AmpereOne A192-32X Benchmarks: 192 Core ARM Server Performance & Power Efficiency
The server processors tested for this initial AmpereOne comparison with all servers running Ubuntu 24.04 LTS included:
- Xeon Platinum 8490H
- Xeon Platinum 8490H 2P
- Xeon Platinum 8592+
- Xeon Platinum 8592+ 2P
- Xeon Max 9468
- Xeon Max 9468 2P
- Xeon Max 9480
- Xeon Max 9480 2P
- Xeon Platinum 8380 2P
- Xeon Platinum 8380
- EPYC 9684X 2P
- EPYC 9684X
- EPYC 9654
- EPYC 9654 2P
- EPYC 9754
- EPYC 9754 2P
- EPYC 8534P
- EPYC 8534PN
- Xeon 6766E 2P
- Xeon 6766E
- Xeon 6780E
- Xeon 6780E 2P
- Ampere Altra Max M128-30
- AmpereOne A192-32X
Each server was running with memory at its maximum supported channels and memory speed. All of the EPYC and Xeon testing were freshly conducted over the summer on Ubuntu 24.04 LTS. The Ampere Altra Max testing was carried out using the Gigabyte G242-P36 server kindly provided by Giga Computing. The AmpereOne server was running with the Ubuntu 64K page size kernel for optimal server/HPC AArch64 performance.
With each configuration the CPU package power consumption was monitored for generating CPU power consumption and power efficiency metrics. For the Ampere Altra Max and AmpereOne servers the total AC power consumption (wall power) was also monitored. There weren't power consumption reporting for the Xeon or EPYC servers due to using the reference server platforms for those tests. Intel/AMD/Ampere all recommend against total power consumption reporting for reference servers due to typically not being reflective of production/retail server platforms that are better tuned for power efficiency. With the AmpereOne and Ampere Altra Max both being tested on production server platforms, the total system power consumption was monitored there but unfortunately I don't have any non-reference Intel/AMD motherboards/servers for the other platforms tested.
So let's move on with the long-awaited look at AmpereOne server performance and power efficiency.