So its reasonable to assume P11 with 1024 sp's = ~ gtx960/ R9 380(nonX)
and with over 100% more shaders the P10 = ~ 2x gtx960/380 or a gtx980/390.
I can see this.:thumbsup:
I'm really not expecting a massive jump in per shader performance. Look at it this way, they demoed P11 vs a 950, a card with similar performance to 1280shader@1GHz, 212mm² Pitcairn. All indications are that P11 will be in the 100-120mm² range, which if P11 was also 1280 shaders (but clocked much lower in the test) would be a reasonable 2x scaling in density. If P10 does come in at 232mm², 2304-2560 shaders makes a lot of sense. If its shaders also perform as well at 800MHz at Hawaii at 1GHz, it might just be a hair below Hawaii in current D11 games.
Shame with the low memory bandwidth. Hopefully its just ES and retail will be at least 7Ghz.
I'd be shocked if final clocks ended up being 800MHz and 6Gbps, but who knows.
Shame with the low memory bandwidth. Hopefully its just ES and retail will be at least 7Ghz.
The shaders are not going to be the same. I am predicting more performance out of less shaders. Anyone who is saying that a number of shaders on Polaris will perform the same as the same number of shaders in Hawaii (clock equalized) will be proven wrong, just watch.
Improved shader efficiency was a key feature of Polaris. So yeah the shaders are definitely better.
http://www.anandtech.com/show/9886/amd-reveals-polaris-gpu-architecture/2
Polaris11-1280SP 128bit 5-6Ghz DDR5 120mm SKU
Cutdown version 1024sp
Polaris10-2560SP 256bit 6Ghz DDR5 232mm2 SKU
cutdown version 2304SP
FULL polaris11-R9 285/380 performance
FULL Polaris10-390x/GTX980 performance
But even if the full chip is 2304SP, there is no reason it cannot outperform Fury X. 2048SP GTX 980 outperforms the 2880SP GTX 780 Ti. Remember that due to a flawed design, the 4096SP Fury X doesn't get the full benefit you would expect over a 2816SP 390X. The 980, compared to the 780 Ti, also has more ROPs and a higher clockrate. I doubt Polaris 10 will have more than 64 ROPs, but it should be able to achieve a higher clockrate if they designed it right especially with a node shrink.
980 also had less memory bandwidth than the 780 Ti. It has better compression. Fury X is actually less efficient with its memory than a 390X. It still tests better thanks to having so much more bandwidth, but it is less efficient. Further efficiency gains closer to Maxwell or even better will mean the bandwidth gap isn't as important.
It better beat the Fury X in performance or it will be a brand failure though. Higher efficiency will help the AMD image, but they cannot afford to let Nvidia get a large lead in raw performance on the top product because of the halo effect.
Head1985 said:polaris 10(if its 2560SP) can be maybe slighly faster than GTX980 but only if its not memory bandwidth bottleneck.I dont see it match Furyx.
if there is any truth to this. another FAILED.
need more power, not less wattage.
Ah, but improved shader efficiency doesn't necessarily imply improved shader performance. They likely will, especially in those areas that can use the new features (heavy tessellation, etc), but the exact same performance per shader would still be more efficient even node-normalized if it does so using less power.
AMD needs to improve it value propositions, and that means the node shrink and architectural improvements to mitigate the need for such high width memory buses. But the node shrink will allow AMD to possibly boost it's dedicated mobile graphics - a market Nvidia absolutely owns because Maxwell is just so much more power efficient and has less TDP at comparable levels of performance with AMD dGPUs.
in the mobile gpu market. yes tdp does make sense. the balance between performance and tdp is important. battery capacity is only so big.
as for the dedicated gpu market. for the 5-10 hours that a gamer manage to even put in per week. does it even matter how much wattage it pulls.