By using this site, you agree to our Privacy Policy and our Terms of Use. Close

A small, fast eDRAM or eSRAM used basically like a 3rd level cache offers a speed advantage when it scores a "hit", when you have a "miss", the speed drops down to that of main RAM. Now, if the data and/or program code in the eSRAM and those in the main RAM are independent enough from each other, the overall speed is the weighted average of the speeds of the two RAM weighted with their hit rate, but if you must process an amount of data larger than the eSRAM and they can be used only when they all have been processed, or if the result of computations on data in one of the memories is necessary to do other computation on the other memory's data, the slowest memory becomes a bottleneck just like when there is a "miss" looking for data in the fastest one, overall speed drops to the slowest one. Small and fast caches are great, and having some L1 and L2 cache has been essential to use efficiently the computing power of CPUs and GPUs since the 486, but L3 caches show their limits when amounts of data larger then them must be processed altogether. Nevertheless, even in the most extreme of these cases, the typical work of GPUs, a small but fast memory can still be very beneficial if it's large enough to fully contain a frame buffer and possibly also z-buffer and double or triple buffer, and a texture cache too if there is some space left.



Stwike him, Centuwion. Stwike him vewy wuffly! (Pontius Pilate, "Life of Brian")
A fart without stink is like a sky without stars.
TGS, Third Grade Shooter: brand new genre invented by Kevin Butler exclusively for Natal WiiToo Kinect. PEW! PEW-PEW-PEW!