When do you think AMD will jump to 128bit?

[ToxicTaZ]

:confused: You do know who MIPS are, and what they do, don't you? :confused:

I mean, well...



...they're in the embedded market (ignoring SGI supercomputers that is). I don't know where MIPS tech appears in AMD products, I'd be *very* surprised if it's anywhere near AMD64.


AMD Adopts MIPS Technologies' 64-Bit Architecture
MIPS Technologies Licenses MIPS64™ Microprocessor Architecture to AMD for the Development of Personal Connectivity Solutions


PLEASE LOOK!! :eek:

SAN JOSE, Calif., Embedded Processor Forum, April 29, 2002 - MIPS Technologies, Inc. (Nasdaq: MIPS, MIPSB), a leading provider of industry-standard processor architectures and cores for digital consumer and business applications, and AMD (NYSE: AMD) today announced the MIPS64™ Instruction Set Architecture (ISA) has been licensed to AMD. With this license, AMD plans to develop 64-bit processors that implement the MIPS64 architecture standard in addition to its current product line of processors based on the MIPS32™ architecture standards. :eek:

MIPS Technologies is the only company in the embedded processor industry to openly license a 64-bit architecture. It is supported by widely used operating systems and embedded tool chains, as well as hundreds of applications. Such widespread support makes it a compelling solution for designers who want to quickly get to market with their next-generation products. :eek:



http://www.mips.com/content/PressRoom/PressReleases/2002-04-29


Sorry to AMD fans if this link hurts :(

And one year later... http://www.amd.com/us-en/Weblets/0,,7832_8366,00.html

[nutball]

AMD Adopts MIPS Technologies' 64-Bit Architecture
MIPS Technologies Licenses MIPS64™ Microprocessor Architecture to AMD for the Development of Personal Connectivity Solutions


PLEASE LOOK!! :eek:

<snippage>

Sorry to AMD fans if this link hurts :(

And one year later... http://www.amd.com/us-en/Weblets/0,,7832_8366,00.html

Please, you look :eek: please, understand what the press release actually said :eek:

Which bit of the following (that you didn't put into blue) did you not understand?

With this license, AMD plans to develop 64-bit processors that implement the MIPS64 architecture standard in addition to its current product line of processors based on the MIPS32™ architecture standards

MIPS32? x86? Hello?

[Filibuster]

The MIPS technology did not go into the PC processor products.

AMD64 is x86 with extentions, not anything like MIPS. They are completely different architectures. As a cpu architecture, AMD64 is closer to an Intel 386 than MIPS anything.

[moshpit]

Remi, you're MOSTLY right, but the next to last part was not at all correct. Only the Prescott P4's are 32 bit CPU's with 64 bit extensions added on. The K8 is a TRUE 64 bit CPU with full BACKWARDS compatability to 32 bit mode and was designed with 64 bit in mind from the ground up (even tho less then the FULL 64 bits isn't used in addressing memory, it's still much closer to 64 then to 32). The very EASY and obvious proof of this is visable in all the early benchmarks showing AMD64 chips getting small but noticable performance increases when running true 64 bit binaries, versus the loss of performance the P4's with EM64T demonstrate, even tho thier instruction sets are completely compatible. And I'll take AMD's word on that point since the proof seems to bear thier explanaition out a bit better then it does for yours.

The P4 and the K8 are truley not comparable in a proper apples to apples comparison considering the P4 is still only a 7th generation x86 CPU versus the K8 being a TRUE 8th generation x86 from the get-go.

But your other explanaition on why 128 bits are useless was completely on target. Even the most demanding of applications will barely push a 64 bit CPU for processing bandwidth and will instead be hitting other bottlenecks in the system long before pushing the CPU's memory addressing limits and internal bandwidth capabilities.

[moshpit]

:confused: You do know who MIPS are, and what they do, don't you? :confused:

I mean, well...



...they're in the embedded market (ignoring SGI supercomputers that is). I don't know where MIPS tech appears in AMD products, I'd be *very* surprised if it's anywhere near AMD64.

