RussianSensation
Elite Member
- Sep 5, 2003
- 19,458
- 765
- 126
One would naturally hope that a more efficient card would require less power to achieve similar performance.
Sheesh. D:
GF104 is less efficient than GF100, so your comment makes little sense. You may think GTX460 is more efficient than GTX470/480, but it isn't. It only appears so because they downclocked the GPU clocks and reduced the number of CUDA cores, thus simplifying the complexity of the chip. Once you factor performance per CUDA core, the performance per watt of GF104 is hardly any more efficient. Plus both cards are still on the same 40nm process. It's pretty clear that NV will have to do a lot more than reshuffling of TMUs before they can claim improved efficiency for Fermi architecture.
FYI, GTX480 often comes awfully close to GTX460 768MB/1MB in SLI. So people need to take into consideration why exactly GF104 appears to be more "efficient" - in stock form, it's about a 50% slower chip than the 480 CUDA core GF100.
Last edited: