What do you expect of the NV35? Here's what I expect.

[StealthHawk]

Originally posted by tazdevl
Aren't AMD and Intel using SOI, even for Prescott?

AMD isn't using SOI yet, it was planned for Barton and scrapped. Hammer will use it, but it's not out yet. don't know about Prescott.

[GlowStick]

Originally posted by StealthHawk
AMD isn't using SOI yet, it was planned for Barton and scrapped. Hammer will use it, but it's not out yet. don't know about Prescott.

Hmm, just my belif is, though it may be wrong.

That the Big boys (Intel and AMD) research how to make chips, their chips are the most complex and intel is allwasy on the leading edge for size .13 and .09 etc.

So id imagine that Intel or AMD would start useing a new process then it would eventually be cloned or adopted by othter companies.

Then again TMSC could have a really good R&D department if they can handle enough orders from companies.

[StealthHawk]

Originally posted by GlowStick
Hmm, just my belif is, though it may be wrong.

That the Big boys (Intel and AMD) research how to make chips, their chips are the most complex and intel is allwasy on the leading edge for size .13 and .09 etc.

So id imagine that Intel or AMD would start useing a new process then it would eventually be cloned or adopted by othter companies.

Then again TMSC could have a really good R&D department if they can handle enough orders from companies.

yeah, AMD has been on .13 long before nvidia, and Intel long before AMD.

[Lezmaka]

Prescott supposedly use strained silicon

[borntosoul]

i think intel will have .09 before the end of this year so maybe nvidia wil have the nv40 by mid next year or later than that with .09 ,i dont see how much further they can go with .13 after the nv35 comes out , maybe they worked out a deal where they will be on .09 much sooner who knows

[mikechai]

Originally posted by borntosoul
i think intel will have .09 before the end of this year so maybe nvidia wil have the nv40 by mid next year or later than that with .09 ,i dont see how much further they can go with .13 after the nv35 comes out , maybe they worked out a deal where they will be on .09 much sooner who knows

I think using an improved and optimised 0.13um process should be sufficient for NV40.
NV45 will definitely use .09 process though.
________
ZR1200A (http://www.cyclechaos.com/wiki/Kawasaki_ZR1200A)

[mikechai]

Originally posted by StealthHawk
yeah, AMD has been on .13 long before nvidia, and Intel long before AMD.

Yeah, Intel has always been the leader in this area. If only NVIDIA/ATI has access to that kind of technology ...
Since ATI/NVIDIA doesn't have its own manufacturing plant, it has to rely on TSMC/UMC(third party) to manufacture its GPU/VPU. Something goes wrong then it affects the whole product cycle.
NVIDIA's NV30 is one example. They can't afford to overestimate TSMC capability. NVIDIA can only make the mistakes once.

So NV35 has to be good unless they are smoking something hallucinogenic ...
________
head shop (http://headshop.net)

[tazdevl]

Intel was 15 months ahead of AMD when it came to the .13m shift. They are already sampling .09m chips and should be announced the end of 2Q, perhaps seeing product in the channel in early 3Q. The gang out of Portland has a pretty solid policy regarding paper launches.... there's a very short lag between the announcement and product hitting the street.

Granted a lot depends on the competitive situation.

With this shift, Intel will be in a position to truly beat the hell out of AMD. The increase in production capacity with the .09m die shift and moving to 300MM wafers are pretty significant. Intel's core strategy has always been to push manufacturing capacity... this is the next step. With all the problems regarding the Hammer, it's going to put a lot of pressure on AMD to move even quicker.

Whoever mentioned it, you're right. It is strained silicon.

Also, doubling the width of the bus and changing to a 8X1 wouldn't double the speed of the NV30. There are too many other constraints within your system to have that kind of increase.

I doubt we will see .09m until late 2004 from other folks, not including AMD.

