I7-7700k, at first I was not surprised, and then I decided that my Schwarzenegger was not so useless yet and found out that in fact this instruction is still supported.
Try uninstalling SystemInfo manually, then deleting the folder if one still exists under c:\Program Files (x86)\Futuremark\SystemInfo and then rebooting & reinstalling it. For Steam version, that is via revalidating 3DMark files.
There is an unfortunate further issue with the latest version ending up with jank from the previous version if updating SystemInfo on top of the previous install.
If uninstalling, then installing, it works.
Yes, we are facepalming over this over here. Sometimes Windows is just... groan.
Try uninstalling SystemInfo manually, then deleting the folder if one still exists under c:\Program Files (x86)\Futuremark\SystemInfo and then rebooting & reinstalling it. For Steam version, that is via revalidating 3DMark files.
There is an unfortunate further issue with the latest version ending up with jank from the previous version if updating SystemInfo on top of the previous install.
If uninstalling, then installing, it works.
Yes, we are facepalming over this over here. Sometimes Windows is just... groan.
it say my cpu does not support SSE3 LOL .... never saw this before and my cpu is i9 9900K !
Totally crap
Same here noticed this today trying to run timespy. Got a reply from them just now...
Try uninstalling SystemInfo manually, then deleting the folder if one still exists under c:\Program Files (x86)\Futuremark\SystemInfo and then rebooting & reinstalling it. For Steam version, that is via revalidating 3DMark files.
SystemInfo 5.60 is now out and should fix this no matter what the previous state was (if you had 5.58 or 5.59 installed, or even something older, in all cases 5.60 will update correctly)
Now on top of this, we are still trying to figure out very rare reports of bluescreens with the new versions (newer than 5.57) and if you are seeing anything like that, please contact us at ul.benchmarksupport@ul.com
And a workaround for those with this rare bluescreen issue is to manually uninstall and revert to the 5.57
Different problem here: 3Dmark apparently works alright, but I have a massive drop in the TimeSpy scores (both regular and extreme) in a Ryzen 5900X + 7900 XTX. Example:
Both tests with same general config, CPU is auto-overclocked by Ryzen Master but GPU is stock.
Important detail that I made this test after upgrading the Adrenalin driver to 23.3.1 released yesterday, so this could be a driver problem but I ran many other benchmarks, from 3DMark and others, also played some games, nothing else drops performance, only the TimeSpy which is a cataclysmic >50% loss.
Let me know if you need more help, I can make better apples-to-apples testing - the higher-scoring result above is from a few weeks ago and there are always Windows updates etc. not to mention that I recently installed an AIO, that';s why the CPU temperature is 11C lower in the new test and CPU clocks 50Hz higher - still even the CPU score is lower. But if this is a TimeSpy versus Adrenalin 23.3.1 bug, might be easy to repro.
Forwarded this to be verified in our test lab tomorrow, but for now first thing I'd recommend is to rule out driver wackiness (especially if you upgraded the driver) by doing a DDU wipe and reinstall.
Thanks, I will do that, testing again with both 23.2.2 and 23.3.1 both reinstalled after DDU and ensuring everything else in the system identical; will post results later tonight.
Hi again, I now ran all my benchmarks (all 3DMark tests and a few other) in a very controlled system, after DDU + reinstall of the Adrenalin driver 23.3.1. Problem confirmed, specifically (compared to 23.2.1):
* CPU Profile Max Threads: -4.82% * Intel XeSS: -36% * TimeSpy: -55%, Extreme: -11% * VRMark Blue Room: -12%
All other 3Dmark tests have neutral to positive results, including:
Comparing only the Graphics score for tests that have a combined CPU+Graphics scoring like TimeSpy. Most of the improved scores are because my older/reference scores for 23.2.1 are from before I upgraded my cooling to an Arctic Liquid Freezer II 360, so most things with some CPU load get a small bump.
As for actual gaming, I played a couple hours of Returnal yesterday, normal performance. I also have a good set of previous bench scores for many games (I'm a small-time/amateur GPU reviewer), no time to re-test those today but I'll do it maybe tomorrow.
