Jump to content
HWBOT Community Forums

_mat_

Members
 View Profile  See their activity

  • Posts

    1003

  • Joined

  • Last visited

  • Days Won

    41

 Content Type 

Posts posted by _mat_

    The official BenchMate support thread

    in Benchmark software

    Posted

    The result was made using a graphics card that can't be correctly matched or the integrated HWiNFO library can't read the GPU's sensors correctly.

    That's why the GPU and GPU memory frequency rows are marked as red (invalid).

    Please submit a bug report from within BenchMate (in the menu of the main dialog) after you have finished a 3DMark run that would lead to an error like this.

    Thank you in advance!

    The official BenchMate support thread

    in Benchmark software

    Posted

    Thanks for your kind words. I will integrate y-cruncher and x265 with the next release. Both are a little bit harder to integrate because their wrappers are built with Java, which is impossible to secure. So another solution has to be found, I'm not sure yet how this will turn out, but I'll find something.

    As for the SuperPi bug. I have encountered this but could never reproduce it two times in a row. Thanks for the bug report, I will look into it!

    • Like 1

    The official BenchMate support thread

    in Benchmark software

    Posted · Edited by _mat_

    1 hour ago, ADVenturePO said:

    If I want to switch to BM 0.10.5 can I copy my results from old BM "results" folder to a new one? 
    Will it create some problems or can I do It?

    Yes, you can move your old results into the new BenchMate installation.

    You can always move/copy the following files to a new BenchMate installation directory. Close BenchMate first, then move/copy, start BenchMate again and all your benchmarks and results should up after restarting BenchMate.

    Old-BenchMate\results\*          =>    New-BenchMate\results\
    Old-BenchMate\config.json      =>    New-BenchMate\

    • Thanks 1

    The official BenchMate support thread

    in Benchmark software

    Posted · Edited by _mat_

    4 hours ago, ADVenturePO said:

    Would you like me to report differences in results made through Benchmate and without it or you do not consider it a bug of any sort?

    It's not a bug, security always has its performance penalty. Although there is next to no impact on most benchmarks, wPrime is different here. That's why I initially refused to add it, but got convinced by the popular vote.

    wPrime is coded in Visual Basic 6, a programming language not suited for multi-threaded applications. To achieve threading it uses a COM object server, which actually starts two wPrime processes per thread to calculate and communicate the progress and result. The startup of these processes is part of the time measurement as well, so when you run wPrime you are also benching the COM object client spawning in addition to the normal workload.

    To secure all these processes, BenchMate has to add a security layer to each process. This is normally done only once when the benchmark application is started. On wPrime with Threadripper this has to be done dozens of time. I optimized the hell out of this, but I can do only so much.

    In any case, you can't bench wPrime with AMD on Windows 10, so you should compare it against a Windows 7 score.  According to the scoreboard it seems like Windows 10 + BenchMate is still winning.

    • Like 1

    The official BenchMate support thread

    in Benchmark software

    Posted · Edited by _mat_

    @MrGenius

    Thanks to your bug reports I could fix two problems. The QPC Skew "bug" led to an improved auto detection algorithm to restrict QPC usage in benchmarks depending on the hardware platform and OS used.

    Additionally I've added the possibility to configure the Timer Emulation used for benchmarks. This is only available for all CINEBENCH and 3DMark versions as they depend on very fast timer facilities to avoid a negative impact on performance. Please check the changelog for more details: https://benchmate.org/changelog.html (see version 0.10.4)

    The skewing of QPC on Windows 7 is actually not a bug! All Intel CPUs below Skylake will show wrong benchmark results if QPC is used for timing and QPC itself just wraps the CPU's TSC. The assumption that all is safe on Windows 7 was never justified. We are just lucky that most benches are so old, that they don't use QPC. Unfortunately that is not true for Geekbench 3, Geekbench 4 and Geekbench 5. Or for example the new Intel XTU 6.5. All of these benchmarks can NOT be benched on Pre-Skylake platforms on Windows 7. Same goes for AMD of course.

