DDR5 (general compute memory) actually isn’t the same as GDDR5 (graphics compute memory). I am no expert in this field, but you can read about the differences on wikipedia. This excerpt sums it up:
Their primary characteristics are higher clock frequencies for both the DRAM core and I/O interface, which provides greater memory bandwidth for GPUs.
Basically, GPUs need a ton of bandwidth but can withstand latency due to the predictable timings of the framerate.
I also assume that because the GPUs processing and memory are on the same board they have less friction in using newer standards. When dealing with CPUs the memory standards need to line up between different physical parts (CPU, motherboard, RAM).
DDR5 (general compute memory) actually isn’t the same as GDDR5 (graphics compute memory). I am no expert in this field, but you can read about the differences on wikipedia. This excerpt sums it up:
Basically, GPUs need a ton of bandwidth but can withstand latency due to the predictable timings of the framerate.
I also assume that because the GPUs processing and memory are on the same board they have less friction in using newer standards. When dealing with CPUs the memory standards need to line up between different physical parts (CPU, motherboard, RAM).