Jump to content
HWBOT Community Forums

FM_Jarnis

Members
  • Posts

    188
  • Joined

  • Last visited

Posts posted by FM_Jarnis

  1. Okay, SystemInfo 4.20 is now available

     

    http://www.futuremark.com/downloads/Futuremark_SystemInfo/latest

     

    SystemInfo Version 4.20

    Updated September 4, 2013

     

    * Updated CPUID module for improved hardware detection

    * Updated AMD ADL library for improved compatibility with latest hardware

    * Improved hardware scan speed by about 50%

    * Added Windows RTC clock verification, results with inaccurate time measurement can now be rejected on 3DMark.com

     

    ----

     

    ...and yes, the last bit means that if you try doing some monkey business with BCLK in Windows 8 on-the-fly and break the time-space continuum of your system (RTC clock skew), 3DMark.com will reject the result as invalid.

     

    We're considering the option of adding this also "built-in" to the benchmark for at least 3DMark (so online viewing would not be required) but patching the benchmark itself is a very long and complex process so this is best we could do on a short notice.

     

    Also doing it like this gives us a "free" benefit - this validation check now applies to ALL Futuremark benchmarks that use SystemInfo (so 3DMark 03, PCMark05 and everything newer than that).

     

    Report any issues you may have to FM support ( http://support.futuremark.com )

     

    Steam versions will be updated early next week but you can already manually install this standalone update before that even when using Steam version.

  2. We started simple with 3DMark on Android on purpose - we wanted to ship a benchmark that is useful to the industry (who couldn't really care less about hardware detection) and developing full blown "SystemInfo"-style component which actually works on all the SoCs out there would have taken far too long. And it's not just doing that on Android - we have iOS and RT to think about as well (those are coming soon - together with Android update that adds another test mode, which is the reason why iOS and RT got delayed over the summer holidays...)

     

    We're actually discussing with the programmer of CPU-Z (who just got Android version out) on the possibility of integrating his work to 3DMark on Android as well, so who knows... We're already pretty annoyed that CPU core frequency is not detected as same SoC may be running at various clock speeds depending on the device and we can only show "up to X GHz" which is the listed max for the SoC in question.

     

    And yes, there was little point in having any kind of online functionality or "toplists" without any kind of proper hardware detection. So we instead went with the "hardware channel" comparison featureset - offering score comparison against averages out of all scores (where any cheats and other outliers get buried in the vast sea of stock results)

  3. Just updated it, the Steam patch is differential, weigh in at 84.6MB :D

     

    Yep. Sadly there is no standalone differential patcher as the InstallShield version used cannot do binary patches to files and the main DAT file was modified so... it would mean a huge installer even for just a patch.

     

    So stand-alone is only available as full installer.

     

    We're actively looking into using another installer in the future that can handle these things better.

  4. Each card requires all assets (they cannot share memory).

     

    As I understand the problem, in case where there are GPU physics simulation going on, each card has to have a copy of the simulation for each card. So in 4x SLI/Crossfire. Every card has to contain the state and assets of all four simulations (or really bad things happen...) - this takes extra RAM as stuff is copied between cards.

     

    Mostly this shows up as small performance penalty for 2GB cards on Fire Strike Extreme in triple/quad sli/crossfire. Two card sli/crossfire is not a problem and even three might not show any major penalty but four and extreme preset pretty much needs 3GB cards to ensure this limitation doesn't influence performance.

     

    Performance preset is no problem unless you have only 1GB cards (should be a fairly rare case these days)

     

    ...and even then it will work. It just won't have quite as good performance as cards with more memory would have.

  5. Looks like Hall of Fame didn't get the memo about 1.1 yet... We'll have to fix that tomorrow. In the meanwhile, here are direct links to best 1.1 scores so far;

     

    Fire Strike:

     

    http://www.3dmark.com/hall-of-fame-2/fire+strike+3dmark+score+performance+preset/version+1.1

     

    Fire Strike Extreme:

     

    http://www.3dmark.com/hall-of-fame-2/fire+strike+3dmark+score+extreme+preset/version+1.1

     

    I would imagine that the first run with 4x GTX Titan will be a new WR :)

  6. BTW on a related note, if you are doing records on 4GPU with Fire Strike Extreme, note that each card should have at least 3GB onboard memory or you will see a performance hit from GPUs running out of memory.

     

    Non-extreme Fire Strike should be fine with 2GB per card.

