| greenmedic88 said: Lost it right here: "• We have more memory bandwidth. 176gb/sec is peak on paper for GDDR5. Our peak on paper is 272gb/sec. (68gb/sec DDR3 + 204gb/sec on ESRAM). ESRAM can do read/write cycles simultaneously so I see this number mis-quoted." Doesn't work that way. Even if every bit of data ran directly through the ESRAM, which it can't, due to the relatively tiny allotment, the speed of the ESRAM wouldn't be "boosted" an additional 68gb/sec by the much slower DDR3 memory. |
Yep: the only way to add the speeds of two different pools of memory is putting them in parallel and doubling the width of the memory bus, it's commonly done on most motherboards, and to be able to use that feature you need to use couples of RAM modules of the SAME type, size, speed and latency (probably different speed and latency are allowed by some chipsets, but they simply would be made work by lowering the whole speed and latency to the lowest of them, as they must work synchronously, and this isn't a secondary thing, so much that many RAM manufacturers sell their modules also in couples carefully matched to get the best performace from them when used in dual channel configuration). Trying to use RAM of DIFFERENT size, type, speed and latency in parallel is just an engineering nightmare.
Or maybe they are meaning that CPU and GPU can access those two RAM separately and simultaneously, but even then, whenever they must work together, providing data or instructions each other, the unit working with the slowest memory would bottleneck the whole system.