    Regarding your statement on Geekbench 5, I've set these categories back to "Alpha" status, which hides them from HWBOT's benchmark overview page. There were only a handful of results and the whole category depends on the automatic submission by BenchMate, which is no longer supported.

    • Like 1

    The official BenchMate support thread

    in Benchmark software

    Posted

    BenchMate 0.10.5 is available for download. ?

    579614929_BenchMate-TimerEmulation.thumb.png.9003e4cc9f85f6f4d7e0101349c7728b.png

    This is probably the last release for 2019 as I am as exhausted as one can be. It was a good year though, but as unhealthy as an out-of-control drug addiction. ?

    I haven't announced the last two minor releases to spare you the hassle of too many downloads. But this one is very close to what I had in mind for 0.10 and fixes all known bugs, so please redownload for the best experience.

    • Like 2
    • Thanks 3

    The official BenchMate support thread

    in Benchmark software

    Posted · Edited by _mat_

    13 minutes ago, MrGenius said:

    So you can execute your little vendetta? You know what? FUCK YOU! Leave me, and the rest of us out of that shit.

    It's not my vendetta. It could be a pipe dream though, but I like to call it a better future. As you please, your opinion is still welcome.

    13 minutes ago, MrGenius said:

    To that I'll add...the more I run BM the more I hate it. Bugs bugs bugs. One gets squashed...another pops up. I'm starting to strongly doubt it will ever work fully as intended. Making your pipe dream of having it be the standard for all HWBOT submissions just that. Wishful fucking thinking. Might want to get off your high horse...before it bucks you off. Just sayin'...

    Oh yeah...what I originally came here for. The bugs. It keeps spitting errors about a skewed timer with RTC enabled on W7 and any change in BCLK. So we can't OC with BCLK unless we use HPET? Is that the story...or what?

    Are we talking latest version, BenchMate 0.10.2? It is hard to say anything without a bug report. Please submit one so I can work on it. Otherwise this is not really of much use. I can't reproduce anything you say with the latest version on any of my available test platforms.

    The number of submission out there in the field also tells quite a different story. And please ask anybody, I take every bug report as serious as possible and try to fix bugs as quickly as my time allows (this is still a hobby).

    Btw, it is NOT MANDATORY to use BenchMate on HWBOT. If you don't like it, don't use it. Use Windows 7, you'll be fine.

    • Like 2

    The official BenchMate support thread

    in Benchmark software

    Posted

    HWBOT can and will decide for itself what's best.

    My hands are tied with GB and not for the lack of trying. It was all already in there, working and as stable as it can be directly securing the inner benchmark executable and parsing its result. The problem is that whatever this goes next, if GB stays active it will always fall back to BenchMate being not able to handle it and therefore less useful as a solution to unify benchmarking.

    I doubt that GB decides whether the bot can be successful or not. There are many very worthwhile benchmarks out there waiting to be tortured with developers happy to see them in good use and support them long-term. I know because GPUPI has been one of my happy places since it got added here. :)

    GB is officially not one of these benchmarks (decided by the dev itself) and because of that it will end up the same having its points removed sooner or later. It will just hurt much more if it is later and without an announced transition period.

    The official BenchMate support thread

    in Benchmark software

    Posted

    2 hours ago, ADVenturePO said:

    Hi OCers,
    @_mat_

    2xQ:
    Do I have to submit score manualy still for like Cinebech R15? Or it will be rewritten from Cinebench R15 with Benchmate Score?
    Are we able to add manually GB3 and GB4 into Benchmate by add button below or is it totally out?

    Thank You

    Yes. HWBOT has not managed yet to integrate BenchMate correctly, because all legacy benches are apparently hand-written and do not support the submisssion API.

    You can use BenchMate 0.9.3 and manually add GB. BenchMate 0.10+ does not support GB at all, as I won't get my ass sued over a benchmark that is developed by somebody who doesn't care about result validity or security to begin with.

    If you do use an old BenchMate version, I recommend you run it on a different OS install. Mixing two BenchMate versions with bluescreens and instable processes will at some point result in some broken pieces of the old version lying around, that the new version can't remove. You will have to manually remove the BenchMate services then to be able to start the other version.

