Announcement

Collapse
No announcement yet.

Benchmark for 12th Gen CPUs

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

  • Benchmark for 12th Gen CPUs

    For some of the 12th gen CPUs, I always see the warning note that "Baseline has been excluded from average results due to anomalies in the submitted results. [Num CPU Processes - Tested: 24, Expected: 16]". I have two questions for this.

    Q1: I know many people want to test their CPUs / GPUs for several times so that they could submit the BEST results. Unfortunately, it seems that PassMark only gives each person / computer one chance to submit, and once you have done so, you cannot submit a new score, even if your new benchmark score is higher. I know that PassMark requires a new benchmark score to be at least 5% higher than the previous benchmark score before it can be resubmitted, but this is based on the score of the whole computer, not some specific part of that computer, e.g. considering only the CPU or GPU etc, so it is very hard to submit your new exciting CPU score (for example, 20510 points comparing to the previous 18520 points) when your whole PC's score is not 5% higher than before.This leads to a lot of inexperienced people submitting low scores and underestimating the performance of the hardware. I wonder why Passmark cannot allow new scores to be submitted but has to set such restrictions? This is kind of not very user-friendly and statistically not inclusive / accurate.

    Q2: Why this happened for the i9-12900K, i7-12700K, and i5-12600K? "Num CPU Processes - Tested: 24, Expected: 16" - is this because PassMark needs to be updated to fully support the big.LITTLE hybrid architecture of the Intel 12th gen CPUs?



  • #2
    1. The 5% difference check was used to prevent people constantly submitting multiple results to manipulate the score. This will be removed in the next build (probably next week).

    2. PerformanceTest is in the process of being updated to support 12th gen Intel CPUs. Web site is already mostly updated, but it depends on PT as well, as that is where the data comes from.

    Comment


    • #3
      Originally posted by Simon View Post
      1. The 5% difference check was used to prevent people constantly submitting multiple results to manipulate the score. This will be removed in the next build (probably next week).

      2. PerformanceTest is in the process of being updated to support 12th gen Intel CPUs. Web site is already mostly updated, but it depends on PT as well, as that is where the data comes from.
      Thank you for the nice advice! Wish PassMark will do better and better.

      Comment


      • #4
        Does this mean the single-thread performance benchmark for the 12900K is not correct?

        Rank: 1
        No. of Samples: 13
        Cores: 8 Threads: 16

        Looks like the Performance Test is not adding the efficient cores, 8, and only posting performance cores 8.

        Cores should say 16.



        Comment


        • #5
          Does this mean the single-thread performance benchmark for the 12900K is not correct?
          No, the benchmark is correct.

          The above discussion was just around the correct display of the different core types. The page for the 12900K now displays the core count & breakdown correctly. But the display issue never affected the single threaded benchmark. But you are correct we should also change the total core count displayed in the graph.

          Comment


          • #6
            The cores and thread core in the drop down has been updated to take into account the efficient cores as well.

            Comment


            • #7
              Can you explain why the new desktop scores are very low for the 12900k. A 3950X is beating the 12900k.

              https://www.cpubenchmark.net/desktop.html

              But the overclocked scores are very high for the 12900k.

              https://www.cpubenchmark.net/overclocked_cpus.html


              Thanks

              Comment


              • #8
                The Performance core / Efficiency core split is a new thing.
                Depending on the software application and the operating system version some load distribution between the P & E cores can happen automatically. It is further complicated by the fact that now only some cores have HyperThreading. This has the potential to confuse a lot of applications, that previous assumed that virtual cores were ALWAYS double physical cores if HT was available. In short it has suddenly got a lot more complex to code for optimal performance (or optimal efficiency) on an arbitrary CPU.

                We have been trying to buy a 12900K / DDR5 system for testing, but there is zero stock worldwide. So we have been unable to do any testing yet on real hardware.

                But nonetheless, you can expect some real world apps to be sub-optimal on this CPU. i.e. more variable levels of performance compared to other releases of new CPU families.
                Some developers will take the time to optimise their app (but is it a fair amount of effort) and the O/S itself might be smarter at load allocation.

                We will probably be doing a new patch release of PT next week. We aren't sure at this point how much more performance we can get out of these CPUs.


                Comment

                Working...
                X