That is what i was refering to. If Sony would incoporate NVidia Ip (i.e. e.g. their next generation ps-units) into a PE they'd certainly talk about it as IP-licensing. This, though, sounds more like NVidia is about to supply a whole discret GPU/VPU/Graphics chip to the system and i do not think (at this point) this means NVidia desgining the whole graphics PE (i.e. gaining access to all Cell IP) as otherwise you'd clearly have that pointed out in their press release (they have got stockholders to treat to after all).
CONFIRMED: PS3 to use "Nvidia-based Graphics processor&
- Thread starter passerby
- Start date
- Status
nAo said:When? Where? Who?
Hofstee, in San Fransisco, Yesterday. I'm inclined, knowing more from the IBM angle, to think that this is NOT akin to the PlayStation3 having a conventional PC topology. Hofstee, a cool guy, was talking directly about this, firstly the moving of the North|South-bridge and Graphics/Vizualization Acceleration on-die.
He then talked about the movement from the the current "one-way" rendering pipeline:
Code:
[CPU] --> [GPU] => [Output] paradigmtowards a Cell based, "two-way" one:
Code:
[CPU] <=> [integrated Acceleration] => [Output]He specifically talked about doing away with texture mapping and spoke of more "physics" based modeling.
Pinky said:
Think about it, Sony Group & Toshiba want to use the architecture in CE applications in the scaled down Cell designs. IBM wants to use it in workstations and vizualization. Your thinking, IMHO, doesn't seem to be compatible or realistic in this regard to STI's ambitions and what they desire in a common|modular and scalable artchitecture.
Shogmaster
Regular
Vince said:
Well, this is what people are now talkign about with PCI Express.
nAo said:When? Where? Who?Vince said:
http://www.beyond3d.com/forum/viewtopic.php?p=425217#425217
BTW, nAo, was THIS the famous "rumor" we disscussed a few weeks a ago? Or is it something totally unrelated?
Oh, and yeah, there goes Pana theory about an in-house GPU...
But i'm certain he'll be more than happy that he was wrong about this one.
Too bad it's not possible to bold the "" icon... I feel this thread should have needed it.
Anyway, bit of mixed feelings here really. Dunno what to think of this. NV has no real experience with console GPUs, the work they did for XB was just an off-the-shelf part of theirs with minimal tweaking, whereas all the other consoles have always had custom jobbies done for them.
The NV27 1/2 with its external RAM worked in the XB because it was mated to a slow CPU with little bandwidth to feed. Cell on the other hand will be a bandwidth monster, so will they go eDRAM instead for the GPU? NV hasn't done any products so far with that tech, will that be a problem for them?
I'm also hesitant to cheer NV on because they still seem to want to steer the DX API to conform to their designs, that shows they haven't really learned from their mistakes with NV30 IMO.
Well, we'll see what happens, I guess.
" icon... I feel this thread should have needed it. Anyway, bit of mixed feelings here really. Dunno what to think of this. NV has no real experience with console GPUs, the work they did for XB was just an off-the-shelf part of theirs with minimal tweaking, whereas all the other consoles have always had custom jobbies done for them.
The NV27 1/2 with its external RAM worked in the XB because it was mated to a slow CPU with little bandwidth to feed. Cell on the other hand will be a bandwidth monster, so will they go eDRAM instead for the GPU? NV hasn't done any products so far with that tech, will that be a problem for them?
I'm also hesitant to cheer NV on because they still seem to want to steer the DX API to conform to their designs, that shows they haven't really learned from their mistakes with NV30 IMO.
Well, we'll see what happens, I guess.
DaveBaumann said:
I think you're not comprehending the level of bandwith Hofstee is talking about when he talks of moving Acceleration on-die.
PCI-Express isn't relative; it isn't even relative to the off-die bandwith that Cell's gets from it's Redwood interface as presented at ISSCC. PCI-Express is 500MB per lane (32 max). Cell's external communications (ala Rambus/Stanford) is 800MB (6.4Gbit) per pin.
THat's the first time I see this gamespot's article!Vysez said:BTW, nAo, was THIS the famous "rumor" we disscussed a few weeks a ago? Or is it something totally unrelated?
I was just referring to the the fact some guy (a couple of months ago I believe..) told us that the PS3 GPU was indeed developed by nvidia..ehhe..he was right
ciao,
Marco
Vysez said:
http://hardware.gamespot.com/Story-ST-9398-1266-x-x-x
he treats those southbridge-northbridge etc. structures as 'old' things. Hmm, Xbox2
Martin Eddy
Regular
PiNkY said:
They did say it was a licensing deal. I can't imagine NVidia producing another
console chip after the XBox debacle. I think it's more likely that they will be sharing IP.
i knew it!!!!!!!!!!
Everything about the GPU in PS3 screamed Nvidia.
Jaws said:
Still room for at least two more, one for the I/O chip and one for sound...
- Status
