It’s a silly thing to say now, but it’s kinda been true since Zen 1.
Their CPU architectures have all been designed first and foremost for the server and datacenter, with smaller sets of datacenter parts for consumer products. For a little while they split off CDNA and RDNA trying to make more headway in datacenter GPUs. But that’s being re-merged into UDNA where consumer tech will be cut down server GPUs, like NVIDIA does.
CPUs are somewhat flexible, as desktops, servers, and a few laptops all use the same compute die.
But with GPUs (for the moment), they have to make long term development/production decisions between server dies and gaming dies. It’s not like older GPUs where server cards literally shared the die with Vega, the GTX 780 TI or whatever, and they could essentially slap a different sticker on to change market allocation.
It’s a silly thing to say now, but it’s kinda been true since Zen 1.
Their CPU architectures have all been designed first and foremost for the server and datacenter, with smaller sets of datacenter parts for consumer products. For a little while they split off CDNA and RDNA trying to make more headway in datacenter GPUs. But that’s being re-merged into UDNA where consumer tech will be cut down server GPUs, like NVIDIA does.
Allocation still matters though.
CPUs are somewhat flexible, as desktops, servers, and a few laptops all use the same compute die.
But with GPUs (for the moment), they have to make long term development/production decisions between server dies and gaming dies. It’s not like older GPUs where server cards literally shared the die with Vega, the GTX 780 TI or whatever, and they could essentially slap a different sticker on to change market allocation.