07-27-2015, 11:13 AM
It wasn't just Fudzilla that was saying this about HBM2 memory in the past. All other rumors pointed to double the bandwidth, and quadruple the capacity, but Fudzilla is the first to say that it could even be 8x the capacity using the same amount of chips.
Remember, when all rumors about HBM1 memory was that it would be 1250 MHz? Then it turned out to be only 1000MHz, and a poor overclocker at that. It was supposed to be like a 5000MHz effective GDDR5 with a 1024-bit bus, as one of the rumors stated. That was a lousy rumor, but it wasn't too far off (1000MHz, with a total of 4096-bit bus width).
The memory bandwidth has really NOT been keeping up with the GPU processing power, like TFLOPs, etc. Of course, with compression and optimizations, etc, it wasn't too bad, but still, just think about how much faster the GPU would be, if it were optimized with super bandwidth in mind, where nearly everything that the GPU needed would be loaded in one shot.
Remember, when all rumors about HBM1 memory was that it would be 1250 MHz? Then it turned out to be only 1000MHz, and a poor overclocker at that. It was supposed to be like a 5000MHz effective GDDR5 with a 1024-bit bus, as one of the rumors stated. That was a lousy rumor, but it wasn't too far off (1000MHz, with a total of 4096-bit bus width).
The memory bandwidth has really NOT been keeping up with the GPU processing power, like TFLOPs, etc. Of course, with compression and optimizations, etc, it wasn't too bad, but still, just think about how much faster the GPU would be, if it were optimized with super bandwidth in mind, where nearly everything that the GPU needed would be loaded in one shot.
Ok with science that the big bang theory requires that fundamental scientific laws do not exist for the first few minutes, but not ok for the creator to defy these laws... 