jfpoole
-
Posts
25 -
Joined
-
Last visited
-
Days Won
6
Content Type
Profiles
Forums
Events
Blogs
Posts posted by jfpoole
-
-
5 hours ago, Leeghoofd said:
As we see more and more CPUs with bigger caches it should have an autodetect function so the benchmark can't be run only in the cache and provide weird outputs as what we see happening with low maxmem settings... old Geeks don't look good on the upcoming gens
I worry about runtime checks that make decisions about how to run the benchmark.
Geekbench 3 and Geekbench 4 include runtime checks that modify the benchmark to accommodate older, less-capable hardware. These checks are what folks are using to "tweak" their Geekbench scores.
My preference would be to have a "one size fits all" benchmark that works the same regardless of the underlying hardware. It may limit the range of hardware it supports, but I think it's the best, most robust approach to having a sane benchmark that is resistant to hacks.
-
3
-
-
Over the past few months we've been toying around with the idea of creating a separate memory benchmark. Our initial thought would be to use the Geekbench 3 and 4 memory tests as a starting point. The memory test work and are well-liked, and moving them into a separate benchmark would give us the opportunity to fix the outstanding issues with them (especially on modern hardware).
If we were to create something that focused on testing memory performance, what would people want to see? What is your wishlist for a memory benchmark?
-
6
-
-
2 hours ago, Alex@ro said:
So, overall, what is your big problem seeing geekbench in benchmate? It's still your software, no one interferes with it, you will get the same money, your wife is not going to leave you for this...what's the problem? Be a grown-up and spill it out loud and clear.
I outlined my initial concerns with Benchmate earlier in this thread, and since then we've discovered that Benchmate can cause Geekbench to report unrealistically high scores for some workloads.
-
On 12/29/2020 at 8:06 AM, _mat_ said:
Only BenchMate 0.9.3 supports Geekbench 3, 4 and 5. BenchMate 0.10.5 is the latest stable release, but too old for Ampere.
Ampere and Vermeer are properly supported with the latest version, BenchMate 0.10.6, currently in beta. It also reintroduces Geekbench 3 with an additional legal disclaimer. It will be released very soon, but if you want to grab it today, it's available for all supporters: https://www.patreon.com/benchmate
We're not interested in having Geekbench integrated in Benchmate. Seeing that you're re-introducing support is confusing. I would urge you to reconsider this decision.
-
1
-
-
We've integrated the fixes introduced in Geekbench 3.4.3 into Geekbench 4 and Geekbench 5. Geekbench 4.4.4 and Geekbench 5.2.4 fix the issues that prevented both from being used on AMD systems running Windows 10 on HWBOT.
Both releases are available for download from the Geekbench website:
http://cdn.geekbench.com/Geekbench-4.4.4-WindowsSetup.exe
http://cdn.geekbench.com/Geekbench-5.2.4-WindowsSetup.exe
Let me know if you hit any issues with either release.
-
5
-
-
@Sparky's__Adventure Geekbench 3.4.4 is the latest version of Geekbench 3, and includes all of the fixes included in 3.4.3b3:
-
1
-
-
This is fascinating. Thanks, @Splave, for taking the time to write this up.
A couple of things that might be helpful:
- You can still download Geekbench 3.1.5 from our servers if you'd prefer an official link: https://geekbench.s3.amazonaws.com/Geekbench-3.1.5-WindowsSetup.exe
- Geekbench 3.4.3 is out of beta and is now the default download on the Geekbench 3 website: https://www.geekbench.com/geekbench3/download/
-
3
-
2
-
2 hours ago, _mat_ said:
And your application can still be hooked on every API call, starting with DeviceIoControl to feed you whatever as values you need to rely on. It's not a single feature to do things right, it's a strategy that begins with the first line of code and never ends.
This is true of any application (even Benchmate) provided someone is willing to devote enough time to the endeavour. I'm just glad we were able to fix the big issues that were of concern to the HWBot folks.
2 hours ago, _mat_ said:I helped you for the sake of the countless hours of work from this community that went into Geekbench records. I think my duty here is done.
Thanks for your help!
-
@_mat_ I'm glad you were able to verify that Geekbench is now resilient against both BCLK skews and Win32 API hooks.
It's less important how Geekbench implements these protections and more important whether these protections work. If you don't have the resources to test the edge cases we can certainly look into including them in our regular test plan for Geekbench. I would just need to know how to reproduce these edge cases so our QA team can reproduce them in our lab.
-
1
-
-
@_mat_ I’m assuming this means you weren’t able to reproduce the issues you encountered on Beta 1 with Beta 3?
-
Geekbench 3.4.3 Beta 3 is now available for download:
https://geekbench.s3.amazonaws.com/Geekbench-3.4.3b3-WindowsSetup.exe
Beta 3 includes the following changes:
- Update how valid and invalid results are displayed to the user.
- Update the Browser to display the validation status of Geekbench 3 results.
- Fix benchmark initialization issues on HEDT systems. Beta 3 should run much more quickly on HEDT systems than Beta 2.
- Fix memory frequency detection.
Give it a try and let me know if you hit any issues. If no other issues are encountered we'll most likely make this an official release in early March, and then start work on migrating the validation code into Geekbench 4 and Geekbench 5.