Your completely correct in that. The 64 bit MIPS licensing being refered to is used by AMD in some EMBEDDED products and none of that tech is locatable anywhere within AMD64's instruction set or architecture. I DARE one of ya'll to find an ACTUAL link to AMD64 and MIPS PLEASE, I DARE YOU, hehehehehe. (and I DO mean one that isn't some April Fools day joke...)

[|JuiceZ|]

(popcorn)

[ToxicTaZ]

Your completely correct in that. The 64 bit MIPS licensing being refered to is used by AMD in some EMBEDDED products and none of that tech is locatable anywhere within AMD64's instruction set or architecture. I DARE one of ya'll to find an ACTUAL link to AMD64 and MIPS PLEASE, I DARE YOU, hehehehehe. (and I DO mean one that isn't some April Fools day joke...) :thumbdwn:

:rolleyes:

"The momentum for 64-bit embedded processors is increasing, and AMD's licensing of the MIPS64 architecture is evidence of that. In the past nine months alone, nearly half of our licensing agreements have been for 64-bit architecture and cores," said John Bourgoin, chairman and CEO of MIPS Technologies. "We are particularly pleased that AMD, with its many years of experience in embedded markets and in microprocessor development, recognizes the value that 64-bit technology is bringing to these markets."

SAN JOSE, Calif., Embedded Processor Forum, April 29, 2002

http://www.mips.com/content/PressRoom/PressReleases/2002-04-29

[Remi]

Moshpit: >Remi, you're MOSTLY right,

Thanks!

>but the next to last part was not at all correct.

Huh?

>Only the Prescott P4's are 32 bit CPU's with 64 bit extensions added on. The K8 is a TRUE 64 bit CPU with full BACKWARDS compatability to 32 bit mode and was designed with 64 bit in mind from the ground up (even tho less then the FULL 64 bits isn't used in addressing memory, it's still much closer to 64 then to 32).

.... (/me scratching his head)... You're the first one I hear saying that... I mean, other than the usual PR statements, of course! :p Given that nobody sane programs in assembly language those days, I guess nobody I've chatted with of those procs really knows what's his talking about (and that includes me), so I decided to take a closer look. I think there's little debate about the fact that the number of bits of a processor is precisely the number of bits of its registers and its ALU(s), which should be easy to check from the technical reference manuals.

Ok, so 64 bits registers... let's see... Oh, they do have those... Good! Now the real test... If they have 64 bits, they must have ADD/MUL/DIV/XOR etc. instructions operating on 64 bits registers... ...which... ...they do!!! (source here (http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/24594.pdf) - 4.6MB PDF). Cool! :)

Now, if only they would have included clock cycles in their reference manuals, we could have checked if the ALUs are really 64 bits, or if they simply had packed 64 bits arithmetic and logic instructions in their instruction set which are translated into microcode performing that computation onto 32 bits ALUs - which looks perfectly doable. As I know I'm a bit suspicious there, I'll give them the benefit of the doubt (until, of course, I find a 64-bits assembly and enough time to do a few tests to check that).

>...The P4 and the K8 are truley not comparable in a proper apples to apples comparison ...

Definitely, they're technicaly very different beasts... but we can't avoid comparing them, because they're competing on the same markets...

So it looks like we do have a x86 64-bits processor here after all...

Hey, the more the merrier! :p

Now, all we need an OS which will use it in true 64 bits mode... ...and all our favorite software adapted to it, of course! :D

[nutball]

Now, all we need an OS which will use it in true 64 bits mode... ...and all our favorite software adapted to it, of course! :D

Linux? :p

[moshpit]

:rolleyes:

"The momentum for 64-bit embedded processors is increasing, and AMD's licensing of the MIPS64 architecture is evidence of that. In the past nine months alone, nearly half of our licensing agreements have been for 64-bit architecture and cores," said John Bourgoin, chairman and CEO of MIPS Technologies. "We are particularly pleased that AMD, with its many years of experience in embedded markets and in microprocessor development, recognizes the value that 64-bit technology is bringing to these markets."

