Nvidia announces x86 chip *edit: not true*

[Nemesis 1]

Originally posted by: apoppin

Nvidia licensed LongRun2 from Transmeta (power management), not core CPU architecture licenses/code morphing.

Not so, SunnyD

http://news.cnet.com/nanotech/?keyword=Transmeta
http://www.deviceguru.com/transmeta-seeks-buyer/

In August of this year, video-chip powerhouse Nvidia paid Transmeta $25 million for a ?non-exclusive license to Transmeta?s Long Run and LongRun2 technologies and other intellectual property? for incorporation into future Nvidia chips.

and worth reading:
http://www.computerworlduk.com...ndex.cfm?articleid=225

According to the complaint, Intel and Transmeta were working together until a dustup over the value of Transmeta?s IP (intellectual property). Transmeta says that Intel folded up its chequebook but kept using Transmeta IP in Intel designs. Having just declared itself the new papa of the green x86, it looks bad that Intel might have let a little of Transmeta?s 1-watt 32-bit x86 leak into Intel?s performance-per-watt chips.
.... In 1995, Transmeta set out to create a metaprocessor, a CPU that could assume the personality of another. Transmeta first created Crusoe, a uniquely flexible CPU with a native VLIW (very long instruction word) architecture. Itanium is another VLIW design, but as opposed to Intel, Transmeta never required developers to code to its CPUs? native architectures. Instead, Transmeta wrote Code Morphing software to translate x86 instructions into native VLIW operations on the fly. Any 32-bit x86 software you choose runs, unmodified, on a Transmeta CPU. Translated code is cached, so Transmeta processors -- the current being Efficeon -- speed up as they learn the instruction mix of your applications.

The Code Morphing software not only translates x86 code to VLIW in real time; it analyses the code it?s translating and makes fine-grained adjustments to CPU voltage and clock frequency based on performance demands and thermal conditions. It?s key that Efficeon doesn?t rely on the OS to measure load and change speed and voltage. Efficeon and Code Morphing measure and adjust to demand by themselves. Transmeta calls this LongRun, and LongRun2 pushes power-saving technology further by greatly reducing the amount of current that transistors leak while they?re in the ?off? state.

Nvidia has this now

Now do you see why intel is panicking and suing Nvidia .. and Nvidia says, "screw you" we can do better than you can with our own CPU; - 1-3 years time frame
- it is a huge slap in intel's face and designed to throw them completely off balance
- imo .. of course, it is would i would do if i was gambling

I really don't see the point here. Intels has full access to all of transmeta patents. AL of them .

http://www.macworld.com/articl...2007/10/transmeta.html

 

[waffleironhead]

Originally posted by: Nemesis 1


I really don't see the point here. Intels has full access to all of transmeta patents. AL of them .

http://www.macworld.com/articl...2007/10/transmeta.html

While that may be true. Intel does not have access to nvidia's patents. So whom do you think can benefit better from the transmeta ip, intel or nvidia?

 

[Nemesis 1]

Thats easy question. Intel . Why you might ask . Thats easy EpiC. . Its also interesting AMD also did some deals with transmeta. I find that as interesting s Intel . Same reasons. ATI = EPIC . Transmeta = VLIW(EPIC) morphing x86. It is Intel and AMD that will benefit the most from transmeta. AMD is along ways from being beaten. End of 09 should be interesting. The Big differance in all of this is Intel buying Elbrus. Now Boris is a Fellow at Intel . In charge of Compilar division. AVX should prove very intersting in tying together all the lose ends.

 

[Nemesis 1]

Here is Info on IA-64. Compilar used in Itanic. For now lets just discuss. Loop fusion for larrabee and Sandy.

Since this is most relavant to the discussion.

Loop fusion in the Intel IA-64 compiler is more aggressive
than that in compilers for IA-32 or RISC processors,
Bottom page 4 top page 5






http://download.intel.com/tech...41999/pdf/compiler.pdf

 

[Nemesis 1]

Than we can move on to this .