    I'm looking forward to the day where Leeghoofd comes through with his promise to remove GB's points. After all this shit this has put me through, I have to answer questions like these on a daily basis. Please, enough is enough.

    • Like 2

    Is overclocking DEAD or dying?

    in HWBOT News

    Posted

    2 hours ago, richba5tard said:

    I know visitor stats do not say everything, but they are certainly not depressing for HWBOT. We went from 1.0 to 1.2 million unique visitors in 2019 compared to 2018. +20% is healthy.

    Is that 1.2 million uniques per year  measured with Google Analytics as users? Or webstats from the server?

    • Like 1

    The official BenchMate support thread

    in Benchmark software

    Posted

    BenchMate 0.10 is finally here!

    867958229_BenchMate0_10.1.thumb.png.2dd0f2e193c46e587a9b0d33f50dcb4b.png

    New benchmarks, Splave's ColdbugMate, benchmark integration completely rewritten, full Windows 7 compatibility, an auto-updater, fancier result screenshots, 3DMark03/05/06 support with LOD, GODLOD, Triple Buffering, ...
    You will love it, I promise.

    There is over 3 months of full-time work in this release. I'm talking 12 hour days, next to no weekends, girlfriend pissed - the whole deal. The changelog has 53 (!) lines solely dedicated to new features, improvements and bugfixes. If you like my work, please consider to donate via PayPal or become a Patreon. Thank you! ?

    Don't bench unprotected.

    Download: https://benchmate.org/?v0.10.1
    Changelog: https://benchmate.org/changelog.html?v0.10.1
    Donate: https://bit.ly/2LQgc5S
    Patreon: https://www.patreon.com/benchmate

    • Like 5
    • Thanks 2

    The official BenchMate support thread

    in Benchmark software

    Posted · Edited by _mat_

    This the right folder, but the prerelease version of BM 0.10 does not create .hwbot files yet for wPrime and 7-Zip.

    Tomorrow the new version will officially launch and I have already added the "wPrime * with BenchMate" categories and the 7-Zip category as well as 3DMark03, 05 and 06 as beta benchmarks to HWBOT. This version will produce . hwbot files, which you can upload (or submit inside the client).

    • Like 3

    rsannino - Ryzen Threadripper 3970X @ 5649MHz - 31sec 47ms GPUPI for CPU - 1B

    in Result Discussions

    Posted · Edited by _mat_

    Crazy good!

    This is a great example why BenchMate should be used instead of just GPUPI + enabling HPET. The latter will perform worse because HPET is slow on these HCC chips and enabling it system-wide has a serious impact on overall benchmark performance. BenchMate just injects HPET into GPUPI, so only a few calls for time measurement inside the benchmark are using HPET while the system timer interrupt is still based on the faster LAPIC PM timer.

    This is how it's done!

    Edit: Love the Wahaha!.exe on the desktop. ?

    Is overclocking DEAD or dying?

    in HWBOT News

    Posted · Edited by _mat_

    13 hours ago, ShrimpBrime said:

    So on that note, being a wrapper as it is, when a benchmark fails, how will the wrapper handle it? Will it automatically close itself if errors are found such as a driver crash ect?  

    BenchMate monitors the benchmark process and it has very strict rules on what data needs to be available and valid to show a score as successfully verified.

    If the benchmark process is unstable, it will most definitely crash. BenchMate recognizes that and will not let you save any result (from a process that is not even there anymore).

    If the benchmark doesn't crash but shows weird behaviour, there is a good chance this will show up somewhere in BenchMate's result dialog. Like for example a benchmark run that is cut short, but the benchmark itself doesn't realize that and still shows a result (very rare, but has happened in CB R15 for example). Due to the fact that we now have so much more data for this run than any screenshot could possibly show, the moderators can make a better decision if this was valid or not. A run cut short for example will show a very different runtime and less timer calls for the whole run. This is data we never had before and we can use it for good.

    • Like 1
    • Thanks 1
×
×
  • Create New...