Announcement

Collapse
No announcement yet.

Benchmarks Of 2nd Gen AMD EPYC On Amazon EC2 Against Intel Xeon, Graviton2

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Benchmarks Of 2nd Gen AMD EPYC On Amazon EC2 Against Intel Xeon, Graviton2

    Phoronix: Benchmarks Of 2nd Gen AMD EPYC On Amazon EC2 Against Intel Xeon, Graviton2

    Today AMD and Amazon announced the general availability of 2nd Gen AMD EPYC "Rome" processors available via the Elastic Compute Cloud. AMD EPYC "Rome" on EC2 with the new "C5a" instance types offer very competitive performance against the latest Intel Xeon instance types, Amazon's own Graviton2 Arm-based instances, and a big upgrade compared to the first-generation EPYC processors in the cloud.

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    Pricing?

    Comment


    • #3
      Once place r5a instances fell apart was memcached performance. The high memory latency made 7xx1 a terrible fit.

      Comment


      • #4
        Yeah, what's the performance/dollar? Sounds like a relevant metric?

        Comment


        • #5
          I'm more interested vs bare metal performance, as I see a lot of discussion over renting vs ownership for IT business in my country as the laws of personal data protection will be enforced next year and data must be hosted locally in the country

          Comment


          • #6
            Originally posted by smitty3268 View Post
            Pricing?
            "That's the quick look at the raw performance of these new AMD EPYC C5A instances. More tests and looking at the performance-per-dollar pricing will come in the weeks ahead."

            Comment


            • #7
              Originally posted by milkylainen View Post

              "That's the quick look at the raw performance of these new AMD EPYC C5A instances. More tests and looking at the performance-per-dollar pricing will come in the weeks ahead."
              I was just interested in what the raw prices were, not any kind of analysis. I looked it up myself here https://aws.amazon.com/ec2/pricing/on-demand/:

              Arm m6g.8xlarge $1.232
              EPYC c5a.8xlarge $1.232
              EPYC m5a.8xlarge $1.376
              Xeon m5.8xlarge $1.536
              Arm m6g.16xlarge $2.464
              EPYC c5a.16xlarge $2.464
              EPYC m5a.16xlarge $2.752
              Xeon m5.16xlarge $3.072

              Interesting. Seems they set the prices for these rome configs exactly the same as the tested ARM options, and cheaper than the prior generation Epyc servers were.
              Last edited by smitty3268; 05 June 2020, 02:56 AM.

              Comment


              • #8
                The c5a instances are meant to be an alternative for the c5 instances. Why did you not benchmark c5a vs c5?
                Last edited by BS86; 05 June 2020, 04:36 AM.

                Comment


                • #9
                  Originally posted by smitty3268 View Post

                  I was just interested in what the raw prices were, not any kind of analysis. I looked it up myself here https://aws.amazon.com/ec2/pricing/on-demand/:

                  Arm m6g.8xlarge $1.232
                  EPYC c5a.8xlarge $1.232
                  EPYC m5a.8xlarge $1.376
                  Xeon m5.8xlarge $1.536
                  Arm m6g.16xlarge $2.464
                  EPYC c5a.16xlarge $2.464
                  EPYC m5a.16xlarge $2.752
                  Xeon m5.16xlarge $3.072

                  Interesting. Seems they set the prices for these rome configs exactly the same as the tested ARM options, and cheaper than the prior generation Epyc servers were.
                  Prices are more than likely heavily subsidized by Amazon for their own solution.
                  You can't make your own ultra-high-end CPU, motherboards etc and fab it at low volume thinking it will compete price-wise with families made in counts of millions.
                  Even if you could make it dirt cheap, RnD costs over units in this volume range is stupid high.

                  That being said, I don't think the customer could care less.
                  Price is price. But it's an interesting situation nonetheless.

                  Comment


                  • #10
                    Originally posted by smitty3268 View Post

                    I was just interested in what the raw prices were, not any kind of analysis. I looked it up myself here https://aws.amazon.com/ec2/pricing/on-demand/:

                    Arm m6g.8xlarge $1.232
                    EPYC c5a.8xlarge $1.232
                    EPYC m5a.8xlarge $1.376
                    Xeon m5.8xlarge $1.536
                    Arm m6g.16xlarge $2.464
                    EPYC c5a.16xlarge $2.464
                    EPYC m5a.16xlarge $2.752
                    Xeon m5.16xlarge $3.072

                    Interesting. Seems they set the prices for these rome configs exactly the same as the tested ARM options, and cheaper than the prior generation Epyc servers were.
                    Originally posted by milkylainen View Post

                    Prices are more than likely heavily subsidized by Amazon for their own solution.
                    [...]
                    Price is price. But it's an interesting situation nonetheless.
                    c - instances have only half the RAM that m - instances have at the same core-count. So comparison is a bit off here.
                    Last edited by BS86; 05 June 2020, 04:41 AM.

                    Comment

                    Working...
                    X