While the latest Intel Core Ultra processors have done away with Hyper Threading (HT), Intel Xeon CPUs continue supporting HT/SMT, including with their latest Xeon 6300 series budget server processors. As the new AMD EPYC 4005 “Grado” processors also support Simultaneous Multi-Threading (SMT) and can be found at the same core/thread count count as the flagship Xeon 6369P processor, it makes for an interesting look at comparing the SMT/HT performance impact and power efficiency. Here are some benchmarks showing the Xeon 6300 against the AMD EPYC 4005 in SMT performance.
Over the past week I carried out a fresh look at the Intel vs. AMD SMT/HT performance with the Xeon 6300 series continuing to support Hyper Threading unlike the newest Arrow Lake and Lunar Lake desktop/mobile processors foregoing HT. On the AMD side, SMT has continued to prove very effective from mobile/desktops to servers. Just a few weeks ago was looking at the great SMT advantage with the AMD Ryzen AI Max “Strix Halo” while SMT is very beneficial too for high core count AMD EPYC servers.
As for the testing today, the top-end Intel Xeon 6300 series processor is the Xeon 6369P providing 8 cores / 16 threads with HT enabled. With the AMD EPYC 4005 line-up there is the top-end EPYC 4565P and EPYC 4585PX processors with 16 cores / 32 threads. For making this SMT/HT comparison as 1:1 as possible, the AMD EPYC 4345P was used that is an 8-core / 16 thread CPU like the Xeon 6369P.
With the AMD EPYC 4345P and Intel Xeon 6369P they were tested both at their defaults (SMT enabled) and then again after rebooting the system with SMT disabled so just 8 cores/threads. Both the AMD EPYC 4005 and Intel Xeon 6300 series server processors were tested from similar Supermicro servers and with 2 x 32GB DDR5 ECC memory (Xeon 6300 being limited to DDR5-4800 while EPYC 4005 allows DDR5-5600), Solidigm P41 Plus NVMe SSD storage, and running on Ubuntu 25.04 with the Linux 6.14 kernel.
While running dozens of different multi-threaded benchmarks on each AMD EPYC and Intel Xeon processor with SMT on/off, the CPU power consumption was also monitored on a per-test basis for seeing the SMT/HT impact on power efficiency.