PARALLELIZATION AND
VECTORIZATION
Support for OpenMP*, automatic parallelization,
vectorization, and load-pair optimization are all included in
the design of the IA-64 compiler. The design takes
advantage of native support for parallelism on the IA-64,
which includes semaphore instructions such as exchange,
compare-and-exchange, and fetch-and-add, in addition to
the fused multiply accumulate instruction (fma). The
support for parallelism on IA-64 also includes SIMD, i.e.,
parallel arithmetic operations on 1, 2, and 4 bytes of data.
In order to exploit the fine grain locality of data access in
applications, IA-64 provides load instructions that
simultaneously load a pair of double floating-point
precision data items.
You will see all of this in AVX. FMA comes at the 22nm shrink. Intel is really taken along road to get to AVX. But it should pay dividends.

( Idon't care) the bolded part above . Larrabee has 256bit vectors as does Sandy as you can see the IA-64 can do a pair of double floating-point
precision data items

 

[Nemesis 1]

Loop Fusion
Loop fusion combines adjacent conforming nested loops
into a single nested loop [16]. Loop fusion is effective in
improving cache performance, since it combines the cache
context of multiple loops into a single new loop. Thus,
data reuse across nested loops is within the same new
nested loop. It also increases opportunities for reducing
the overhead of array references by replacing them with
references to compiler-generated scalar variables. Loop
fusion also improves the effectiveness of data prefetching.
Loop fusion in the Intel IA-64 compiler is more aggressive
than that in compilers for IA-32 or RISC processors, for
Intel Technology Journal Q4, 1999
An Overview of the Intel IA-64 Compiler 6
example, since loop fusion in the IA-64 takes advantage of
a large number of available registers. In the loop on the
right-hand side of Figure 9, cache locality is improved
because the accesses to array a are reused within the same
loop. Further, it enables the compiler to replace references
to arrays a and d with references to compiler-generated
scalar variables.
So what does IA-64 have to do with larrabbee or Sandy . Not much other than its a perfect fit. How intel pulls this off . Not sure . But I am sure they will pull it off. Larrabee /Nehalem should be great . Sandy /larrabee takes us to new places. Thats in the embedded market . All the way to the top processors. AVX is the future.


I while back Idon't care. I posted a article on what went wrong with transmeta. Had transmedia used more cache and vectorization . In that article . It was discribed how the elbrus compiler would of transferred Transmeta to a super power in CPU design. It would have given intel fits. As it was Crusio died on the island and there was no companion.

If you followed that article . Which you did read. You can see with Intels latest cpus . They are on the exact road that article described . Its almost a perfect fit . Other than I don't know were Epic and Intel stand. We do know were AMD and EPIC stand however. I see intel and Amd going the same direction but differant paths. I won't bet who is on the best path. Both look promising.

 

[chronodekar]

Coming back to the earlier discussion of AVX, (nice pdf link Nemesis!) it seems to me, that AVX will be another instruction set about to be released by Intel. By micro-architecture, intel refers to the actual implementation on silicon. From a software view-point that's irrelevant. As a programmer, we see the processor as a black-box. i,e. we send instructions into it and expect output. A point made in the pdf is that AVX will be fully compatible with ISA. So that means that if we send in ISA based instructions to an AVX chip, it will respond as we normally expect.

On the matter of recompiling, Intel is saying that many of the ISA-based instructions can be replaced instruction-for-instruction with AVX ones. The key point to note here, is that it requires an additional step from a software developer. In simple english, "If you just recompile your program with this new compiler of ours(from Intel), you will be able to increase program efficiency by xx.xx%!!" What is not being mentioned is that you will end up being tied down to Intel chipsets.(or at the very least, Intel licenses, if someone-else uses it too)

If Intel manages to cover 95+% of the desktop market-place with AVX chips by 2011, then Nemesis1, your prediction will be fulfilled. i.e. AVX will have replaced ISA, and compilers WILL have to optimize for it.