-
1
-
1
-
@Splave, it may take a while for the benchmark to initialize on TR3 systems (it's proportional to the number of hardware threads). How long did you let Geekbench sit at the "Benchmark Progress" dialog?
-
Geekbench 3.4.3 Beta 2 is now available for download:
https://geekbench.s3.amazonaws.com/Geekbench-3.4.3b2-WindowsSetup.exe
Beta 2 improves the in-app validation routines, and makes it far more obvious whether Geekbench considers a result valid (or invalid or questionable) due to timer issues. Beta 2 will still let users view invalid results as well as upload them to the Geekbench Browser. The Browser still needs to be updated to display the validation status of the results, but that should be straightforward once the client code is working properly.
Give it a try and let me know if you hit any issues.
-
4
-
-
1 hour ago, Splave said:
- We have extra protection via benchmate at no cost to primatelabs and they dont have to mess around paying people man hours with EOL geekbench3 anymore.
For what it's worth we still sell and support Geekbench 3. For the vast majority of users we recommend Geekbench 5 as it has much more modern benchmark tests, but if people hit bugs with Geekbench 3 we'd like to hear about it.
-
4
-
2
-
How did you run Geekbench? On our test systems that drastic a change to BCLK would’ve been flagged.
-
1
-
-
6 hours ago, _mat_ said:
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.
I disagree -- I think this is something we should fix in Geekbench itself, as then these fixes benefit all Geekbench users.
23 minutes ago, keeph8n said:I never saw an issue with GB3 running on 3970X or 10900X/10980XE? Is there a particular case that wouldn't allow GB3 to be run? 3990X perhaps?
I don't know the full list of processors, but the 3990X (along with the W-3175X) was mentioned explicitly by @Splave.
-
2
-
-
We're working on a new Geekbench 3 release that addresses issues that affect the overclocking community. The first preview release, Geekbench 3.4.3 Beta 1, is now available for download:
https://geekbench.s3.amazonaws.com/Geekbench-3.4.3b1-WindowsSetup.exe
Geekbench 3.4.3 Beta 1 fixes the issues that prevented Geekbench 3 from working on recent AMD and Intel HEDT processors.
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.
Give it a try and let me know what you think.
-
4
-
8
-
-
Geekbench 3, 4, and 5 only use CPUIDSDK to report system information. Swapping out the DLLs should have no impact on the benchmark score.
I'll take a look and see if we can't do a release of Geekbench 3 that includes the latest CPUIDSDK DLLs.
-
2
-
4
-
-
On 2/6/2020 at 10:19 AM, Splave said:
well I'll bite, the elephant in the room is also changing benchmark files is a no no. DLL files from what I know can have quite an effect. In order for Geekbench3 to work on 64 core AMD and for intel 28 core ES 3175x's is to swap the DLL files from geekbench 4 latest version to the geekbench 3 folder and overwrite them. ??
Which DLLs did you have to replace? The CPUIDSDK DLLs?
-
1
-
-
@Mythical tech I am being paranoid here but at the same time we've had legal threats for weirder things in the past so maybe the paranoia is justified!
@_mat_ FWIW removing the profiling code (which wasn't controlled via a setting) didn't change Rigid Body Physics performance on any of our ~150 test systems. I don't expect this issue will affect the majority of our users. It's trivial for us to filter out results that are potentially affected by the LAPIC timer bug -- we can exclude results gathered from Windows 10, or from systems that have a non-standard BCLK, or any one of the number of system metrics we collect. What we can't do is filter out Benchmate results because there's no way for us to identify those results!
Again, I'll reiterate and say I understand that the community here is upset but I'm not comfortable with the Benchmate integration. If that means Geekbench is removed from HWBot that's disappointing but it's most certainly not my call to make.
-
Benchmate integration introduces a number of issues for us, both technical and legal. On the technical side, Benchmate has introduced issues that have prevented Geekbench from working properly, or have prevented Geekbench from accurately measuring performance. On the legal size, what if Benchmate's modifications to Geekbench cause Geekbench to delete data from a user's system and we get sued because of it?
We may be shooting ourselves in the foot here, but I fear this is very much a damned if we do, damned if we don't decision in which case I'd rather make the decision that gives us the most control over our destiny.
-
1
-
-
Sorry, I'm still not comfortable with Benchmate integrating Geekbench. If that means HWBot no longer uses Geekbench then so be it.
-
1
-
1
-
-
Thanks to @GeorgeStorm for getting me back into my old account!
Thanks, @mickulty, I appreciate the kind words. We've spent a lot of energy dealing with mobile vendors who manipulate Geekbench scores (see, e.g., https://www.geekbench.com/blog/2018/09/huawei-benchmark-boost/) but this has been a blind spot for us. I feel really dumb that we just assumed everything was ok. The team's been discussing strategies on how we can address this issue and I hope we'll have something to share in the coming weeks.
-
2
-
3
-
-
What sort of cheats are you wondering about? Geekbench is pretty good at preventing users from cheating on the benchmarks and preventing users from submitting fake results to the Geekbench Result Browser. However, users might be able to trick Geekbench's system information library Geekinfo into reporting the wrong hardware configuration (i.e., I've not spent any time "hardening" the library against cheating).
Running Geekbench in Benchmate (Possible?)
in HWBOT Software and Apps
Posted
That error message means Geekbench cannot connect to the Geekbench Browser to upload your benchmark result. Can you access the Geekbench Browser using your web browser?