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_

    Geekbench 3.4.3

    in Benchmark software

    Posted

    I appreciate the update as well. A benchmark developer showing interest to work together with HWBOT and the overclocking community should always be appreciated. In my opinion it's an absolute necessity to cooperate to be able to keep the benchmark in the ranks long-term.

    On 2/12/2020 at 6:29 AM, jfpoole said:

    Geekbench 3.4.3 Beta 1 also fixes several issues that prevented Geekbench 3 from being used on AMD systems running Windows 10 on HWBOT.

    May I ask what the specific changes are?

    21 hours ago, Leeghoofd said:

    If they can fix some stuff on Matt's recommendations, maybe we don't need to go to this step

    It is impossible to do this in userland applications. Timers can not be verified correctly by using WIN32 API functions alone, direct access to hardware devices is needed (a kernel driver). The only workaround is to enforce HPET system-wide like I introduced with GPUPI years ago when I had no other solution. I doubt that this is the right way for Geekbench as a mainstream benchmark.

    And even if a userland solution is found, 30 seconds with Cheat Engine are enough to change the timers to your liking. I've been there and if you think this through this is not the way to go with competitive benchmarks.

    My recommendation is to integrate Geekbench into BenchMate again, which solves all problems at once. I'm up for it, because in the end it's the best for overclocking and the community to not remove points from Geekbench if not necessary.

    • Like 4

    The official BenchMate support thread

    in Benchmark software

    Posted · Edited by _mat_

    Thanks, very much appreciated. :)

    BenchMate itself is not doing anything nasty, I was very careful with direct hardware access. My guess is that HWiNFO is intervening with the SMBus communication of EleetX. This should only be a problem during the benchmark run though, because otherwise HWiNFO is not doing anything after successful startup.

    HWBOT 2020 Make it great again Thread:

    in HWBOT 2021 Edition

    Posted

    4 hours ago, yosarianilives said:

    Definitely would love to see this, I think with datafile benches it may be technically possible, others idk.

    BenchMate already has it if you submit online from within the client. If you bench offline, just copy the result folder to another BM install on your normal PC (json and hwbot files at least). Select all results in the result list and submit.

    Or use BenchMate on a USB stick and just plug it into your PC after the session.

    The official BenchMate support thread

    in Benchmark software

    Posted

    It is not just a single executable, it is so much more. It is a launcher application, three services, a kernel driver, the client application and 5 DLLs. Then there is the bug report tool and a hardware detection debug tool.

    I made it portable, because I knew that this will be the best for overclockers. You can even run it on a USB stick, it works just as fine. But if for example the system hangs and leaves broken services behind, these will interfere when you start a different BenchMate versions afterwards.

    In any case, this has been fixed and will be soon available for download. I'm really close with finishing BenchMate 0.11, which will make our lifes better for sure. Don't forget that this is an alpha software from a single guy that is so low on funding, that it's not funny anymore. This is early stages for something that nobody has every tried before. It is supposed to be no fun.

    TR 3990X

    in HWBOT News

    Posted · Edited by _mat_

     

    3 hours ago, l0ud_sil3nc3 said:

    I thought it was you :)

    I was the first to break 3G on AMD and started the whole Bulldozer memory frequency crap by taking the DDR3 crown with 3311 MHz. :D

    That was a wild time for memory overclocking for sure, lots of hours and money spent to be on the top. And all that's left are 1.5 hardware points. I personally couldn't give more of a fuck, I'd rather see these achievements displayed/promoted in a better way. Being listed in the record list a lot more  satisfying than hardware points: https://hwbot.org/benchmark/memory_frequency/

    So that's why I think that a better  timeline for records, separated for example by platform could be a way out to have more flexibility for decisions regarding benchmarks and its points. Additionally a list of lifetime achievements on the user page would be nice. Something like:

    * First to achieve 100.000 points in 3DMark01

    * Held the Super Pi global record for 35 days in Nov 2015

    Benchmarks and points come and go, achievements are here to stay.

    • Like 3

    TR 3990X

    in HWBOT News

    Posted

    It amazes me that everybody thinks XOC is dying while demanding that nothing should be done about it at the same time. That's not how it works. HWBOT needs to adapt to the new challenges today, not dwell on the past.

    We have a saying in Austria: "Was sich nicht bewegt, ist tot", roughly translated into "If it's not moving, it's dead".

    Moving on does not devalue the achievements of the past. Those were celebrated until the next record comes along. That's XOC. The only thing that sucks is that these achievements are not properly frozen in time. So let's ask the question how we can honor benchmark results from the past. Two things come to mind:

    1. Implement a nice(r) timeline to revisit the records of each generation.
    2. Freeze the ranking or at least seperate it from new results after the benchmark was frozen. Seeing records made with zombie-modded beauties on liquid nitrogen getting replaced by a workstation card two years later isn't fun to watch. For what? A few hardware points?

