Announcement

Collapse
No announcement yet.

Suggestion

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

  • raydenwins79
    replied
    Thank you for the infos. Its good to read that one memory benchmark is multithreaded and I understand why you are using DirextX and DirectCompute.

    Thank you for the Optimus workaround; I have finally run all tests with the NVIDIA GT 540M,
    but one thing is strange: there were still no DirectX 11 and DirectCompute scores, but this GPU does support both features. The composite 3D score increased from 367 (Intel HD 3000) to 616 (NVIDIA). Do you have an idea why the last two scores are missing?

    Leave a comment:


  • David (PassMark)
    replied
    According to nVidia the best video card should be auto selected by their driver. But this doesn't seem to work some (or maybe even all) of the time. You can force a card to use using this process.
    http://www.passmark.com/support/perf...ia_optimus.htm

    We used DirectCompute for the same reason we use DirectX and not OpenGL. PerformanceTest is Windows only software so it makes sense to use the interface that was designed for Windows. For many years OpenGL was dysfunctional, and development of new features was very slow. This has changed a bit in the last year or so, but years of slow development on OpenGL pushed everyone into DirectX. In any case we are testing the GPU hardware, the choice of using DirectCompute or OpenCL to load up the shader code into the GPU shouldn't greatly impact on the relative performance of the GPU hardware.

    There are seven different memory tests. The one called 'threaded' is multi-threaded. The rest are single threaded.

    Compression test uses an Adaptive encoding algorithm based on source code from Ian H. Witten, Radford M. Neal, and John G. Cleary in an article called “Arithmetic Coding for Data Compression”. The system uses a model which maintains the probability of each symbol being the next encoded. It reports a compression rate of 363% for English text, which is slightly better than the Huffman method. This test reports its results in Kbytes/Sec compressed.

    We looked at using MD5 hashing as part of the encryption test, but instead decided on including SHA256. Which is the newer (more secure) version of a MD5 hash. So the encryption test uses SHA256, AES and a couple of others as well.

    At this point I think it is unlikely we'll be making significant changes to the tests, it will be mostly just bug fixes, help file improvements and graphing improvements from this point until release.

    Leave a comment:


  • raydenwins79
    replied
    another suggestion

    I really love new PT8.0 beta, but I have a laptop running Windows 8 and Sandy Bridge/Intel HD 3000 + NVIDIA GT 540M. I cannot select which GPU should be used for complete tests - this selection seems to be present in advanced tests only. But its a pity, because when I want to create my baseline, only the weaker Intel GPU is used where DX11 and DirectCompute is not supported. Is there a way to implement such a selection?

    Why did you decide to use DirectCompute instead of OpenCL?
    Is the memory benchmark multithreaded?
    Which library is used for compression benchmark?
    Is there a way to suggest additional practical FPU benchmarks such as MD5 checksum creation or H.264 encoding/decoding speed?

    Thank you and please keep up the good work.

    Stephan

    Leave a comment:


  • David (PassMark)
    replied
    We would like to do something like this.
    We are still defining what the online page for each baseline will look like.
    So something like this, or a graph of some sort could form part of the page.

    Leave a comment:


  • godfather
    started a topic Suggestion

    Suggestion

    Hi,
    I recently tried the Performance Test 8 beta and it works well and I like the new features.

    I do have a suggestion. Would it be possible to provide testers with an idea of where they rank compared to other testers? This can come simply in the form of a percentile.

    For example:

    Passmark Rating: 2000 (64th percentile)

    CPU Mark: 5000 (70th percentile)

    etc...

    I know benchmarkers won't be a representative sample of PCs in the world, but I'm sure many would find it interesting to know where they rank.

    Doesn't sound too difficult to implement, so just throwing it out there...

    naturally, this would need to be constrained to a certain time frame. ex: in the past 60 days or so.

    edit: you can also add a rank. for example: you rank 12,323.

    meaning 12,322 tests were run (in history) that yielded a higher score.
    Last edited by godfather; 08-23-2012, 04:17 AM.
Working...
X