SAN JOSE, Calif., Embedded Processor Forum, April 29, 2002

http://www.mips.com/content/PressRoom/PressReleases/2002-04-29


Nope, sorry, that doesn't connect AMD's x86 CPU's to the embedded MIPS architecture at all. All that does is support what I'm saying that AMD is running 2 seperate 64 bit architechtures at the same time for differing markets. You've proven my point, not denied it at all. Nowhere in that press statement does it show a connection to AMD64 and MIPS64. All it says is that AMD is doing good in the embedded market thanks to thier licensing of MIPS, that's it.

And to Remi: You sir, are a minch! (a minch is a good thing!) Thank you for taking my comments in stride and not being insulted by them and instead taking them as a spur to do further research and find out for yourself. You REALLY are the kind of enthusiast who will learn alot in this industry, THAT I have no doubt of. Your quickness to check what I was saying and admittion that I was probably correct based on the evidence publicly available impressed the hell outta me. I guess I'm too used to trying to point reason out to fools and having them bite my hand for it. So, once again, thank you for listening objectively and then forming your own opinion based on new information.

Marketing terms are indeed often an attempt to fool the public into believing something that isn't true, but SOMETIMES, if you read between the lines of the marketing crap they feed us, you can find jewels of truth in there that lead to further research on the subject that shows the REAL merit of said product. In this case, AMD's claims to 64 bit-ness are actually true, and, for the record, there's NO connection to thier MIPS32 and MIPS64 lines of embedded CPU's to thier line of x86 CPU's, 32 bit OR 64 bit...

edit: Oh yeah, one last thing. PLEASE O PLEASE, Remi, make the time to perform that assembly language test you suggested to determine internal processing bandwidth! That is a test I've been wondering about for SOME serious time. I do understand that AMD could claim 64 bit and still be doing some kind of internal translation, and have been curious to know about that little question myself. Indeed, all the public information points to a fully 64 bit CPU, but your question of the ALU's internal functionality is simpy one I have no answer for and would LOVE the answer to ;)

[moshpit]

At the risk of endangering myself to ridicule for double posting instead of just editing, I felt this part was worthy of it's own post instead of adding to the already long last post.

AMD's processor sales are not just targeting the mobile/desktop/server x86 market like you seem to think, Taz. That article you point to is referencing the Alchemy line of MIPS based CPU's that AMD makes for hand-held and other embedded designs that will NOT be running an x86 based OS. The Alchemy line actually DOES consist of 2 seperate lines also tho. There is a MIPS based line based on lower clocked, low power consumption processors that are meant for very advanced cell phones, hand held PDA's, and other specialized jobs. The other line of Alchemy based CPU's has more architectual similarity to the K6 line of CPU's and is another NON-MIPS based CPU with slightly higher clock speeds and power consumption, but also slightly higher performance for larger devices that may require Window CE or other light x86 OS duties. Both lines of Alchemy CPU's are targeted at slightly differing markets, but have NO cross-over into AMD's x86 lines of CPUs. Opteron, Athlon 64, Turion, and Sempron are all AMD designed (with some help from IBM and a few ex DEC Alpha designers) with no external licensing required to make them(altho I think SOI does have some cross-licensing between IBM and AMD going on there). Only part of the Alchemy line has a licensing requirement due to it's MIPS based heritage. The 64 bit related quote from that link made NO reference to the x86 market what-so-ever and it's only 64 bit reference was to the MIPS based CPU cores only (considering MIPS corp wouldn't mention a competing standard like x86 to begin with, what with x86 Alchemy's making inroads in the embedded market and AMD begining to position the newer Alchemys as a superior product for perfromance).

Please get this equation strait : MIPS does not = x86 and the two are mutually exclusive of each other with VERY different requirements. There is NO MIPS based tech or licensing in AMD64, at all. Period, end of story.

[Rakeesh]

Hmm...I just noticed something. It looks like theres a trend in that every year and a half the amount of needed memory for games seems to double. In about 3 years from now its probably safe to say that we'll be playing games that require 4GB of ram.

