Quantcast
NeoGAF: Wii U GPU 160 ALUs, 8 TMUs and 8 ROPs

Forums - Nintendo Discussion - NeoGAF: Wii U GPU 160 ALUs, 8 TMUs and 8 ROPs

One of the Project CARS developers has stated that the ALU figure is wrong for this rumour. It has 192. Still doesn't explain why the space on the die is too large though, even for 192 ALUs.



Around the Network
Pemalite said:
fatslob-:O said:
Pemalite said:
fatslob-:O said:

Wait a minute. Does this mean that AMD also has silicon graphics engineers.

Sure does.

I wonder if their the ones who were more involved in designing the R7 260X and the R9 290 as well as the R9 290X chips. 

I still wonder why silicon graphics went belly up. 


They wen't belly-up more than once. :P
Basically not enough design wins.

Matrox took over the professional market, nVidia and AMD wen't with the gamer then compute markets.
3dfx was bought by nVidia, PowerVR went mobile, S3 was bought by Via then sold off to HTC.
XGI was spun-off from SiS and tried to break into the market and failed, SiS then bought them out.
Intel and AMD started to include a free-GPU in all their platforms which squeezed out IGP's from Via, nVidia, SiS from the market.
NEC faltered and left.
Tseng was bought out by the then-ATI.
Cirrus Logic, Trident and Texas Instruments sold off/left the GPU market.

In the end we are left with a Duopoly.
PC gamers essentially fund the technology that goes into the consoles, console's are low-profit business for GPU manufacturers it's not enough to cover R&D costs if the big three wanted a new GPU built from scratch, AMD might make a couple Billions dollars over 10 years from all three consoles if they are lucky, which for AMD is a great thing, they need all the cash they can get.

As for who designed the R9 290 series, well. Not allot of *new* engineering wen't into it, most of the work was done with Bonaire from an architectural perspective, just AMD took advantage of the mature 28nm to blow up the die size.
AMD has also cut engineers to try to become profitable again, which might explain WHY we will have had the same GPU's for 3 god damn years.

@Bold Your dead on right about that LMAO. The only profitable time that they had ever since core 2 duo released was from spinning off their fabs to global foundry. Ahhh the blunders of crappy AMD management. 

Eh, the volcanic islands series GPU is already more advanced than the nvidia kepler. It is nvidia who needs to step up their game to compete with amd because alot of phone manufacturers are going to be using AMD graphics IPs which will threaten nvidia's already paltry amount marketshare over there. AMD by then should get a noticeable increase in profitability by then. 



fatslob-:O said:

@Bold Your dead on right about that LMAO. The only profitable time that they had ever since core 2 duo released was from spinning off their fabs to global foundry. Ahhh the blunders of crappy AMD management. 

Eh, the volcanic islands series GPU is already more advanced than the nvidia kepler. It is nvidia who needs to step up their game to compete with amd because alot of phone manufacturers are going to be using AMD graphics IPs which will threaten nvidia's already paltry amount marketshare over there. AMD by then should get a noticeable increase in profitability by then. 


Well. Allot of Phones do use AMD's GPU tech. My phone sure does.
Hint: Adreno. - Re-arrange the letters for Radeon. :P
However, Qualcomm now owns that tech and iterates upon it.

It's a big opportunity missed by AMD in my opinion.



Pemalite said:
fatslob-:O said:

@Bold Your dead on right about that LMAO. The only profitable time that they had ever since core 2 duo released was from spinning off their fabs to global foundry. Ahhh the blunders of crappy AMD management. 

Eh, the volcanic islands series GPU is already more advanced than the nvidia kepler. It is nvidia who needs to step up their game to compete with amd because alot of phone manufacturers are going to be using AMD graphics IPs which will threaten nvidia's already paltry amount marketshare over there. AMD by then should get a noticeable increase in profitability by then. 


Well. Allot of Phones do use AMD's GPU tech. My phone sure does.
Hint: Adreno. - Re-arrange the letters for Radeon. :P
However, Qualcomm now owns that tech and iterates upon it.

It's a big opportunity missed by AMD in my opinion.

Adreno is an anagram for radeon.