[StealthHawk]

Originally posted by borntosoul
i think intel will have .09 before the end of this year so maybe nvidia wil have the nv40 by mid next year or later than that with .09 ,i dont see how much further they can go with .13 after the nv35 comes out , maybe they worked out a deal where they will be on .09 much sooner who knows

both Intel and AMD will have .09 this year.

[borntosoul]

i didnt think amd would have it this year but if they did it would be cool

[MuFu]

Ok, it seems that TSMC's 0.13u fab process uses Cu interconnects only - NV30 does not use Al as has been suggested. That Jen-Hsun interview was a bit misleading.

So NV35 is fundamentally the same architecture (with the exception of the 256-bit mem interface) and uses same process tech as NV30. Hmm.... =\

MuFu.

[tazdevl]

Originally posted by StealthHawk
both Intel and AMD will have .09 this year.

Very doubtful in terms of AMD.

Intel is a certainty.

[StealthHawk]

NM, you guys are right. AMD pushed its .09 CPUs into next year when everything got delayed. sorry for the confusion :o

[MikeC]

Originally posted by MuFu
So NV35 is fundamentally the same architecture (with the exception of the 256-bit mem interface) and uses same process tech as NV30. Hmm.... =\

MuFu.

If NV35 also uses 500MHz DDR2 memory, then its physical memory bandwidth would be twice that of the GeForce FX or 32GB/sec.

There might be a couple of other enhancements like 2nd generation s94$%k and 88sw((30sdf. Man, I've got to get a better keyboard...heh.

[GlowStick]

Originally posted by MikeC
If NV35 also uses 500MHz DDR2 memory, then its physical memory bandwidth would be twice that of the GeForce FX or 32GB/sec.

There might be a couple of other enhancements like 2nd generation s94$%k and 88sw((30sdf. Man, I've got to get a better keyboard...heh.

I really dont beleive the NV35 will have a 256bit DDRII bus.

It goes under my same list as the R400 this fall.

[Uttar]

What matters most for the NV35 now is its MIPS ( Million Instructions Per Second ) in the Pixel Shader ( and Vertex Shader, to a lesser extent )
The R300 can do 3500 MIPS in the PS
The NV30 can do 450 MIPS in the PS
That is with a 96 instruction program

Here are the theorical figures, in case anyone is interested:
R300: 325*8 = 2600
NV30, FP32: 500*4 = 2000
NV30, FP16: 500*8 = 4000

As you can see, something definitively ain't right with the NV30 PS architecture... Maybe the drivers could fix it, but then that would be a godly driver set.

The NV35 could still run at 500Mhz and it would be fine if its practical figures were nearer of its theorical figures.
Oh, and AA quality is important too ( Gamma Corrected AA, *please* ! And RGAA for the default 4x mode, *please* )


Uttar

[kyleb]

no kidding, the fx's aa is the biggist turnoff in my book, i was hopeing the adoption of 3dfx technology would have solved that. :(

[borntosoul]

Originally posted by MikeC
If NV35 also uses 500MHz DDR2 memory, then its physical memory bandwidth would be twice that of the GeForce FX or 32GB/sec.

There might be a couple of other enhancements like 2nd generation s94$%k and 88sw((30sdf. Man, I've got to get a better keyboard...heh.
mike a few more clues would be nice :)

[nutball]

Originally posted by GlowStick
I really dont beleive the NV35 will have a 256bit DDRII bus.

It goes under my same list as the R400 this fall.

It's been sold as fact in Another Place by a few sources who are well regarded as knowing what they're talking about :)

[kyleb]

i sure hope it is, i can't see it being very competitive without it. :(

[GlowStick]

Originally posted by nutball
It's been sold as fact in Another Place by a few sources who are well regarded as knowing what they're talking about :)


Hm so you can confrim that NV35 is 256 DDR II?
Any links would be nice, or just the name of hte site if u dont wanna look for the exact link.

If the NV35 dose have the 256 DDRII bus i think ill wait for the NV35.

But could the rumor be wrong, and its 256DDRI? hehe