Since other devices in your system need a physical memory address to function, you wouldn't be able to utilize the full 4 gigabytes in your system, therefore you'd presumably need more memory than that. Only problem is 32-bit CPUs wont be able to address memory higher than that in one clock cycle, thus there'd be a significant slowdown when accessing memory in that range. (you could probably pull this off, it would require some kernel tricks and the OS would have the additional burden of tracking which memory belongs to which process without losing things)

64-bit CPUs on the other hand can address up to 16EB (ExoBytes, 2^64/2^60=2^4=16) of memory in one clock cycle. So whoever says that 64-bit CPUs wont be needed for a while now is a bit wrong :D

[moshpit]

You right on target there, AlphaWolf. But AMD64 and EM64T CPU's have nowhere near the full 16 exobytes available to them because the full memory range of a 64 bit CPU won't be even a consideration for maybe as much as 10 years on the desktop given the last 10 years rate of development to compare against. I believe they both used a 40 bit addressing scheme, but am unsure of that totally because I remember hearing somewhere that Intel even short-changed on that and went more with a 38 bit pseudo-40 bit addressing setup.

Just historically, look at the average PC of 10 years ago and you see 4mb ram is considered spacy and almost overkill. Now we're seeing 512mb and 1Gb as the standard, and 4Gb is considered that same kinda overkill for a desktop. Looking at that trend alone, and factoring that nothing major shakes the computer industry up and suddenly requires an exponential increase of ram amounts suddenly and out of the blue, then we'll possibly be seeing 512Gb to 1 Tb become standard in desktops with 4Tb being overkill (not to mention somewhere along the way a SLIGHT widening of the memory address bit-width to maybe 44 bit or something like that).

That's all just an educated guess tho, based on historical reference...

[Rakeesh]

40-bit addressing should probably be fine for now. I am sure that can be extended within the next 10 years or so when it becomes necessary. In fact, I'd bet we will probably see a change in most chipset architectures that would accomodate a 64-bit addressing just for hard disks that grow above the 1TB range.

[moshpit]

Good point! We cannot forget our beloved (sarcasm) Mechanical Storage Devices...

If we don't increase thier addressing too, they may begin to feel unloved. And that would never do ;)

[Filibuster]

The CPU doesn't limit hard disk capacities now, if it did the largest hard disk would be 4GB. ATA5/6 use 48bit addressing currently...theoretically a maximum 144 petabytes. (source:WDC)

I can't believe there is still discussion about where AMD64 came from. Its still x86, they've been using it for 20 years, just the last in a long series of added features.

[moshpit]

No, u missed the point. We had switched to discussing the drive controllers keeping up with CPUs, that's all. And of course we're arguing where AMD64 came from! Some silly dolt tried to say AMD licensed it from MIPS which is complete malarky. x86 was designed by IBM, stolen by Intel and made thier own thru patent infringment because IBM couldn't defend themselves legally at the time due to a anti-monopoly ruling they were operating under. AMD won several lawsuits years later against Intel and forced the right to be able to make x86 CPUs. And AMD extended the 32 bit ISA to 64 bit, all on thier own. AMD gets full credit for 64 bit x86. Period. And it's BLARINGLY funny that IBM is backing AMD so heavily now. Looks like the thieves at Intel shouldn't have been so quick to bite the hand that was feeding them...

(if that last sentance wasn't flame bait, I don know what is ;) )

[nutball]

I can't believe there is still discussion about where AMD64 came from. Its still x86, they've been using it for 20 years, just the last in a long series of added features.

There isn't a discussion, there's an Intel-fan-boy trying spread some fantasy about AMD pinching tech from MIPS, and then there's the rest of us. :D

[moshpit]

Watch it, Nutball. I think Filibuster is trying to say that Intel invented x86 and therefore gets the credit for x86-64. I'm hoping I'm wrong in thinking that, but I'm VERY used to people giving Intel all the credit for x86, when all Intel did was rip it off themselves (and indeed improved it up to 32 bit). AMD reverse engineered it and ripped it off from Intel in return, and then made it better with 64 bit. Intel gets NO credit for 64 bit x86, they tried to abandon us x86 users for the next gen with Itanium and at one time was preaching that tech would be the desktops eventual way of doing things.

