| shikamaru317 said: 11.4 teraflops sounds way too high for PS5/XB4 imo. PS4 was roughly equivalent to a $150 GPU when it released in 2013. 5-6 years earlier, a $150 GPU had somewhere around 500-600 gflops if I remember correctly. So that shows about a 3 to 3.6x improvement in mid range graphics in 5-6 years. 3 times the PS4's 1800 gflops comes out to about 5400 gflops, 3.6 times PS4's 1800 gflops comes out to about 6400 gflops. I think that's the range we're looking at for PS5 and Xbox 4, 5400-6400 gflops, and that's AMD gflops, not Nvidia gflops. I'd like to think that PS5 and Xbox 4 will be more powerful than that in 2018-2019, but I just don't think it'll be possible to top that for $400-500 without selling their consoles at a loss. |
The HD 7850 released for $250 at the beginning of 2012.... wouldn't it be reasonable to expect that in 2018 AMD manages to release a 10 nm series whose $250 card gives more than 6 TFLOPS, for a console release later on 2019? The R9 290x is already sub-$250 and packs 4 TFLOPS. And like I said, the only way I see your scenario happening is if we are stuck on 14 nm for a time an yields are bad enough to push prices up, like it happened to 28 nm when the PS4 and XOne were being designed. This, and Sony and MS taking the risk of releasing a console with few noticeable graphic improvements due to diminishing costs.
80 nm to 28 nm improved power consumption from 2.5 GFLOPS/W to 20 GFLOPS/W. Am I too optimistic to expect maybe 75% of that 8x improvement going from 28nm to 10nm, which is the same distance as 80 to 28nm? I'll admit that maybe I am... though in my defense, the strongest argument might be really what Sony and Microsoft are willing to release. I used to think this generation might be shorter because of underpowered consoles, and a lack of competition, but now I think the opposite might happen, solely on the grounds of diminishing returns and slower GPU cycles.
