NV48 not cancelled according to Taiwanese sources

[kemosabe]

Digitimes might be referring to what we've been calling NV47, but in any case it sounds like 0.09u vs. 0.11u for next spring. Makes you wonder which 0.09u process NVDA will adopt and how far along they are on their next design.

 

[phenix]

Is this Nvidia's answer to X800XL?

 

[Kaotik]

Is this Nvidia's answer to X800XL?

nope, it's nVidia's answer to R520

 

[Bjorn]

Is this Nvidia's answer to X800XL?

nope, it's nVidia's answer to R520

I find it rather unlikely that NVidia will release their main competitor to the R520 using a 0.11u process. Especially since it's a non low-k process.

 

[Pete]

Well, 110nm doesn't use low-k, and nV seems to have had more luck without it (NV43 vs. RV410). A 24-pipe version of the NV40 should compete pretty well with a, say, 16-"super"-pipe R20, no?

It's curious why JHH said the PS3 GPU took a mere 50 engineers, though they were apparently working on it for two years. I wonder if that was a way of assuring investors nV wouldn't experience an innovation slowdown around its release, as with the NV2A? I mean, if nV did move from IBM to TSMC, would they switch to a different process at the same time (when they already have 110nm working well at TSMC with the NV43)?

 

[Ailuros]

Non low-k 110nm has been so far only a process that has been used for mainstream parts and not the high end for a reason.

A 24-pipe version of the NV40 should compete pretty well with a, say, 16-"super"-pipe R20, no?

What's a "super" pipe anyway? R520 sounds so far like a low-k 90nm iteration with higher clock speeds than a NV40. IMHO R520 shouldn't be too far apart in terms of transistor count from a NV40 (in any case not twice as high as R420), which makes me think that ATI had a very good reason for picking low-k 90nm in the first place.

Why would anyone think that let's say a ~300M/6 quad NV4x@500MHz (or higher) on 110nm is "that" easy? What's the transistor count on NV43/2 quads exactly again?

 

[hovz]

i think r520 will be a slightly speed bumped r420 with sm 3.0. i think the main improvement will be in the memory bus supporting gddr4. i dont think sm 3.0 is of any importance right now.

 

[Pete]

No doubt I'm wrong. I just thought nV might be able to be competitive with an older process the same way the R300 handled the NV30, but that's probably overlooking the greater parity of features and speed this round.

 

[AndrewM]

It's curious why JHH said the PS3 GPU took a mere 50 engineers, though they were apparently working on it for two years.

Uhh, I think that's how many engineers they are using to integrate it into the PS3.

I wonder how people twist this stuff in their head..

 

[Ailuros]

It's curious why JHH said the PS3 GPU took a mere 50 engineers, though they were apparently working on it for two years.

Uhh, I think that's how many engineers they are using to integrate it into the PS3.

I wonder how people twist this stuff in their head..

It's not always easy for us laymen to understand and/or combine each and every tidbit of information that comes along. The Sony/NVIDIA cooperation announcement for PS3, marks that the GPU meant for PS3 and NV's next generation GPU have been developed in parallel. Combine that with the tidbit above and of course it makes sense to assume that the 50 engineers were responsible for the PS3 integration.

In any case doesn't it always depend also what one means with "engineers" exactly? IMHO if one would specify a very specific group of engineers that number isn't exactly low on the other hand.

 

[Farid]

It's not always easy for us laymen to understand and/or combine each and every tidbit of information that comes along. The Sony/NVIDIA cooperation announcement for PS3, marks that the GPU meant for PS3 and NV's next generation GPU have been developed in parallel. Combine that with the tidbit above and of course it makes sense to assume that the 50 engineers were responsible for the PS3 integration.

Actually JHH said it took hundreds of engineers to create their next-gen architecture, and he added that there's 50 engineers working, right now, on the integration of this architecture to the PS3.

We don't have any more infos than thoses, for instance we don't know if that next-gen GPU was created with this PS3 deal in mind (very unlikely, but still impossible, for the moment, to rule out.), we don't know to what extent the PS3 GPU will be custom, etc...

 

[Jawed]

I reckon JHH is trying to make everyone think that NVidia has the successor to NV40 in the bag, whereas by "next-gen" he actually means NV40.

Jawed

 

[Pete]

It's curious why JHH said the PS3 GPU took a mere 50 engineers, though they were apparently working on it for two years.

Uhh, I think that's how many engineers they are using to integrate it into the PS3.

I wonder how people twist this stuff in their head..

Sorry, I was unclear. I meant to say "is taking," not "took." I imagine NV2A took about the same number of engineers to "integrate" into the Xbox, but the NV2A/25 surely took a bit more work.

My point was that I thought JHH said it to assuage investors that they wouldn't experience a similar temporary setback in industry leadership as they did after the NV2A. I think it without much experience or insight into this market, though.

 

[Megadrive1988]

what are the differences between NV47 and NV48 ??

my understanding is, both are supposed to be highend refreshes of NV40 / GeForce 6800 Ultra.

the NV47 *supposedly* has 24 pixel pipelines. what about NV48 ??

 

[Fodder]

NV48 was originally slated to be a simple X850-like PCIE-native re-jig of the NV40 core, but I think the general assumption is that it's quite different now.