BUT, if nVidia (or AMD) comes out with an alternative, which is at least, of equal if not superior performance (and we have seen both companies are capable of surpassing Intel's performance benchmarks) AND if they can cover about 20% of the market-place, then AVX will be seriously challenged. In that kind of a scenario (Marketplace: Intel 80% nVidia/AMD 20%), I'd say a programmer/developer will have to be very concerned about getting tied down to a particular license. There will be a push to converge on a "standard" of some sorts. From the above comments, I'd say that Transmeta (with it's code-morphing) could very well be the answer.

Now the above is, of course, highly speculative, and things can turn out very differently.

Addressing a different issue, some might not be convinced that having a 20% market-share is enough to tip the scales in your favour. I agree. But it's enough to force the industry to converge on a common standard. Just look at the web-browser scenario. IE (all of it's variants) have about 60-70% market share. Firefox has some 25%, with the rest going to Safari and others. Before Firefox, Microsoft practically dictated what web-standards were. Now, thanks to the rise of the Fox, every major web-designer is demanding for a common standard. And the result? IE8 is going to be more standards-complaint (With the IE6 rendering engine about to be dumped).

Mark my words, for Intel, what will matter in the end, is how much of the market they will be able to grab before nVidia arrives on the scene.

 

[Nemesis 1]

C+C common enough.

 

[apoppin]

Originally posted by: Nemesis 1
Thats easy question. Intel . Why you might ask . Thats easy EpiC. . Its also interesting AMD also did some deals with transmeta. I find that as interesting s Intel . Same reasons. ATI = EPIC . Transmeta = VLIW(EPIC) morphing x86. It is Intel and AMD that will benefit the most from transmeta. AMD is along ways from being beaten. End of 09 should be interesting. The Big differance in all of this is Intel buying Elbrus. Now Boris is a Fellow at Intel . In charge of Compilar division. AVX should prove very intersting in tying together all the lose ends.

then why is intel afraid of Nvidia and what they are proposing to do with x86 ?

i think Nvidia is going to use the transmeta patents to much greater effect .. it is clear that they have been planning this for a long time

 

[Nemesis 1]

Why do you assume Intel is afraid of Nv . I think you got it all wrong fella.

Intel is afraid of no one right now. They have a set path. Why did AMD buy ATI . Why is NV bad mouthing Intel You guys all act like Intel started the larrabee project after AMD bought ATI. NO friend not true. AMD bought ATI after they found out about project larrabee. Intel had many projects going . Larrabee was the winner. Don't fool yourself. Into thinking Intel is reacting to anything NV or AMD/ATI has planned because its just not so. They are reacting to Intels plans .

I have seen it to many times. Intel intros Metal / highK gates. IBM/AMD announce they to have it . But theirs is better. Where is it. Its been over a year.

NV . making all kinds of noise about getting programmers to code for cuda.

Na. We see things differantly you and I . Its not about whos right whos wrong . But finding the true facts in all the mumble jumble. Thats relavent. Its like trying to hone the senses. Because fact and fiction in todays world is getting harder to discern. Keeping ones senses honed is paramont to success of enlightenment.

One might ask one self . Why did Intel choose Larrabee. Ok intel already has Nehalem in the works . and the team in Israel is working on sandy.

So ya look at nehalem pretty good processor 8 thread capable . That combined with the Larrabee is alot of threads in flight. But its not really a fit ya would look for to tie X86 and Epic together. Intel bought Elbrus in 04. 2 years latter release C2D than NehalemC2D with HT. ondie memory controler and QPI. Its getting more interesting. Vastly improved compiler that you well see at work with larrabee.

Than comes sandy and AVX Now Intels set path is clear to all when its announced.

Why does X86 have to be ported to Sandy AVX? I would think real hard on this one.

As good as Intels Sandy AVX will be. One hugh cloud hangs over it . I am not happy about intels decision to delay FMA until 22nm. To me thats a sales plow. TO transparent it is.

 

[Denithor]

God, I love reading this stuff you guys post.

Fcking brilliant people posting around here (even if the grammar sucks at times).

:beer:

 

[apoppin]

Why do you assume Intel is afraid of Nv . I think you got it all wrong fella.

i am not assuming .. intel is reacting .. they are *suing* Nvidia; i know a preemptive strike when i see one
- one that Nvidia has coldly calculated as a risk.

i believe you are overawed by intel's smoke and mirrors

you were all about itanium and netbust also

 

[taltamir]

i want to know... what do you mean by "coldly calculated as a risk"... thats some negative connotations there...

