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

    On 6/5/2020 at 4:28 PM, _mat_ said:

    This is well known. Old NVIDIA drivers want to inject code into the benchmark process, which is obviously a big no-no!

    Fixes:

    • Use the latest NVIDIA driver if possible
    • Use a more recent NVIDIA GPU or an AMD GPU

    The upcoming version has a fix for this and will not block the result, but still block the access to the process memory.

     

    The official BenchMate support thread

    in Benchmark software

    Posted

    1 hour ago, Alex@ro said:

    No, i don't mix bench-mate versions anymore. The thing is that i started clean, exactly fresh version of window and bench mate 10.5 . It worked, then it didn't. Then i put new os with same bench mate , didn't ran. Download new version, didn't work. New os plus new version works. After few crashes here we go again. 

    There is without a doubt a solution for that. Please send me the log files (whole BM/logs directory as zip) so I can look into it. You can also use the BugReport.exe tool in BM's directory if the system is online.

    You are ditching a new take on benchmarking because of a single bug of an early beta version. If any software developer would stop working on something because of that, the world would not have a single software product. You really have to distinguish between flaws in the concept and the implementation.

    1 hour ago, Alex@ro said:

    And to be honest you are overdoing it. I mean how many people can temper with benchmarks the way you think they do? As far as i know so far we had few scammers but most of the issues were photoshop related. And those who altered anything else were caught by the sharp eye of the optician- Splave :D  So i am sorry, i don't have the time to search and think for cheating solution, i want to actually bench and enjoy the process not getting stuck by whatever bug of a over-zealous programmer.  

    Let me be clear. I respect your work in the way that it's free and you are trying to improve things in order to be cheat-secure. That's awesome! However when you impose things, like bench mate mandatory for certain platforms and your app creates issues, you know what to expect. Ryzen3 sub-zero issues, b-die issues and on top of that bench-mate having issues, oh boy, hell of a benching experience....NOT. 

    Ah, here wo go again. Screenshots are so great, manual input of results is really efficient, nobody is cheating, the future of overclocking is bright! I've heard it all before and watched it quietly while supporting GPUPI for many years here. HWBOT is already bending backwards to manage the situation with the tools at hand. Platform/OS matrix with has to manually checked by the mods against timer skewing? Seasonal ranking needs a text editor open or a certain version of CPU-Z? Beginner's scores are often removed due to the complexity? Benchmarks and results have to be removed when a flaw is found (if it can't be verified that it's valid, it has to be assumed it's not)?

    When will we realize that we at least need an alternative that's being worked on in parallel to tackle these problems FOR THE FUTURE. I'm not saying that BenchMate needs to be mandatory right now (or maybe ever) here on the bot. But I am saying that it's unwise to not explore (over-zealous) solutions, that would allow benchmarking and overclocking to reinvent itself.

    • Like 1

    The official BenchMate support thread

    in Benchmark software

    Posted

    2 minutes ago, Splave said:

    Really sucks and still a move that I dont understand by the dev but it is what it is. 

    I do understand it. Soon every benchmark will have a result database that is much better than the Geekbench database. Better integrity/quality, better hardware detection, nice graphs, screenshots. And it costs ... nothing. ❤️

     

    The official BenchMate support thread

    in Benchmark software

    Posted · Edited by _mat_

    1 hour ago, Alex@ro said:

    First of all i want to say that i am giving away 100$ cash for anyone that can remove this from internet or at least take the website down for at least a week.

    Be my guest. A bounty to take down BenchMate's security is something I always wanted to do.

    1 hour ago, Alex@ro said:

    Yeah, everything is nice and dandy, unicorns are pink and gooses are flying until this software will simply not work.

    It's an early beta version. You can contribute by posting your problems so I can fix them. This is pretty normal, especially with something that difficult like BenchMate.

    1 hour ago, Alex@ro said:

    And for the worst, you cannot debug anything at it, it works or doesn't you don't know why or how to replicate the issues.

    Everything that is not security-related is logged to file. Check the debug log path inside the BenchMate directory.

    1 hour ago, Alex@ro said:

    Example, let alone the fact that if you are trying to run two versions on the same system for example 0.10.5 or 0.9.3 (for geekbench) you will have only issues. Which is weird because they are standalone, you won't install them competing for same registry or cache or whatever. 

    It is well known that this doesn't work. You can thank the Geekbench dev for that, because 0.9 wouldn't even be available for download if it wasn't for his legal threat. But I didn't want to let benchers down, so I kept it online if you really need it. It was never meant to be used in parallel with other versions, so that is by design.

    That said, the next version already supports to run different versions of BenchMate at once.

    1 hour ago, Alex@ro said:

    Benchind amd ryzen 3 on windows 10, clean os, working benchmate. Few restarts and nothing works. Benchmate will start but no benchmarks will load, will be stuck at loading. Of course the benchmarks run if you start them on their own.  

    So, i reinstall os and try again. Nope. Meanwhile i tried cleaning cache, temp files, prefetch, whatever , no go. 

    I download a clean benchmate and try it, not working.

    Final solution, new operating system with clean benchmate, finally it works.

    This sounds like you are mixing BenchMate 0.9 and 0.10. Don't do that! Use another clean OS to keep them separate for now.

