Opteron vs. EPYC Benchmarks & Performance-Per-Watt: How AMD Server Performance Evolved Over 10 Years

For those just wondering about the idle power draw of the EPYC 7601 within the Tyan TN70A-B8026, the system while solely idling within Linux was showing an average power draw of about 98 Watts but with some spikes to around 125 Watts. Keep in mind if you load up this server with the 24 x 2.5-inch NVMe SSD slots, etc, your power numbers will vary quite a bit.

Lastly is a look at the AC system power consumption for all of the systems during the entire duration of the benchmarking process:

The minimum power draw of the EPYC 7601, which is basically when briefly idling between test execution, was right in line with the single Opteron 2356/2384 configurations.

The average power draw throughout these many benchmarks is interesting with the EPYC 7601 coming in at right around 180 Watts with this Tyan Transport SX TN70A-B8026 2U server and its SSD, 128GB DDR4 memory, etc. This average power draw of the EPYC 7601 server is lower than the older dual Opteron configuration.

The peak power draw during the most demanding benchmarks was higher but still not too bad at around 350 Watts. The Tyan Transport SX TN70A-B8026 is backed by redundant 770 Watt 80 plus Platinum power supplies.

For those curious about the EPYC power consumption, hopefully these numbers were able to address your questions. If you have remaining questions or would like other tests, feel free to post in the forums. I also hope you found these numbers interesting or insightful in seeing how the AMD server performance-per-Watt and overall system power consumption has evolved over the past decade since the Opteron 2300 days.

I've done some other long-term performance-per-Watt comparisons already this year if interested too like the Athlon II X3 vs. Ryzen 3 and Core i3 2100 vs. Core i3 7100, for those interested in the power efficiency of modern Linux systems.

If you want to see how your own Linux system(s) compare in efficiency against these server results and have a power sensor supported by the Phoronix Test Suite, it's as easy as running PERFORMANCE_PER_WATT=1 phoronix-test-suite benchmark 1709158-TY-OPTYEPYC154 to see how your numbers would compare side-by-side.

Stay tuned for other interesting EPYC Linux performance metrics in the days ahead and thanks to AMD and Tyan for making this EPYC benchmarking possible at Phoronix. And if you didn't see our earlier Intel vs. EPYC 7601 Linux benchmarks, check those out as well.