 

[Nemesis 1]

Originally posted by: apoppin

Why do you assume Intel is afraid of Nv . I think you got it all wrong fella.

i am not assuming .. intel is reacting .. they are *suing* Nvidia; i know a preemptive strike when i see one
- one that Nvidia has coldly calculated as a risk.

i believe you are overawed by intel's smoke and mirrors

you were all about itanium and netbust also

Hay I don't care who it is in the Industry. But NV needs to be slapped down hard. Amd/ATI has done dam fine job of it I might add. Destroying NVs margins and taking market share. If Intel has their eye on anyone its ATI. ATI is intels threat.

 

[Nemesis 1]

Originally posted by: chronodekar
Coming back to the earlier discussion of AVX, (nice pdf link Nemesis!) it seems to me, that AVX will be another instruction set about to be released by Intel. By micro-architecture, intel refers to the actual implementation on silicon. From a software view-point that's irrelevant. As a programmer, we see the processor as a black-box. i,e. we send instructions into it and expect output. A point made in the pdf is that AVX will be fully compatible with ISA. So that means that if we send in ISA based instructions to an AVX chip, it will respond as we normally expect.

On the matter of recompiling, Intel is saying that many of the ISA-based instructions can be replaced instruction-for-instruction with AVX ones. The key point to note here, is that it requires an additional step from a software developer. In simple english, "If you just recompile your program with this new compiler of ours(from Intel), you will be able to increase program efficiency by xx.xx%!!" What is not being mentioned is that you will end up being tied down to Intel chipsets.(or at the very least, Intel licenses, if someone-else uses it too)

If Intel manages to cover 95+% of the desktop market-place with AVX chips by 2011, then Nemesis1, your prediction will be fulfilled. i.e. AVX will have replaced ISA, and compilers WILL have to optimize for it.

BUT, if nVidia (or AMD) comes out with an alternative, which is at least, of equal if not superior performance (and we have seen both companies are capable of surpassing Intel's performance benchmarks) AND if they can cover about 20% of the market-place, then AVX will be seriously challenged. In that kind of a scenario (Marketplace: Intel 80% nVidia/AMD 20%), I'd say a programmer/developer will have to be very concerned about getting tied down to a particular license. There will be a push to converge on a "standard" of some sorts. From the above comments, I'd say that Transmeta (with it's code-morphing) could very well be the answer.

Now the above is, of course, highly speculative, and things can turn out very differently.

Addressing a different issue, some might not be convinced that having a 20% market-share is enough to tip the scales in your favour. I agree. But it's enough to force the industry to converge on a common standard. Just look at the web-browser scenario. IE (all of it's variants) have about 60-70% market share. Firefox has some 25%, with the rest going to Safari and others. Before Firefox, Microsoft practically dictated what web-standards were. Now, thanks to the rise of the Fox, every major web-designer is demanding for a common standard. And the result? IE8 is going to be more standards-complaint (With the IE6 rendering engine about to be dumped).

Mark my words, for Intel, what will matter in the end, is how much of the market they will be able to grab before nVidia arrives on the scene.

The part Bolded above . I think you have priority cuda mixed up with Intels AVX ISA. Intels has flat out said we don't care what ya write the code in , Keeping X86 simply gives them a leg up. Because of existing Apps. Intel can do everthing except cuda. But whats to stop Intel from doing the same thing NV is trying to pull. Copy NV IP. Why not?

You just don't get it Intel wanted EPIC along time ago. But MS. and AMD didn't and they conspired they did. X dec guys were involved they were. Ms. X dec guys AMD = AMD64.

Now NV AMD All are ready to move away from X86. Because GPUs got better performance If can utilize. Same as EPIC would have. But Intel turned everyones world upside down . By Building an X86 gpu larrabee. Than to top it off They have AVX up to bat. If Intel succeeds with Larrabee NV is over. I for One can't wait for the reviews.