    1 hour ago, Alex@ro said:

    Also i had many crashes from memory when geekebcnh was nice and stable. How come, i mean benchmate is harsher than geekbench or what?

    No. There is next to no impact during a run other than measuring sensors (like GPUPI does). This was optimized for the next release as well to have even less impact for the work that is done.

    1 hour ago, Alex@ro said:

    Why do i have to do this? Why do i have to waste time and ln2 and etc for running a software on top of my benchmarks?

    What happened with pretesting on air? Things should work nicely before you start with cold. This is the same process as it is with any benchmark, tweak, driver or OS install in general.

    1 hour ago, Alex@ro said:

    In my experience over time i found that eventually you can find answers to most common issues with benchmarks. Google it, this forum, ask a fellow overclocker, eventually you know what to do. But with benchmate, you are left blind and dumb, both at the same time.

    I'm always happy to help as quickly as possible. Sometimes BenchMate is not even the issue, but that's fine with me. I help anyways.

    1 hour ago, Alex@ro said:

    My ranting stops here, i would rather pay for supporting hwbot, hire a full-time guy to check for cheaters and everything rather than use this junk. It gives issues most of the times, when nothing works you don't know what to do to fix it. While the idea is great, i hate cheaters  and i would send them to guilotine,  the execution leaves a lot to be desired. Sorry, i would rather hope that the developer can fix the issues by being annoyed at my post rather than praise the software when it doesn't work. 

    Please do that. If it helps competitive benching, then it's a good thing.

    HWBOT (as it is now) and BenchMate are two very different approaches to competitive benchmarking. HWBOT uses screenshots, CPU-Z started after the benchmark, uploading with manual form entry. That was always problematic to begin with, because the data to moderate scores is simply missing. So even if you have ten paid fulltime guys, you will fail. The concept is flawed, but without any universal alternative it had to make due.

    BenchMate however tries to create a new workflow similiar to that of Geekbench and 3DMark, but for every benchmark out there. This even works for benchmarks that are not supported anymore. Time measurement, process memory protection, result saving and uploading - these features are hard to do and very time consuming. With BenchMate's eco system the benchmark developers can simply use the actively supported infrastructure and focus on the workload. Is that hard to do? Fuck yes. It's a nearly impossible, non-trivial task. Security alone is hard enough, but BenchMate needs to be fast as well to have no impact on benchmark performance. So the only thing I need to ask for is patience so I can make it happen.

    I like feedback. Good, bad, ugly, it's all fine to me as long as it helps to improve BenchMate in the long run. To me it just seems that you have mixed 0.9 and 0.10 to make Geekbench work, which is no longer officially supported. Neither is version 0.9.

    HWBOT GPUPI for CPU - 100M Benchmark Validation Regulations

    in Old HWBOT Rules (pre 2021)

    Posted

    6 hours ago, viper said:

    Why can't we just upload the full screenshot as we do for so many other benchmarks?

    To quote myself from the mission statement on benchmate.org:

    Quote

    Benchmarking is in a horrible spot right now. Even on HWBOT, the world's biggest result database, most results have to be submitted as screenshots and system data is entered manually. Due to undetectable time measurement bugs and easily abusable vulnerabilities of benchmarks, results can not be correctly judged by the inconclusive data available. Additionally, manual submission is error-prone, inconsistent, time consuming and hard to learn while still being far from trustworthy. Although there are better submission methods and online rankings available like 3DMark's Hall of Fame or the Geekbench Browser, they all set their own opaque standards for validation and are only available for their (latest) products.

     

    HWBOT GPUPI for CPU - 100M Benchmark Validation Regulations

    in Old HWBOT Rules (pre 2021)

    Posted · Edited by _mat_

    15 hours ago, viper said:

    Hey is there going to be any update soon?

    There is going to be a major BenchMate release first. All my time and effort is going into that.

    GPUPI 4 will still arrive this year though.

    15 hours ago, viper said:

    It's not fair not to be able to save any newer scores. It's a waste of time and effort for better results.

    I have no clue what that means. Is this about 3.2 and 3.3?

    The official BenchMate support thread

    in Benchmark software

    Posted

    This is well known. Old NVIDIA drivers want to inject code into the benchmark process, which is obviously a big no-no!

    Fixes:

    • Use the latest NVIDIA driver if possible
    • Use a more recent NVIDIA GPU or an AMD GPU

    The upcoming version has a fix for this and will not block the result, but still block the access to the process memory.

    M/L-azul - Ryzen 7 2700 @ 4142MHz - 20.36 points Cinebench - R11.5

    in Result Discussions

    Posted

    @ML_azul This indicates that something is accessing the benchmark's process memory, which is forbidden of course. The same thing could for example change/overwrite the score in the window.

    I've already seen this quite a few times in the past months and the problem is an old version of NVIDIA's GeForce driver for old graphics cards (like 8000 series), that tries to hook into the process for whatever reason. You have a few options:

    1. Update the driver to the latest version. I'm not sure if that works for older GPUs.
    2. Use another GPU. AMD is fine for example.

    This bug has been fixed in the upcoming version, I just disallow the access from the driver. It doesn't harm anything but won't trigger the error.

×
×
  • Create New...