AmpereOne Performance In The Cloud With Oracle Cloud A2
([Processors] 5 Hours Ago
1 Comment)
- Reference: 0001485769
- News link: https://www.phoronix.com/review/ampereone-oracle-cloud-a2
- Source link:
After talking about [1]AmpereOne for the Oracle Cloud Infrastructure since last year, earlier this month Oracle finally announced general availability on their OCI Ampere A2 instances for tapping into these latest-generation Ampere Computing AArch64 server processors in the public cloud. Here's a brief look at the OCI Ampere A2 performance with AmpereOne compared to their prior A1 instances powered by Ampere Altra.
[2]
With the Oracle OCI Ampere A2 instances they are offering up to 156 core instances, up to 946GB of DDR5 memory, and up to 78 Gbps Ethernet bandwidth and up to 24 vNICs. Oracle's own claims about AmpereOne in their cloud is that there is around 28% more performance on average than the Ampere Altra Max powered A1 instances. Oracle advertises around 13% better performance, around 18% better performance for Nginx, around 27% better performance for PostgreSQL, and 59% better performance for MySQL Point Select.
[3]
One interesting item to note that with the AmpereOne A2 instances, Oracle is treating each "OCPU" as two AmpereOne cores. AmpereOne doesn't employ Simultaneous Multi-Threading (SMT) / Hyper Threading (HT) but they are treating two AmpereOne cores as one OCPU, which is unlike their handling of A1 instances.
[4]
With A2 one Oracle CPU (OCPU) is two cores, one OCPU with A1 is one core, and then for the AMD and Intel x86 instances one OCPU is two vCPUs (SMT/HT included).
[5]
From the Oracle Cloud I ran some benchmarks of the AmpereOne A2 instance with 8 OCPUs (16 vCPUs). I then compared that to the Ampere Altra Max A1 instances at both 8 OCPUs (8 vCPUs) and 16 OCPUs (16 vCPUs). Oracle Linux 9 was running across these instances.
Given I was running these benchmarks on the cloud out-of-pocket at my own expense, I just ran a few basic benchmarks and will leave more benchmarking until having AmpereOne hardware in the lab. Additionally, with Oracle Cloud the CPU power consumption monitoring isn't available for assessing the overall CPU power consumption and power efficiency of these AArch64 server processors. So until being able to dive into the power and performance more closely with actual hardware and compare to the AMD EPYC Bergamo and Intel Xeon Sierra Forest competition, this article was just exploring the A1 Ampere Altra to A2 AmpereOne generational difference.
[6]
[1] https://www.phoronix.com/search/AmpereOne
[2] https://www.phoronix.com/image-viewer.php?id=ampereone-oracle-cloud-a2&image=ampereone_a2_1_lrg
[3] https://www.phoronix.com/image-viewer.php?id=ampereone-oracle-cloud-a2&image=ampereone_a2_2_lrg
[4] https://www.phoronix.com/image-viewer.php?id=ampereone-oracle-cloud-a2&image=ampereone_a2_3_lrg
[5] https://www.phoronix.com/image-viewer.php?id=ampereone-oracle-cloud-a2&image=ampereone_a2_4_lrg
[6] https://www.phoronix.com/image-viewer.php?id=ampereone-oracle-cloud-a2&image=ampereone_a2_51_lrg
[2]
With the Oracle OCI Ampere A2 instances they are offering up to 156 core instances, up to 946GB of DDR5 memory, and up to 78 Gbps Ethernet bandwidth and up to 24 vNICs. Oracle's own claims about AmpereOne in their cloud is that there is around 28% more performance on average than the Ampere Altra Max powered A1 instances. Oracle advertises around 13% better performance, around 18% better performance for Nginx, around 27% better performance for PostgreSQL, and 59% better performance for MySQL Point Select.
[3]
One interesting item to note that with the AmpereOne A2 instances, Oracle is treating each "OCPU" as two AmpereOne cores. AmpereOne doesn't employ Simultaneous Multi-Threading (SMT) / Hyper Threading (HT) but they are treating two AmpereOne cores as one OCPU, which is unlike their handling of A1 instances.
[4]
With A2 one Oracle CPU (OCPU) is two cores, one OCPU with A1 is one core, and then for the AMD and Intel x86 instances one OCPU is two vCPUs (SMT/HT included).
[5]
From the Oracle Cloud I ran some benchmarks of the AmpereOne A2 instance with 8 OCPUs (16 vCPUs). I then compared that to the Ampere Altra Max A1 instances at both 8 OCPUs (8 vCPUs) and 16 OCPUs (16 vCPUs). Oracle Linux 9 was running across these instances.
Given I was running these benchmarks on the cloud out-of-pocket at my own expense, I just ran a few basic benchmarks and will leave more benchmarking until having AmpereOne hardware in the lab. Additionally, with Oracle Cloud the CPU power consumption monitoring isn't available for assessing the overall CPU power consumption and power efficiency of these AArch64 server processors. So until being able to dive into the power and performance more closely with actual hardware and compare to the AMD EPYC Bergamo and Intel Xeon Sierra Forest competition, this article was just exploring the A1 Ampere Altra to A2 AmpereOne generational difference.
[6]
[1] https://www.phoronix.com/search/AmpereOne
[2] https://www.phoronix.com/image-viewer.php?id=ampereone-oracle-cloud-a2&image=ampereone_a2_1_lrg
[3] https://www.phoronix.com/image-viewer.php?id=ampereone-oracle-cloud-a2&image=ampereone_a2_2_lrg
[4] https://www.phoronix.com/image-viewer.php?id=ampereone-oracle-cloud-a2&image=ampereone_a2_3_lrg
[5] https://www.phoronix.com/image-viewer.php?id=ampereone-oracle-cloud-a2&image=ampereone_a2_4_lrg
[6] https://www.phoronix.com/image-viewer.php?id=ampereone-oracle-cloud-a2&image=ampereone_a2_51_lrg