Announcement

Collapse
No announcement yet.

NVIDIA Reportedly Near Deal To Buy Arm For $40+ Billion Dollars

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

  • Originally posted by PerformanceExpert View Post

    And if you want to own a Neoverse N1 based server that is faster than Graviton 2, Ampere is happy to sell you one.
    Where? How? The newest product available from Ampere is the eMAG that gets stomped by EPYC, power9 and Intel. https://www.phoronix.com/scan.php?pa...ome-power9-arm They **announced** an product that according to their own (integer only) benchmarks barley reaches current EPYC performance, but will be up against EPYC's next generation.

    Comment


    • Originally posted by QuImUfu View Post

      Where? How? The newest product available from Ampere is the eMAG that gets stomped by EPYC, power9 and Intel. https://www.phoronix.com/scan.php?pa...ome-power9-arm They **announced** an product that according to their own (integer only) benchmarks barley reaches current EPYC performance, but will be up against EPYC's next generation.
      Here for example? EPYC 3 is said to be a 10% throughput gain, nothing amazing, and certainly not enough to compete with the 128-core Altra Max.

      Comment


      • Originally posted by PerformanceExpert View Post

        Here for example? EPYC 3 is said to be a 10% throughput gain, nothing amazing, and certainly not enough to compete with the 128-core Altra Max.
        Only that this site looks like this if they actually offer/stock a product.
        Amperes side says you can only pre-order (that links to the site you linked) or get an evaluation sample: https://amperecomputing.com/ampere-d...tion-programs/
        Sadly these evaluation samples require an NDA, otherwise Phoronix probably would already have proven you wrong and that Altra is a joke compared to EPYC in most workloads.

        Comment


        • Originally posted by QuImUfu View Post

          Amperes side says you can only pre-order (that links to the site you linked) or get an evaluation sample: Sadly these evaluation samples require an NDA, otherwise Phoronix probably would already have proven you wrong and that Altra is a joke compared to EPYC in most workloads.


          We have benchmarks on the prior generations of ampere chips and they were not weak in all workloads vs EPYC. We can except the results to be mixed and we can expect EPYC 3 not to win all the time against Altra.

          Comment


          • Originally posted by oiaohm View Post



            We have benchmarks on the prior generations of ampere chips and they were not weak in all workloads vs EPYC. We can except the results to be mixed and we can expect EPYC 3 not to win all the time against Altra.
            For some workloads that might be true.
            Keep in mind that benchmark is against EPYC Naples, not Rome. Rome performs much better then Naples in performance and power efficiency. And Altra will have to compete with Milan. They more or less admitted that they won't be competing for the performance crown in their performance slide, placing their product minimally faster then Naples, while Milan is said to be ~15% faster.

            Comment


            • Originally posted by QuImUfu View Post
              For some workloads that might be true.
              Keep in mind that benchmark is against EPYC Naples, not Rome. Rome performs much better then Naples in performance and power efficiency. And Altra will have to compete with Milan. They more or less admitted that they won't be competing for the performance crown in their performance slide, placing their product minimally faster then Naples, while Milan is said to be ~15% faster.
              Altra also performs far better than the previous generation and beats Rome. We don't know exactly by how much, but it will be very close to Milan for sure. The Altra Max starts sampling this Q4, and will have the performance crown in 2021. Neither AMD nor Intel have any part on their roadmap that could match it until well into 2022.

              Comment


              • Originally posted by QuImUfu View Post
                For some workloads that might be true.
                Keep in mind that benchmark is against EPYC Naples, not Rome. Rome performs much better then Naples in performance and power efficiency. And Altra will have to compete with Milan. They more or less admitted that they won't be competing for the performance crown in their performance slide, placing their product minimally faster then Naples, while Milan is said to be ~15% faster.
                Might be Naples but some of the wins of the older generation from Ampere against Naples allowing for the gains of Rome and Milan they still lose. So the fact that Altra is meant to have increased performance a lot over even the prior generation at the same clock speed.

                Basically EPYC 3 will not completely beat the older generation of the Ampere stuff at everything so we cannot expect EPYC.3 to always be the best choice over the Altra. We are going to have to wait for Altra to be tapped out to know how often it wins and that is Q4 this year for the 80 core. Yes most vendors have benchmarking restrictions before final production run in the way Ampere is no different.

                Comment

                Working...
                X