It's honestly not that big of a missed opportunity because AMD at the time didn't have an ARM license so their Adreno graphics divison was mostly useless for AMD on most fronts and they thought that the desktop divison can handle every market segment and it did hold up pretty well in the laptop segment remaining competitive with nvidia there. What's more is that AMD was also in debt at the time so they also couldn't bother to be in even more debt and plus they didn't have much experience in designing ARM processors to for the most part. It was actually the better choice for AMD to sell Adreno if you know what I mean from the financial standpoint. What I'm not sure if going fabless for AMD was a good idea in the first place. 



fatslob-:O said:

Adreno is an anagram for radeon.

It's honestly not that big of a missed opportunity because AMD at the time didn't have an ARM license so their Adreno graphics divison was mostly useless for AMD on most fronts and they thought that the desktop divison can handle every market segment and it did hold up pretty well in the laptop segment remaining competitive with nvidia there. What's more is that AMD was also in debt at the time so they also couldn't bother to be in even more debt and plus they didn't have much experience in designing ARM processors to for the most part. It was actually the better choice for AMD to sell Adreno if you know what I mean from the financial standpoint. What I'm not sure if going fabless for AMD was a good idea in the first place. 

Oh I agree completely for the most part.
But instead off selling it off, they could have licensed purely the GPU architecture out, I don't think AMD at the time understood how big mobile would get, which is hindsight really.
nVidia intends to license it's GPU technology out to ARM manufacturers for instance, then the ARM manufacturers can modify and improve it if they want, much akin to how many companies license ARM and modify it to their own needs, ARM still owns the I.P to the instruction set however.



Around the Network
Pemalite said:
fatslob-:O said:

Adreno is an anagram for radeon.

It's honestly not that big of a missed opportunity because AMD at the time didn't have an ARM license so their Adreno graphics divison was mostly useless for AMD on most fronts and they thought that the desktop divison can handle every market segment and it did hold up pretty well in the laptop segment remaining competitive with nvidia there. What's more is that AMD was also in debt at the time so they also couldn't bother to be in even more debt and plus they didn't have much experience in designing ARM processors to for the most part. It was actually the better choice for AMD to sell Adreno if you know what I mean from the financial standpoint. What I'm not sure if going fabless for AMD was a good idea in the first place. 

Oh I agree completely for the most part.
But instead off selling it off, they could have licensed purely the GPU architecture out, I don't think AMD at the time understood how big mobile would get, which is hindsight really.
nVidia intends to license it's GPU technology out to ARM manufacturers for instance, then the ARM manufacturers can modify and improve it if they want, much akin to how many companies license ARM and modify it to their own needs, ARM still owns the I.P to the instruction set however.

I don't think you understood how dire AMDs financial position was in.

AMD had to undergo alot of restructuring over the past few years ever since 2007. The only way they could sustain themselves over a long period of time was by selling alot of assests by combating losses after losses each quarter. Adreno was not what you'd call worthwhile division. AMD could still also license out it's graphics IPs if they wanted too even if the technology mostly came from the desktop side. The manufacturers could just scale down GCN so that it will take lower power consumption. Adreno was not worthwhile in the first place because it was alot smaller at the time while also not very vested so it was infact mostly useless for AMD. Selling it to Qualcomm was the right thing to do since AMD also couldn't be bothered to also invest in that graphics divsion. 



its obvious that it cant just have 160 alus cause that would mean only 176Gigaflops and 360 is 240gigaflops
its true that wii u is more optimized and has new features and has better performance but tyhat doesnt really matter when we talk about lazy ports cause all those adventages are nulled, not to mention that wii u is also port is downgraded to work according to the charactreistics of the old consoles and developers are still not used to the new hardware and also the engines have not been 100% tailored and optimized for wii u and ports are still working so its clear you cant achieve that with just 176gigaflops

no, at least for ports to work and even considering the use of undepowered engines like the unreal engine 2.5 you would need about a gpu of at least 400gigaflops

besides, the guys that are speculating this are also reducing the numbers of spus, which firstly were 320spus giving something like 352gigaflops, by putting interpolators on wii u gpu, which make no snese there cause amd already mentioned a couple of years ago that interpolators were inefficient and were slowing down the performance of the rv770 and thats why the rv870(basically the same old rv770 with minor changes to optimize the performance) didnt pack interpolators and instead the spus were assigned to do the interpolation, and since wii u gpu is based on teh rv770 but not an rv770 its obvious that the was perfected and at least has to be as good as an rv870 or the e6760, not to mention that rv770 wasnt designed for shade model 5.0 and wii u was made for that in mind