     

    (single GPU still works fine with 1GB on performance and 1.5GB on extreme but per card memory use goes up with multi-GPU)

  7. Time to re-do those world records, now with 4 GPUs..

     

    ---

     

     

     

    What's New in 3DMark Windows Edition v1.1.0

     

    This update adds the Ice Storm Extreme test and fixes issues when testing systems with multiple GPUs. 3DMark scores will increase slightly on systems with two GPUs and significantly on systems with three or four GPUs.

     

    NEW

    The Ice Storm Extreme benchmark test has been added to 3DMark Advanced Edition. Ice Storm Extreme is a Direct3D feature level 9 benchmark test for Windows tablets, ultrabooks and notebooks. Ice Storm Extreme raises the rendering resolution from 720p to 1080p and uses higher quality textures and post-processing effects in the graphics tests to create a more demanding load for the latest mobile devices. You can compare Ice Storm and Ice Storm Extreme scores from Windows and Android devices.

     

    FIXED

    - 3DMark now works correctly on systems with up to four GPUs.

    - Fixed the issue caused by Windows update KB2670838, which added partial DX11.1 support to Windows 7.

    - Fixed a problem with the bloom post-processing effect when using very high rendering resolutions in custom settings.

     

    Our mirrors should start deploying the new installer over the next hour or so. Steam version is already available. There is no differential installer - main DAT file was modified so the patch installer would be almost as large as the full installer anyway.

  8. Didn't post about this because been just a tad busy with 3DMark and PCMark related things but I guess word spreads quickly.

     

    If it breaks, let us know :) Currently it's somewhat experimental and error messages can be somewhat misleading (tho if you ping me with the problem ID, I can look the issue up and give more detail)

  9. Couple of small benchmark updates for older 3DMarks are now available. Also SystemInfo 4.17 is now available. New 3DMark will get it's first patch also soon but that is still being tested and tweaked.

     

    -----

    3DMark 11 v1.0.5

     

    This version fixes a few minor issues. Benchmark scores are not affected by this update.

     

    Compatibility

    - SystemInfo updated to version 4.17 to improve compatibility with latest hardware.

     

    Bug fixes

    - Fixed display mode identification for systems with a very large number of supported display modes.

     

     

     

    3DMark Vantage v1.1.2

     

    This update adds Windows 8 compatibility and fixes a few bugs. Benchmark scores are not affected by this update.

     

    Compatibility

    - Fixed Windows 8 compatibility issues with online result service connection.

    - SystemInfo updated to version 4.17 to improve compatibility with latest hardware.

    - Updated PhysX software to version 9.12.1031.

     

    Bug fixes

    - Fixed various GUI bugs.

    - Updated exporter.dll to fix problem loading saved results.

     

    Professional Edition only

    - Added support for infinite looping through the command line interface.

    - Added save and submit result options to the command line interface.

  10. You are using an obsolete CPU-limited benchmark on a modern system that was designed for GeForce 6800/7800 age of cards.

     

    06 game tests are nothing more than badly-threading CPU tests at this point on 670M. Use 3DMark Vantage, 3DMark 11 or new 3DMark.

  11. Maybe this will be fixed in the future, but for now:

     

     

    If you submit a 3DMark result to Futuremark..... that is your one and ONLY shot with that particular save file. If you submit it to FM, decide to delete it then change your mind..... tough. If you submit a result, just keep it in your account.

     

    This is by design and part of our improvements towards better result file security - every result has unique ID and that unique ID can only appear in the DB once. If result is deleted, the ID still remains (in the table of deleted results).

     

    We recommend using "hide" instead of "delete" to get a result off various searches and lists.

     

    If you accidentally delete a result that is important to you, contact FM support with details - we can restore recently deleted results.

  12. In general the scores should go up and they go up substantially enough that the scores cannot be compared to previous versions.

     

    The only case where they could go down is with storage that uses compression on the hardware level and somehow benefited about it a lot on the previous version -then the storage test might go down a few percent. On normal HDD or SSD there should be no major differences here either.

  13. PCMark 7 v1.4.0 released

     

    PCMark Vantage v1.2.0 released

     

    These feature official Win 8 compatibility. PCMark 7 has some major bugixes and scores of v1.4.0 are not comparable with older versions. Video transcoding test, web browsing test and storage test all feature bug fixes that influence scoring.

     

    PCMark Vantage has some minor fixes, no changes in scores there. Windows 8 full compatibility for PCMark Vantage requires Windows 8 Media Center Pack as by default Win 8 does not include a MPEG2 codec.