Should be Intel getting spanked in todays games . Luckily 6 games are coming out with Larrabbe release. Man intel has deep pockets. 1 of the games project Offset is going to change how we see gaming . Not long to wait to see now.

 

[chronodekar]

Originally posted by: Nemesis 1
If Intel succeeds with Larrabee NV is over. I for One can't wait for the reviews.

This I also agree with. But it's a BIG "IF". We all know intel graphics by now. Their only +ive point (IMO) is that they are cheap. For gaming or anything requiring a graphics card you need an ATI or nVidia solution. (in today's market)

It will be a big thing if intel manages to pull it off. But, I've listened to intel about it's graphics for nearly 8 years now. I relented and bought an nVidia chipset. Now, I wish I had made that decision sooner.

You seem confident in intels graphics capabilities Nemesis1, but not me. Nope, definitely not me. At least, ... until I see it for myself.

I like your posts Nemesis 1, and I've read them all (in this thread anyway). But about graphics, I say Intel is just hot air.

Now, if we are talking about processors, then it's a different question.

 

[Nemesis 1]

We shall all bare witness to software render . Going by intels past efforts in GPUs . Has nothing to do with larrabee. Larrabee takes us back to the beginning. Back to the Future. Software render is coming from Intel . NV promised us BIG BANG and a CAN of Whoop ass. Than delivered nothing. Intels only claims I have heard is they can do software renders better than hardware. Hell the debates about over. Soon we shall see. One thing for fact . Intels drivers will be easier than NV or ATI. Maybe not on todays games but future is software render. You would have to be blind not to see that.

 

[theeedude]

Larrabee is Itanium for graphics. Here is Intel coming on like a freight train, if Intel is doing it, it's inevitable, right? Everyone else might as well close shop, right? Looks great on slides, great hype, I mean what could go wrong? In reality Intel is just starting to do what Sun, Sony, Azul, etc did 5 years ago, and pretending like it's god's gift to humanity. Lots of general purpose cores on a die. Wow, what a novel idea.

 

[Keysplayr]

Originally posted by: Nemesis 1
We shall all bare witness to software render . Going by intels past efforts in GPUs . Has nothing to do with larrabee. Larrabee takes us back to the beginning. Back to the Future. Software render is coming from Intel . NV promised us BIG BANG and a CAN of Whoop ass. Than delivered nothing. Intels only claims I have heard is they can do software renders better than hardware. Hell the debates about over. Soon we shall see. One thing for fact . Intels drivers will be easier than NV or ATI. Maybe not on todays games but future is software render. You would have to be blind not to see that.

This isn't a cult Nemisis, so can you nix the biblical "we shall all bear witness" fruitcake gab?
Intel's x86 patent expires within 2+/- years. What does this mean for the rest of the world? Is it true that Intel cannot renew these patents? I don't presume to know, but this is what the general jist is around the web. Is it true? If so, does this mean that anyone and his brother can now produce x86 CPU's if they had the know how? Including Nvidia?

Then, miraculously, AVX. Intel is going to try to change the technology and make it the standard so it can have another 17 years of patent dominance. It's current x86 patents will run out, so they will try to kill x86.

"If I can't have you, no one else cannnnnn..........!!!! "

Even if I'm way off here, don't think for a second that Intel has not been concerned with it's patents running out. Yes, that was/is a fear. I feel that Nvidia, had they been given access to x86 licensing would have had an x86 CPU by now to right along with their GPU's.
Yet you feel Nvidia needs a "smackdown"? Why wouldn't you want more competition in the CPU arena? We only have two choices that are viable right now. Been that way for about forever. The industry we are talking about is a dynamic one. Companies will morph and change as needed to survive or remain competitive and have a product people will want.
Intel doesn't want anyone else to have a chance to make something that they do besides AMD, and that's only because they have to.

So yeah, all this reeks of fear. I agree with the others in here. To dismiss it and say Intel is fearless is kind of close minded. Just like people, fear usually keeps us alive. Companies have the same fate. If they fear nothing, they die. Fear is good for them. Keeps them on their toes and on the lookout for a predator. And believe me, that patent running out probably has Intel seeing every other company that has the capability to produce a x86 CPU look like salivating tigers who haven't eaten in 22 years +.

