Announcement

Collapse
No announcement yet.

Anyone suggest a suite for Testing entire system for reviewing hardware

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

  • Anyone suggest a suite for Testing entire system for reviewing hardware

    Can anyone suggest a suite for Testing entire system for reviewing hardware?

  • #2
    It depends what type of system you're talking about (low vs. high or something to cover the entire spectrum), how much time you plan to run it, whether you'll just be using or open or closed source graphics drivers, etc. There's lots of variables in coming up with a good suite.
    Michael Larabel
    https://www.michaellarabel.com/

    Comment


    • #3
      Originally posted by Michael View Post
      It depends what type of system you're talking about (low vs. high or something to cover the entire spectrum), how much time you plan to run it, whether you'll just be using or open or closed source graphics drivers, etc. There's lots of variables in coming up with a good suite.

      An overall system test suite. Something that can be used across hardware. Or maybe the best suite for testing a variety of graphics cards and one for testing a variety of Motherboards. That sort of thing, It would need to be varied enough to be used for each piece of hardware I test.

      The drivers will be closed source.

      The Graphics tests should use only the free (as is cost) stuff. It can be closed or open so long as it is free.

      Comment


      • #4
        Originally posted by KRDucky View Post


        An overall system test suite. Something that can be used across hardware. Or maybe the best suite for testing a variety of graphics cards and one for testing a variety of Motherboards. That sort of thing, It would need to be varied enough to be used for each piece of hardware I test.

        The drivers will be closed source.

        The Graphics tests should use only the free (as is cost) stuff. It can be closed or open so long as it is free.
        I'd recommend then just looking at any of my recent Phoronix.com articles to look at the tests I use for each type of review. To make it easy to run then you can run phoronix-test-suite build-suite to pre-configure it with all of the tests.
        Michael Larabel
        https://www.michaellarabel.com/

        Comment


        • #5
          Thanks. I am trying to put a good one together for Teksyndicate. Wendell from TekSyndicate said he would look into using PTS for their hardware reviews instead of the usual fare. I want to give him a suite to get him started. Can you recommend some tests for me to bundle together?

          Comment


          • #6
            For all around testing:

            phoronix-test-suite benchmark xonotic tesseract build-linux-kernel c-ray pgbench sqlite

            Are among the favorites for starters.
            Michael Larabel
            https://www.michaellarabel.com/

            Comment


            • #7
              Thanks. I just got a message from Wendell. This is what he said he needs. Any recommendations for this? I have looked at the tests and suites. So many look redundant. And there are so many different tests.

              the OpenCL SDK raw compute -- I had a helper intern but IIRC we did some bitcoin mining just for testing and/or maybe some type of floating point benchmarks for 32/64 bit from the CLI (64 bit kinda sucks) .
              bonnie++ and hdparm are fun.. hdparm on for example the Linux 750s show only 700-800mbyte/sec but because those are enterprise drives they have multiple h/w i/o queues so you have to have several parallel i/o threads.
              most of the time this is not too scientific. I stumble around and look for unexpected results or weird results then try t figure it out further, then figure out if it is a defect of some kind, or known behavior
              mostly this type of testing is for GPU and storage. some CPU benchmarks but I had a lot of trouble recently -- mostly fixed in kernel 4+ though where speed step was doing all sorts of weird stuff.
              On windows the speedstep results were wildly different than Linux, for example. But I think i7z showed what was going on and the later kernels fixed it up.
              I often wonder if the folks bitching about Nvida or AMD drivers are using one of the several mutual incompatible versions, or a hilariously ancient kernel
              I am rocking kernel 2.6 on a lot of production h/w still, but man the kernel has been really stable as of 3.9/4/etc so I may have to rethink my position on stable kernel versions. lol.

              Comment

              Working...
              X