The ONLY reason they (Intel) release EM64T is because they found the Intanium sucked ass with anthing that required backwards compatability and customers wouldn't let em get away with that crap this time. AMD was already on the ball to extend the life expectancy of our software by making sure that the move to 64 bits didn't break everything we use daily. So, AMD wins the credit for the proper move to 64 bit. Sorry Intel fan-boys, your Hero tried to screw us all over and got burned for it. YAY for the underdog!

[Viral]

:rolleyes:

"The momentum for 64-bit embedded processors is increasing, and AMD's licensing of the MIPS64 architecture is evidence of that. In the past nine months alone, nearly half of our licensing agreements have been for 64-bit architecture and cores," said John Bourgoin, chairman and CEO of MIPS Technologies. "We are particularly pleased that AMD, with its many years of experience in embedded markets and in microprocessor development, recognizes the value that 64-bit technology is bringing to these markets."

SAN JOSE, Calif., Embedded Processor Forum, April 29, 2002

http://www.mips.com/content/PressRoom/PressReleases/2002-04-29

Someone's probably already corrected you, but this article is referring to embedded processors, not x86 desktop CPU's. AMD obviously licensed the architecture for their 64-bit embedded technology.

But obviously you don't know the difference here and were just happy find something to rag on AMD with. Because afterall, Intel has been implementing pretty close to all of AMD64's new marketable features into P4 just to keep up. Gee, how unoriginal of them.

[Rakeesh]

The CPU doesn't limit hard disk capacities now, if it did the largest hard disk would be 4GB.

Aside from what moshpit already added, the main reason why that isn't very limiting for the CPU is because the hard disks are so slow that you can actually run through several hundred clock cycles before the hard disk manages to perform the write anyways. That would be very taxing for RAM on the other hand.

[moshpit]

Someone's probably already corrected you, but this article is referring to embedded processors, not x86 desktop CPU's. AMD obviously licensed the architecture for their 64-bit embedded technology.

But obviously you don't know the difference here and were just happy find something to rag on AMD with. Because afterall, Intel has been implementing pretty close to all of AMD64's new marketable features into P4 just to keep up. Gee, how unoriginal of them.

Hehehhe, obviously somebody who doesn't like my long writing style. Couldn't sit thru my long version of saying exactly that with nearly the same amount of sarcasm, but nicely spread across half a page, could ya? lol. I understand, I can barely stand reading my long winded stuff either ;)

[Viral]

lol, nah it wasn't that. I just decided to make a reply then a there without reading the rest of the replies. I looked a few posts down and didn't see anything, so I went ahead. I then read your post after :P

FWIW, you went into much more depth than I ever could. I don't know anything about embedded processors, but I could tell what the article was about at least.

[moshpit]

HAHAHAHA, Niether do i!!! ROFL! But, like you, I could EASILY see the stretch that the poster was trying to make between embedded MIPS CPU's and x86 desktop/mobile/server chips. But then I jokingly dared anybody to try to make a connection between MIPS and x86 and reposted the same damn thing again! I think the whole 64bit concept is what's throwing him off tho. He automaticly assumed that AMD has only ONE 64bit direction they're taking and so thought maybe this related. Unfortunately for him, he had to make such a big deal of it on the first page that he inspired my long winded counter to shut him down from his spreading of false information like he was somehow in the know and was educating us. THAT pissed me off enough to eat his lunch ;)

I MAY not know much about embedded CPUs, but I know a TON about AMD64 and it's roots. MIPS has NO connection to it. IBM, a few ex-DEC Alpa engineers, and AMD engineers spent a TON of work spanning 6 years since the first time I read about the concept of x86-64 to make it happen, even with Intel espoucing Itanium as the real 64 bit alternative.

[moshpit]

Damn, call me the thread killer :(

Even though I was purposely inflamitory, looking for a decent debate, it seems I've scared off my potential debator. Damn. That's no fun. Was hoping for a little Intel fanboi chompin and instead scared him off, heheheheh.