All results above with GPU in stock settings. I also tested TimeSpy with my overclocked GPU profile, got almost same score: +1.70% for the base test, +1.36% for Extreme. This is also not normal, my OC-versus-stock improved TimeSpy by +2.42% and Extreme by +6.86% in my last round of testing with the same overclock profile.
Mos abnormal results are for DX12 benchs, including the Intel XeSS feature test, although there's also VRMark which is DX11, and big DX12 benchs like SpeedWay/PortRoyal/DXR with perfectly sane results.
Update on this: We have not been able to reproduce this in our test lab at this time. We're still trying, but nothing so far. So we know at this time that this is not a general issue that affects everyone with 7900XT/XTX. If this were such an issue, it would have also come up during driver approval testing but that process saw nothing odd with the latest drivers when they were tested.
But we also looked at our database to compare results on the previous driver vs. new driver on any result using 7900XT or XTX and can confirm that this appears to be a real, if very rare issue. Among all results with the new driver, approximately 3% of the results show abnormal (very low) scores on Time Spy. No similar group of very low scores appear on the results with the previous driver version.
This would lead us to believe that this is some kind of issue with the latest driver that requires some additional conditions to be true to appear. We've raised the issue directly with AMD and are waiting for their response.
Since benchmark itself has not been modified for ages, our current view is that this is a rare driver bug that somehow causes the system to run in some kind of low power mode while Time Spy is running, which then tanks the performance to a very low level. Benchmark itself is quite simple - it just renders 3D graphics as fast as possible and measures the fps - and the code has been vetted and audited quite extensively, so a bug in the test seems very very very unlikely, so the reality is that for some reason, this specific 3D graphics load on the latest driver can sometimes rarely result in GPU running it very slowly.
At this time I would recommend simply disregarding the result as a driver issue and await for future AMD driver updates to solve it. If you REALLY want to try to get rid of it, next step I would do would be to try a clean OS install, but obviously this is not a good solution for a home system that is in active use, and probably not worth the effort at this time.
Also just to rule it out, has anything related to power profiles been modified from default in Windows? Driver side settings issue seems ruled out if you tried DDU wipe and ran unmodified.
Thanks for the feedback, it does indeed look like a rare driver problem. I use Windows's High Performance power setting but not customization beyond that. This is my before/after comparison for TimeSpy after the DDU & reinstall cleanup:
Easy to notice the much lower GPU clock frequency, looks like the driver is putting the GCD to sleep, also causing GPU temps to be 12C lower.
Today I ran a good number of game benchmarks - Callisto, SOTR, Quake II RTX, Guardians of the Galaxy, Horizon, Dying Light 2, Batman Arkham Knight, Returnal - every one of them performed normally, It seems there is something very specific in the rendering engines used by some of the 3DMark tests that this driver doesn't like and only in some specific circumstances since you can't repro.
Report this post
There is an unfortunate further issue with the latest version ending up with jank from the previous version if updating SystemInfo on top of the previous install.
If uninstalling, then installing, it works.
Yes, we are facepalming over this over here. Sometimes Windows is just... groan.
it works :)
Same here noticed this today trying to run timespy. Got a reply from them just now...
Try uninstalling SystemInfo manually, then deleting the folder if one still exists under c:\Program Files (x86)\Futuremark\SystemInfo and then rebooting & reinstalling it. For Steam version, that is via revalidating 3DMark files.
OOPS I missed the comments above sorry.
Now on top of this, we are still trying to figure out very rare reports of bluescreens with the new versions (newer than 5.57) and if you are seeing anything like that, please contact us at ul.benchmarksupport@ul.com
And a workaround for those with this rare bluescreen issue is to manually uninstall and revert to the 5.57
https://benchmarks.ul.com/downloads/Futuremark_SystemInfo_v557_installer.msi
But do contact us if you have this issue, you could help track down the cause for a proper fix.
https://www.3dmark.com/compare/spy/35755350/spy/36463560
Both tests with same general config, CPU is auto-overclocked by Ryzen Master but GPU is stock.