 

[Idontcare]

Originally posted by: Nemesis 1
If Intel has their eye on anyone its ATI. ATI is intels threat.

That would be the ranking of threats to Intel as I would place them as well. NV is a threat, anything that breaths is a threat to Intel, but they are less a threat than AMD/ATI.

 

[Keysplayr]

Originally posted by: Idontcare

Originally posted by: Nemesis 1
If Intel has their eye on anyone its ATI. ATI is intels threat.

That would be the ranking of threats to Intel as I would place them as well. NV is a threat, anything that breaths is a threat to Intel, but they are less a threat than AMD/ATI.

Salivating Tigers. Yup. All of them are in Intel's eyes.

 

[Idontcare]

Originally posted by: keysplayr2003

Originally posted by: Idontcare

Originally posted by: Nemesis 1
If Intel has their eye on anyone its ATI. ATI is intels threat.

That would be the ranking of threats to Intel as I would place them as well. NV is a threat, anything that breaths is a threat to Intel, but they are less a threat than AMD/ATI.

Salivating Tigers. Yup. All of them are in Intel's eyes.

Salivating and bleeding out. Intel is the only one turning a profit in their semiconductor sales IIRC.

 

[Keysplayr]

Originally posted by: Idontcare

Originally posted by: keysplayr2003

Originally posted by: Idontcare

Originally posted by: Nemesis 1
If Intel has their eye on anyone its ATI. ATI is intels threat.

That would be the ranking of threats to Intel as I would place them as well. NV is a threat, anything that breaths is a threat to Intel, but they are less a threat than AMD/ATI.

Salivating Tigers. Yup. All of them are in Intel's eyes.

Salivating and bleeding out. Intel is the only one turning a profit in their semiconductor sales IIRC.

Of course. Because they have the very best out there right now. People are still buying despite the recession, it would appear. Every computer needs a CPU. That is AMD or Intel only. And every computer needs a display device. And that is Intel, AMD or Nvidia. And not all need gaming GPU's. Intels IGP's do just fine for most.

 

[apoppin]

Originally posted by: Nemesis 1

Originally posted by: apoppin

Why do you assume Intel is afraid of Nv . I think you got it all wrong fella.

i am not assuming .. intel is reacting .. they are *suing* Nvidia; i know a preemptive strike when i see one
- one that Nvidia has coldly calculated as a risk.

i believe you are overawed by intel's smoke and mirrors

you were all about itanium and netbust also

Hay I don't care who it is in the Industry. But NV needs to be slapped down hard. Amd/ATI has done dam fine job of it I might add. Destroying NVs margins and taking market share. If Intel has their eye on anyone its ATI. ATI is intels threat.

*needs* a slapdown

:roll:

if anyone "needs" anything like that ... it would be your company

ATi exists only as a division of AMD .. you are saying intel is afraid of AMD now


yes, 'coldly' .. calculated .. planned to get a desired response

But whats to stop Intel from doing the same thing NV is trying to pull. Copy NV IP. Why not?

their CPU engineer's inability to comprehend graphics, perhaps
CUDA, for sure

 

[Nemesis 1]

Na you don't get it. Intel is on a set path . Were others are still trying to find there way.

It doesn't matter what NV or AMD/ATI does. Intel has a course and it must stay on it .

Keys AVX isn't anything intel is forcing on anyone. Its part of the processor its their to take advantage of or not. Show me were Intel is forcing AVX on anything. You use it or you don't its about performance and efficiency first. AVX was Brillant . AND it gets ALL the monkeys off INTELS back.


I like competition. But I like tech better. Vector processing is very intriguing to myself. When Intel includes statements up to 1024 bit vector units . I get excited. Thats one reason I like ATI . Say what you want their gpu is based on a sound design . That will give Intel fits. AMD really needs Intels compilars. From what I am hereing . ATI and Intel still have very good relations. Don't count cal out. Your about to find out why.


Sandy design . I like it alot it has hugh performance and efficiency potencial.