Announcement

Collapse
No announcement yet.

Amazon EC2 Cloud Benchmarks vs. AMD Ryzen, Various AMD/Intel Systems

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

  • #11
    Originally posted by edwaleni View Post
    Benchmarking cloud based VM's is troublesome because it is simply the performance measurement in one point of time on that particular day.
    For what it matters, this is more or less confirming the results of another similar benchmark. https://www.servethehome.com/amd-ryz...ux-benchmarks/

    Comment


    • #12
      Originally posted by tg-- View Post

      I don't think it's a fair point to make.
      You don't pay a monthly subscription with advertised peak values - you pay hourly rates for different performance classes.
      If an hour costs you 40 cents, and the performance drops by 40% in the next hour, why would you pay 40 cents for this one?
      The whole point is to pay hourly rates for hourly compute performance. Why would you accept a one hour job to take 2 hours, especially when physical hardware is payed off in a limited time period and will guarantee you time to complete the job?
      Because it's still far cheaper than buying the actual hardware.

      If you are just doing a few renderings or something per month you won't find much profitable to have your own bigass rendering node as it will be idle most of the time. If the VM ends up costing even twice of its standard rate at random times, it's still insignificant if compared to making/maintaining your own infrastructure for such very spikey load.

      If you need to render stuff on a daily basis then yeah, what you said becomes an issue, but as others said it's a situation where it also starts to become more profitable to just make your own infrastructure and not use the VM at all.

      These services make the most sense for companies that have spikes of activity, not for companies that load them 24/7.

      Comment


      • #13
        Well, the biggest point (IMO) is that it's scalable. So if your demand doubles in a week, you aren't scrambling to assemble servers while your customers get timeout issues. It's an easy way to temporarily expand your capability.

        A lot of traffic is also time sensitive. You might have predictable peaks of use. So if your "hot" period is 6x your average, and only lasts two hours a day, doesn't make sense to have 6x your average in-house capacity that 92% of the day is idle.

        Things like AWS are allowing small businesses to compete with giants. Sure, you pay more, but at least you can play in the field, and not just watch from the sidelines.

        Comment


        • #14
          Originally posted by starshipeleven View Post
          Because it's still far cheaper than buying the actual hardware.

          If you are just doing a few renderings or something per month you won't find much profitable to have your own bigass rendering node as it will be idle most of the time. If the VM ends up costing even twice of its standard rate at random times, it's still insignificant if compared to making/maintaining your own infrastructure for such very spikey load.

          If you need to render stuff on a daily basis then yeah, what you said becomes an issue, but as others said it's a situation where it also starts to become more profitable to just make your own infrastructure and not use the VM at all.

          These services make the most sense for companies that have spikes of activity, not for companies that load them 24/7.
          I agree.
          It still means, that said companies might evaluate when it _actually_ is the cheaper variant, so it's very fair to compare performance and price.
          Owned physical hardware can in many cases be the cheaper and more effective solution, but it certainly doesn't have to be.

          Since for example a Ryzen box can easily outcompete some of the more performant EC2 offerings, it might save money in a rather short period of time.
          If you use it for 3 hours each month it you'll likely do better with one of the cloud services.
          Last edited by tg--; 27 April 2017, 03:47 PM.

          Comment


          • #15
            I just wanted to say a quick "Thank you!" for this comparision. Detailed fact-based comparisons like these are one of the reasons why I subscribe to Phoronix Premium.

            Comment


            • #16
              Originally posted by csett86 View Post
              I just wanted to say a quick "Thank you!" for this comparision. Detailed fact-based comparisons like these are one of the reasons why I subscribe to Phoronix Premium.
              Thanks, glad you liked it!
              Michael Larabel
              https://www.michaellarabel.com/

              Comment


              • #17
                Michael, did you try setting the P-States on the c4.8xlarge, m4.10xlarge, and m4.16xlarge?

                Comment


                • #18
                  This is a great comparison! I understand benchmarking cloud environments can be tricky, but the information can still be valuable. I would have never thought of comparing it to desktop performance, but with Naples not far away this comparison is very interesting. Articles like this are why I subscribe.

                  Would be great to see benchmarking of more cloud providers (DigitalOcean/Linode/Azure/etc) if possible.

                  Comment


                  • #19
                    Originally posted by starshipeleven View Post
                    Because it's still far cheaper than buying the actual hardware.
                    I've checked their pricelists and a mere glance at it screams at you that Amazon don't have even slightest intention of charging you just an workhour.
                    Also, it's not clear how can you be sure that you are consuming just what you see on your end.

                    What if you use say 2-3 hours and then get an invoice for 2.537 workhours ?

                    For smaller loads, owning your gear isn't thaqt big of a deal. Nice Ryzen at 3.6+ GHz with cooler, RAM, mobo and gehouse should be in €700 range. If you are prepared to cut some corners, you can probably have two 65W ones for a €1000. And both can hang on Gig-E, waiting fo a magic packet to boot.

                    No worries for cloud availability and prices, packages, privacy and business secrets, invoicing etc etc.

                    And if you have weaker workstations and do heavy stuff on such central CPU muscle, you can spare some $$$ on workstations too and amortize CPU requirements over those few workplaces, so you can have your own, controlled and optimized minicloud.

                    These services make the most sense for companies that have spikes of activity, not for companies that load them 24/7.
                    If you are, e.g. small producer of mini aircrafts and spend up to €1M fo nice laser deposition 3D printer, and you need to do simulations ( liquid/gas flow, thermal, mechanic effects etc), would you really feel comfortable exposing your work to the friggin cloud ? A those prices, how much of a problem would be to spend €50-100k to a decent CPU/GPU cluster. ?
                    Last edited by Brane215; 28 April 2017, 03:03 AM.

                    Comment


                    • #20
                      Originally posted by starshipeleven View Post
                      The title is a typo, it should have been "Ryzen is quite competitive with Amazon cloud and even Intel's high end stuff".
                      For me, yes. For someone else, something else might fit their equeation better than Ryzen or Cloud. Or Cloud of Ryzens. ;o)

                      Comment

                      Working...
                      X