Yes, synthetic benchmark results in general are to be taken with a grain of salt, leaving a lot of potential for things like these that can easily produce the r

Yes, synthetic benchmark results in general are to be taken with a grain of salt, leaving a lot of potential for things like these that can easily produce the result you desire than what really is.. there is a big question of how this translates to what your workload needs/does. We just bought our first AMD Epyc server for our HPC cluster. I cant wait until next week to test it against our Skylake nodes. 1 socket EPYC 16C cpu vs 2X Xeon Skylake. I already have a suite of our internal workloads to hammer them with.. one of them is a key in house software compiled usually with intel compiler and intel mkl for performance reasons. Really curious how it will do with the EPYC cpus.