     

    Also note that this update ditches PCMark Vantage trial version - the basic edition is now freely available with no limitations on number of runs and works the same as, for example, PCMark 7 Basic Edition.

  14. It actually boils down to "can we enforce the rules we make?", rather than a philosophical debate. We all want the image quality maintained, but if you can't check if a certain IQ "optimisation" was used ... not possible to enforce.

     

    Why don't Nvidia and AMD, two members of your BDP program, provide you with all the information to ensure image quality?

     

    Most likely reason? Because our benchmarks are primarily tools for them and their customers and the press. None of these would even consider "cheating" with IQ-degrading tweaks when benchmarking hardware as it would pretty much defeat the whole point. From their perspective, this is a complete non-issue.

     

    The only case where the problem raises its ugly head is when benchmarks are used for hobby/competitive benchmarking and the objective is no longer to compare hardware performance (or test system stability) but to get biggest possible numbers.

     

    It definitely is a known use case for 3DMark and we've done a lot in that regard already (all the online submission validation code is there almost entirely to support this). The unfortunate problem is that without server side code doing validations, any such checks would be absolutely trivial to hack. And you guys seem to eminently dislike having a network connection while benchmarking.

     

    The only theoretical option for such validations offline would be to have punkbuster-style monitoring processes running in the background, trying to figure out if someone is messing up with the benchmark but such system would be very complicated to code, hard to maintain and almost impossible to do without influencing benchmark results - something that would diminish the usefulness as a tool for BDP, OEM manufacturers and the press.

     

    For 3DMark we added in-UI support for online result validation so you can see result validity for screenshot purposes without opening a browser. While there are no new checks (the checks are same as for 3DMark 11), it would be possible to expand this further, for example against the LOD tweaks. But again, it has to be server side or it is pretty much pointless. And since HWBot community seems to be against network connection requirement, it is hard to get through an argument inside FM that we should dedicate resources to do, for example, NVIDIA LOD setting checks for 3dmark.com result validation. Nobody else except competitive/hobby "record breaker" benchmakers have any need for it and notable competitive benchmarkers refuse to use it (because you must be online), so... why do it?

     

    Just trying to explain the situation from our perspective...

  15. FM view is this; If you change something in driver/registry that changes the visuals of the test, you are changing the test and the score becomes fairly pointless.

     

    Of course it boils down to "are we trying to get big numbers out of a piece of software?" or "are we trying to tweak hardware to perform a well-defined and repeatable workload as fast as possible?"

  16. That's just how InstallShield does it - if it already finds SI 4.15 installed, it skips the installation of that. Any doubts about it being correctly installed is fixed by the usual uninstall->reinstall.

     

    Any older version and it will update it.

  17. Combined test has always been a problem. The 6970 had the same Combined issue you're describing. As far as I can remember it was indeed because of a driver issue.

     

    What happens if you disable SystemInfo and manually select the amount of cards to use for rendering? Just run combined.

     

    If you do manual GPU count select to something other than the actual GPU count in system that is going to participate in rendering, you just make the combined test completely pointless. It is a debug option only for cases where you get somehow bogus results with the autodetect (most commonly, 2-4 GPUs in system but Crossfire not enabled on purpose would require manual setting to 1 to get correct single card score).

     

    6xxx series Radeons have a driver issue of some sort with 3DMark 11 combined test - more than 3 cards combined test score suddenly tanks. We've told AMD. I actually reminded AMD person about this just today.

     

    This doesn't happen on 7xxx series, by the way (nor on any NVIDIA setups). It is specific to HD 6xxx cards and specifically with 4GPUs.

  18. That's not true I was on 13.1 when 3dmark came out and was they weren't approved but the 13.2 beta where lol

     

    13.1 did go through approval process - it actually did so before the launch.

     

    It was specifically NOT approved because it didn't render the Fire Strike test correctly (part of the particle effects were missing). 13.2 betas fixed this problem. Approval is not just a rubber stamp; the driver actually has to render things correctly.

  19. Overclockers -love- beta drivers :D

     

    //edit: how can you not see LOD as a major sticking point for the 3DMark benchmark? Tesselation "optimisations" are clearly invalid for 3DMark submissions, but the LOD adjustments are not? Why?

     

    Tessellation adjustments are in default driver UI. You could do these without realizing that you are messing up the benchmark load -> high priority to recognize and flag.

     

    LOD tweaks require third party tools mucking up with internal settings that the vast majority of people have no idea about -> not so high priority.

×
×
  • Create New...