platform?) and ?x86-64? (now ?AMD64 ISA?).

Markering or not, it is the correct term.

B) You can't disable SSE2 and be compliant to the AMD64 ISA. Period. It's right there in the docs I linked, black on white. Would the AMD64 ISA still be functional without SSE2? Yes, but that's not the point. Do you have any idea how many x86 instructions go unused today but have to be present for IA32 compatibility?

C) SSE2 IS part of the AMD64 ISA, see my last post for the link and B for an explanation

D) There is nothing saying that there has to be an SSE2 unit anywhere on a chip that can execute the instructions. For all the computer cares, SSE2 could be run as sequential FP ops (although that would defeat the use of even supporting SSE2).

E) Your last sentence shows that you're just not understanding what I'm writing (probably because you don't want to). AMD's 64-bit extensions aren't competition to SSE2, they are a complement to them. As such making a direct comparison between them is as nonsensical, that was my point. You even validated the point indirectly by saying that SSE2 does have a use.

I'm done debating this, I can't make my point any clearer than this.

 

[MadRat]

I am beginning to sound like a broken record when saying Intel can release a chip compatible to both AMD64 and IA64. The abstraction layer that someone was talking about earlier is exactly how Intel will demonstrate their 64-bit desktop technology. Prescott has been said to be a 32-bit processor with a 64-bit processor mode and its most likely built to include IA64-compatible functions. Using an abstract layer to translate AMD64 to Prescott-compatible micro-ops is most likely what they'll demonstrate. Even a 286 can be used to emulate a 64-bit processor if performance isn't a must. There is no reason Prescott, an IA32 design, cannot run both AMD64 and IA64 code at a decent performance level if given the right support.

 

[AIWGuru]

The speculation about IA64 posted above by me (and written by Ed) was based on information about a possible modular processor design in Tejas not prescott. Also, that president of Intel said that the processor demod would be a "prototype" which would kind of exclude Prescott. I don't know if you were aware of that. I guess now you are.