    Maybe we should have a look into properly preserving our achievements other than trying to hold on to them as long as possible.

    Fun question to prove my point: Who was to the first guy to break the DDR3 3000 barrier?

    • Like 1

    TR 3990X

    in HWBOT News

    Posted · Edited by _mat_

    5 minutes ago, der8urner said:

    If we include a new tweaked gb3 in bm (as @Splave suggests it here) 

    I am not allowed to do anything with Geekbench. The dev threatens with a lawsuit because I'm obviously profiting from derivative work by fixing his product with a free tool. ?

    • Like 1

    TR 3990X

    in HWBOT News

    Posted · Edited by _mat_

    That only shows that we desperately need a real memory bench to legitimize good memory kits. :)

    My concern about Geekbench and other benchmarks that deliberately don't give a rat's ass about competitive benchmark is, that sooner or later it all comes back to HWBOT for not being "fun". We already have no control over the development of Windows, CPUs, hardware timers and much more, and look where that got us:

    windows8-81-10.thumb.png.eba01e25e9b7f91b14e60d728f34417d.png

    Is this what fun looks to you? It is definitely not for a newcomer or somebody that just wants to show off his system. That's why Geekbench is still so popular. You need no effort to make it work, it just does. But popularity doesn't say anything about the quality of the score. That fat "N" in the list above and my research tells a different story. They are going for quantity and their results are not appropriate for competitive benchmarking or in fact worthy of any kind of ranking/list.

    So the last thing we need is that benchmark developers walk all over us as well. It is an absolute necessity that we cooperate with these devs to work around the issues we certainly have no control over. If the developer doesn't want that, we have a single weapon to decide our own future: global points. So please let us them wisely.

     

    The official BenchMate support thread

    in Benchmark software

    Posted

    BenchMate is implemented to use next to no ressources, especially during the benchmark BenchMate does nothing to interfere. All action is triggered by hooks inside the benchmark process, which dictates what BenchMate is showing. The only exception is the sensor measurement, that happens once every second. We need that to define the new benchmark standard that gives more understanding what happens during the benchmark. This will make much more sense with BenchMate 0.12, were all the data will come to life. It's going to be a beatiful graph orgy for sure.

    2 hours ago, Alex@ro said:

    but benchmate would never start, even if i reboot many times.

    What happened exactly? BenchMate applies next to no load on the CPU when it is starting up. The only part that does need some ressources is the verification of the installation files to make sure they weren't tampered with. I doubt that this comes anywhere close to the workloads that Geekbench applies.

    TR 3990X

    in HWBOT News

    Posted · Edited by _mat_

    The red X is due to the many threads using a lot of timer calls during Geekbench, which had the possibility to colide inside BM's kernel driver and produce a skew error, which will ultimately get blocked. This has been already fixed of course, but not in BenchMate 0.9.3.

    This is exactly what happens when I'm not allowed to support Geekbench with the latest BenchMate versions. People go back using old versions with bugs that have already been fixed, mixing up BenchMate versions that leave artifacts due to crashing and inevitably fuck the installation up (unless you know what you are doing and remove the BenchMate services manually). Worst case scenarios that all fall back to BenchMate, exactly as I predicted.

    Geekbench needs to either allow the integration or lose its points. This is not fun for anyone of us.

    • Like 5

    The official BenchMate support thread

    in Benchmark software

    Posted · Edited by _mat_

    Did you use multiple BenchMate versions on the same OS? Switching between 0.9.3 and 0.10.x for example does this.

    The upcoming release allows new versions to be run right next to old ones without any conflicts. Services, driver and all are available seperately for reach version, each component has their own version postfix. So please be patient, the next version will improve on all fronts. :)

    • Like 1

    GPUPI problem

    in Rookies Help Rookies

    Posted · Edited by _mat_

    You need to use BenchMate to save your result file, don't do it within GPUPI. The hardware detection and submission upload in GPUPI hasn't aged well and is unstable.

    For the next BM version I already added a warning and disabled this functionality, so people can't use GPUPI result file saving.

    The official BenchMate support thread

    in Benchmark software

    Posted · Edited by _mat_

    You are probably using a CPU that is currently not supported. The AES-NI instruction set is necessary for now to do fast file integrity checks.

    The next version, BenchMate 0.11, won't need the instruction set and runs even on socket 775. That's the oldest platform I have here for testing. It will be available for download in the next couple of days (asap).

    • Like 1
×
×
  • Create New...