Ampere Altra Performance Shows It Can Compete With - Or Even Outperform - AMD EPYC & Intel Xeon
Code compilation can be very scalable to many CPU cores but really depends upon the number of files as part of the source tree for how well it can scale with obviously larger projects generally enjoying more of a boost. In the case of AArch64 there can also be a difference as a result of possible different code paths being followed depending upon architecture. In any event, the Ampere Altra Q80-33 2P proved to be competitive with the EPYC 7742 2P and Xeon Platinum 8280P.
While in the case of compiling the large LLVM code-base the Ampere Altra came a bit behind the Xeon Platinum 8280, the two Q80-33 CPUs were consuming much less than the EPYC 7742 2P and Xeon Platinum 8280 2P processors. The combined CPU power consumption during the LLVM build test was just 145 Watts on average for the Ampere Altra server (and a peak of 341 Watts) compared to a 257 Watt average on the EPYC 7742 2P and 275 Watt average on the Xeon Platinum 8280 2P.
Obviously the cooling can change a great deal depending upon the server platform and its cooling abilities, but for what it's worth the Ampere Altra CPU temperature with Mount Jade was similar to the AMD Daytona reference server and the Gigabyte Storage Server platform used for the Xeon Platinum testing.
In the case of multi-threaded compression/decompression such as with LZ4, the Ampere Altra 2P was performing right in line with the top-end, current-generation AMD EPYC and Intel Xeon CPUs.
But making these numbers even more promising for the Ampere Altra was the significantly better performance-per-Watt based on the real-time power consumption data during the test.