Ati on Xenos

Jawed said:
Titanio said:
I'd also expect Xenos to have performance much more akin to this coming generation than any coming after it, even if its architecture/approach is different ;)

ATI is saying that Xenos performs like a 32-pipe conventionally architected part. We'll just have to wait and see, eh? But I don't believe that 32 pipes is going to be in the next gen of PC parts (G70/R520 or refresh).

32 = 8 vertex + 24 pixel? ;) Perhaps if you had Xenos entirely working on pixels all the time it could act like a 32 pixel pipe chip but that's not exactly likely to be happening very often..

I think/hope, btw, that we will see 32 pixel pipe parts from ATi and NVidia. In terms of raw shading power, if R520 is a 32 pixel pipe part, I think it should outperform Xenos. It may not be as efficient of course, but that's a different question..
 
Jawed said:
Titanio said:
32 = 8 vertex + 24 pixel? ;)

I'll leave you to clutch at straws :D

Jawed

I think it's a valid question..are they talking about it behaving like a 32-pipe part if the whole chip is working only on pixels? That'd make sense.

Xenos is offering a next-next-gen approach to graphics, but not its performance. Unified shaders is a new way to arrange "power", for want of a better expression..it doesn't dramatically increase the number of execution units you can stick on the chip etc.
 
And not to mention that there's prolly a list as long as your arm of NVidia-specific OGL extensions. That workload might be common to both PC and PS3, but those extensions still need to be weaved into the developer's platform for PS3 - Sony may well have had OGL running on PS3 OS for years now, but not the NVidia-specific parts of it
I don't think you understand how Sony normally deals with software development aspects at all. :D
 
Fafalada said:
And not to mention that there's prolly a list as long as your arm of NVidia-specific OGL extensions. That workload might be common to both PC and PS3, but those extensions still need to be weaved into the developer's platform for PS3 - Sony may well have had OGL running on PS3 OS for years now, but not the NVidia-specific parts of it
I don't think you understand how Sony normally deals with software development aspects at all. :D

Has Sony worked with NVidia on a console before?...

Jawed
 
Jawed said:
Tell me who commonly refers to a 6800 Ultra as a 22-pipe card. There's your answer.

Jawed

I wasn't ragging on them for referring to it as being the equivalent of a "32-pipe card". If they're counting vertex pipes, that's fine, just would like to know.
 
Fafalada said:
And not to mention that there's prolly a list as long as your arm of NVidia-specific OGL extensions. That workload might be common to both PC and PS3, but those extensions still need to be weaved into the developer's platform for PS3 - Sony may well have had OGL running on PS3 OS for years now, but not the NVidia-specific parts of it
I don't think you understand how Sony normally deals with software development aspects at all. :D

Doh beatme to it :)
 
exactly how do you get 2 years in addition to that ? 2 years before that they were most likely still working on the nv3x and just starting the nv40. That would have been 4 years or so of developement and i highly doubt sony would just get a modified g70 but that is what nvidia's comments point too a modified desktop part
 
We will probably won't find out until later about How long Nvidia actually had time to develop RSX, but I think RSX wouldn't be so radically different even if they had extra time to develop. Nvidia is not really in love with unified shader and given that, I don't think RSX would be significantly different part anyways even if they had really long time to develop.
 
scooby_dooby said:
i read they had filed patents regarding using 3 cells in the ps3, 2 would act as the GPU.

Then last year they signed a deal with nvidia to make a GPU, and only went with 1 cell processor.

So i think most of the speculation is based on the patents Sony filed.
ok, it's just been confirmed from the horse's mouth that at one time, they were considering using a CELL chip as a GPU ..

Kutaragi himself states that they were considering using one CELL as CPU , and another CELL as GPU , in part 3 of the Hiroshige Goto (PC Watch) interview, discussed here: http://www.beyond3d.com/forum/viewtopic.php?t=23836&start=180 and here: http://www.ga-forum.com/showthread.php?t=51518
 
Jawed said:
ATI is saying that Xenos performs like a 32-pipe conventionally architected part.

Actually, they aren't. The person that said that that has undergone some training since then...
 
DaveBaumann said:
Jawed said:
ATI is saying that Xenos performs like a 32-pipe conventionally architected part.

Actually, they aren't. The person that said that that has undergone some training since then...

So it is possible to say an equivalent in traditional pipes?
IF it is, can you tell us? (or it is under NDA)
 
pc999 said:
DaveBaumann said:
Jawed said:
ATI is saying that Xenos performs like a 32-pipe conventionally architected part.

Actually, they aren't. The person that said that that has undergone some training since then...

So it is possible to say an equivalent in traditional pipes?
IF it is can you tell us? (or it is under NDA)
I'd be willing to bet there will be an attempt to answer that question in his upcoming article
 
Nope, I'm not even going to make an attempt to go there. Its stupid to do things like that when we are talking about theory and hardware we can't experiment with.
 
I was going to say. Fair comparisons can only be made 'in the field' with proper benchmarks, given unified shaders (plus the other special sauces Xenos has) are an untried technology thus far.
 
DaveBaumann said:
Nope, I'm not even going to make an attempt to go there. Its stupid to do things like that when we are talking about theory and hardware we can't experiment with.
ahhh ... ok, fair enough :)

I guess I would have lost that bet, then :D
 
Back
Top