Feels like DDR5 capabilities have been running away from what boards and memory controllers can reasonably support, bar the lowest amount of ranks/channels and non-optimal clock ratios.
And it’s so painful for me since my workload is completely memory bandwidth bound but does not scale across multiple sockets, much less multiple nodes.
A specialized computational fluid dynamics program. And nominally CFD is great to scale across many sockets/systems but it seems this particular program wasn’t written to be able to do that.
Feels like DDR5 capabilities have been running away from what boards and memory controllers can reasonably support, bar the lowest amount of ranks/channels and non-optimal clock ratios.
And it’s so painful for me since my workload is completely memory bandwidth bound but does not scale across multiple sockets, much less multiple nodes.
96 cores per socket? 😎
12 channels of ram per socket? 🤩
DDR5-4800? 🫠
Just curious, but what workload doesn’t scale across multiple sockets? Gaming?
A specialized computational fluid dynamics program. And nominally CFD is great to scale across many sockets/systems but it seems this particular program wasn’t written to be able to do that.
Ddr6 and new sockets and boards going to come out very soon probably
DDR6 is years away.