R500/NV50
- Thread starter kenneth9265_3
- Start date
chavvdarrr
Veteran
no, its great 
Cyberdigitus
Newcomer
here we go again...
stevem said:
The best I can fathom, as I mentioned here that that the next core is actually and extension of the R300 line. This also corresponds with the comment I heard that the OGL rewrite wouldn't be available "for a couple of product releases". It seems like the rewrite is actually targetting R600, which, if I'm correct, will be the first PC product to be based on the unified shader architecture. They should be able to prototype the drivers on similar silicon as they will have the XBox chips back, which (again, if I'm correct) will be the platform R600 is developed from.
Architectural Lines:
R300 --> R420 --> R520
R400 (not released) --> R500 (XBox) --> R600 (PC / Longhorn)
pat777 said:
Sorry, when I say "the unified architecture" I'm not talking about DirectX, I was talking about ATI's. I'm keen to see if NVIDIA will adopt a physical unified architecture as when I asked Kirk about it in the initial NV40 presentations he was quite heavily against the idea at all - considering the way he said it and the fact that he's probably working on these generations at that time, it made me wonder if they would do it at all (however, it may be the case that the constraints on DirectX make it an inevitability).
Mindriot said:
I firmly believe the original R400 would have been a unified shader archtecture, however its development is now forming the basis of the XBox R500 chip.
Re: shader architecture
IIRC he threw in issues like cache coherancy and the like. The instruction scheduler has to be very good when allocating tasks and keeping the ALU's fed with as few bubbles as possible (and this hardware is a fixed costs across the entire range).
Scott_Arm said:
IIRC he threw in issues like cache coherancy and the like. The instruction scheduler has to be very good when allocating tasks and keeping the ALU's fed with as few bubbles as possible (and this hardware is a fixed costs across the entire range).
Errm, these are not "inherant problems with the idea" they are different considerations when dealing with a new architecture - at the hardware level a unfied shader model can present potentially greater performance per silicon (at the high end) but to achieve it effectively it represtents a different set of challenges to find the optimal solution.
chavvdarrr said:no, its great
I don't see what's so great about killing innovation and competition, cause that's what it's going to do if they end up doing it that way. The IHVs will have no choice but to implement the exact functionality that Microsoft think is good, rather than thinking for themselves. MS becomes the industry dictator, while IHVs end up building so similar products that basically only the brand name will differ.
Um, yeah how would this kill competition?
The details of the architectures and how they achieve a more unified, consistent set of features will differ markedly.
Is a Pentium 4 the same as an Athlon 64 just because they both run x86 with MMX and SSE2?
I don't see how a more consistent and broad set of base features (there can always be more, though they might not be exposed other than OGL or something) will just make all the products out there the same.
The details of the architectures and how they achieve a more unified, consistent set of features will differ markedly.
Is a Pentium 4 the same as an Athlon 64 just because they both run x86 with MMX and SSE2?
I don't see how a more consistent and broad set of base features (there can always be more, though they might not be exposed other than OGL or something) will just make all the products out there the same.
