I gotta say though, after using the SuperMicro MOBO for this series of processors, it's really truly the only MOBO deserving of this chip.
Derek Cooper
2022-07-22 18:22:09 +0000 UTC
Me love zippy
Level1 Techs
2022-07-22 18:18:52 +0000 UTC
At 9:24 into the video, you show nbody CUDA perf. Single precision is shown as running 31.4 times faster than double precision. Any idea why there is such a difference there? E.g., CUDA card, main memory bandwidth, etc.? A factor of 2-8 or so would make sense to me, based on data sloshing stuff, so I am inclined to point my finger at the GPU.