ethomaz said:
It is true... ESRAM have way better latency (it is on die at all) but latency don't beneficiates graphics processing... it can beneficiate post-processing AA for example but GPU tasks are not affected by high or low latency. ESRAM performance is fine... the issue is the size... 32MB is really small but I undestand MS because more could be costed A LOT. |
It's not really the cost, but the way the ESRAM is integrated into the GPU/CPU die. MS can increase the amount of ESRAM, but if they do, they have to shrink one of the other two components. In this case, the GPU will be shrunk because the CPU is already small. MS is between rock and hard place with this type of configuration. You either sacrifice GPU power or you sacrifice ESRAM buffer. Neither choice is good.