Important detail that I made this test after upgrading the Adrenalin driver to 23.3.1 released yesterday, so this could be a driver problem but I ran many other benchmarks, from 3DMark and others, also played some games, nothing else drops performance, only the TimeSpy which is a cataclysmic >50% loss.
Let me know if you need more help, I can make better apples-to-apples testing - the higher-scoring result above is from a few weeks ago and there are always Windows updates etc. not to mention that I recently installed an AIO, that';s why the CPU temperature is 11C lower in the new test and CPU clocks 50Hz higher - still even the CPU score is lower. But if this is a TimeSpy versus Adrenalin 23.3.1 bug, might be easy to repro.
* CPU Profile Max Threads: -4.82%
* Intel XeSS: -36%
* TimeSpy: -55%, Extreme: -11%
* VRMark Blue Room: -12%
All other 3Dmark tests have neutral to positive results, including:
* FireStrike: Extreme +3.12%, Ultra +2.82%
* SpeedWay: +4.47%
* Port Royal: +3.46%
Also ran other benchs, all neutral/positive scores:
* FurMark 1440p: +4.58%
* Superposition 4K Optimized DirectX 11: +1.88%
* Blender 3.4.0: +3.67%
Comparing only the Graphics score for tests that have a combined CPU+Graphics scoring like TimeSpy. Most of the improved scores are because my older/reference scores for 23.2.1 are from before I upgraded my cooling to an Arctic Liquid Freezer II 360, so most things with some CPU load get a small bump.
As for actual gaming, I played a couple hours of Returnal yesterday, normal performance. I also have a good set of previous bench scores for many games (I'm a small-time/amateur GPU reviewer), no time to re-test those today but I'll do it maybe tomorrow.
All results above with GPU in stock settings. I also tested TimeSpy with my overclocked GPU profile, got almost same score: +1.70% for the base test, +1.36% for Extreme. This is also not normal, my OC-versus-stock improved TimeSpy by +2.42% and Extreme by +6.86% in my last round of testing with the same overclock profile.
Mos abnormal results are for DX12 benchs, including the Intel XeSS feature test, although there's also VRMark which is DX11, and big DX12 benchs like SpeedWay/PortRoyal/DXR with perfectly sane results.
But we also looked at our database to compare results on the previous driver vs. new driver on any result using 7900XT or XTX and can confirm that this appears to be a real, if very rare issue. Among all results with the new driver, approximately 3% of the results show abnormal (very low) scores on Time Spy. No similar group of very low scores appear on the results with the previous driver version.
This would lead us to believe that this is some kind of issue with the latest driver that requires some additional conditions to be true to appear. We've raised the issue directly with AMD and are waiting for their response.
Since benchmark itself has not been modified for ages, our current view is that this is a rare driver bug that somehow causes the system to run in some kind of low power mode while Time Spy is running, which then tanks the performance to a very low level. Benchmark itself is quite simple - it just renders 3D graphics as fast as possible and measures the fps - and the code has been vetted and audited quite extensively, so a bug in the test seems very very very unlikely, so the reality is that for some reason, this specific 3D graphics load on the latest driver can sometimes rarely result in GPU running it very slowly.
At this time I would recommend simply disregarding the result as a driver issue and await for future AMD driver updates to solve it. If you REALLY want to try to get rid of it, next step I would do would be to try a clean OS install, but obviously this is not a good solution for a home system that is in active use, and probably not worth the effort at this time.
https://www.3dmark.com/compare/spy/36500772/spy/35770210
Easy to notice the much lower GPU clock frequency, looks like the driver is putting the GCD to sleep, also causing GPU temps to be 12C lower.
Today I ran a good number of game benchmarks - Callisto, SOTR, Quake II RTX, Guardians of the Galaxy, Horizon, Dying Light 2, Batman Arkham Knight, Returnal - every one of them performed normally, It seems there is something very specific in the rendering engines used by some of the 3DMark tests that this driver doesn't like and only in some specific circumstances since you can't repro.