Big difference in gaming performance between win7 & win10 on the same PC

[HD64G]

For the 2 games below the results are telling everything except for the cause of this difference. The settings in the game configuration are exactly the same for both oses. The GPU isn't fully used in either of those in win10 while in both of those in win7 the GPU usage is much higher.

DirtRally

Win7 pro 64

<fps_race min_fps="105.092873" max_fps="172.066467" av_fps="141.146957" min_fps_ms="9.515393" max_fps_ms="5.811707" av_fps_ms="7.084815" />

Win10 ult 64

<fps_race min_fps="116.963440" max_fps="215.634491" av_fps="163.754944" min_fps_ms="8.549680" max_fps_ms="4.637477" av_fps_ms="6.106686" />


The Witcher 3 Wild Hunt

When loading the same save game looking at the central plaza in Novigrad from above, while in win7 the FPS stabilises at 105-107 after a few seconds, in win10 there are some random flactuations from 85 to 98 FPS with the average being 96-97.


In Assetto Corsa I had the same problem and big difference but I don't have benchmarks for win7 atm.

There are several games that don't face the same problem as they are GPU-heavy but on the CPU-sensitive ones there is a significant problem in performance.

The security program I use is the same for both oses. The energy-performance profiles (balanced, ryzen balanced, high performance) have all tested and the problem still isn't solved. The RAM stability is perfect for weeks now after testing and finding proper settings. The games are installed once and are being loaded from the same HDD.

I think that either the CPU scheduler in win10 is not working well for my CPU or something in win10 eats plenty of CPU circles but cannot be found until now. So, anyone that can make an educated and not random guess? Anyone else owning a not 5000 series Ryzen CPU with which can do the same test to confirm those findings below?

 

[BSim500]

So, anyone that can make an educated and not random guess? Anyone else owning a not 5000 series Ryzen CPU with which can do the same test to confirm those findings below?

I take it the "W10 ult 64" is the W7 one (typo?) A lot of people have noticed that some games will definitely run faster on W7. CS:GO is one well known one (though not the only one) running up to +15% faster on W7 on a Ryzen:-

It's possible it's the scheduler. Or simply W10 bloat (ISO install size now +5.5GB W10 vs 3.1GB W7) and there's more than double the services. In fact I find it amazing how after one single "W10 2015 vs W7" test when W10 originally launched, no-one has since bothered to test if the +50% bloat added to W10 (2021 vs 2015) since then has changed things:-

There's nothing you can personally do about the scheduler but you could try one of the W10 debloat / decrapify scripts. Or try out LTSC (can't legally buy it as an individual but a lot of gamers seem to use it for a reason).

 

[Shou Miko]

Have you tried the version of Windows 10 or 11 that Ghost Spectre has modded this a lot of the times see great performance over the original Windows 10/11.

Link: https://www.youtube.com/c/GHOSTSPECTRE/videos

 

[delshay]

This is why I still use my old AM1 & AM2 computers for gaming. I think but not sure, but I think windows 10 & 11 have more background tasks running. Some of windows 10 features I disable them via my antivirus software at startup.

 

[mb194dc]

Try CPU Z benchmark on Windows 7 and 10 and post the results.

I also dual boot windows 7 and 10 and on a Ryzen 5 1600 @3.9 B350 machine and have seem similar.

 

[dgianstefani]

O&O shutup10 + Geek uninstaller (system apps).

 

[qubit]

I'm not surprised that it's slower with W10 and will probably be slower still with W11.

I first noticed this with DX9 between XP and Vista, with Vista being something like 20-40% slower depending on exact circumstance on the same hardware. This meant that cards which were good enough to play a game at 60-80fps suddenly dropped to 40-50fps, giving a really choppy, unacceptable performance.

This performance difference is still there today, although it's less obvious with more powerful hardware. The other thing of course, is that most games have now moved on from DX9 and drivers for modern cards no longer support XP, so one can't actually compare to confirm this is still the case.

 

[HD64G]

Nice to know there are more like me out there. Thanks a lot for the input guys and/or girls! And thankfully I kept my 7 install to run games that don't even start on win10 (Call of Juarez Bound for Blood, Neverwinter Nights 2 and the GTR2 racing game for example). Let's see if any Intel owner can validate our findings or I will post the Tomb Raider benchmark later today as one more clue to the MS sambotage on Ryzen performance that started from win10 and went on with win11.
BSim500 of course that was a typo. There is only one windows ultimate version until today...

Update: The Tomb Raider doesn't have any difference between the oses. Just 3FPS higher for the min and average on win10pro when the win7ult already reaches 128min and 176average FPS.

 

[mobiuus]

don't forget spectre and meltdown patches in win10 that also hinder cpu and memory performance...
i myself reinstalled win10 a week ago and noticed in 3dmark lower cpu performance...
and with exactly the same updates, software, drivers and cpu-gpu clocks

 

[qubit]

don't forget spectre and meltdown patches in win10 that also hinder cpu and memory performance...
i myself reinstalled win10 a week ago and noticed in 3dmark lower cpu performance...
and with exactly the same updates, software, drivers and cpu-gpu clocks

That's a good point. I wonder if one ran an old enough build of W10 without those patches, would it run faster?

 

[BSim500]

don't forget spectre and meltdown patches in win10 that also hinder cpu and memory performance...
i myself reinstalled win10 a week ago and noticed in 3dmark lower cpu performance...
and with exactly the same updates, software, drivers and cpu-gpu clocks

I don't know if they work on not, but the official registry keys for disabling mitigations are:-

reg add "HKLM\SYSTEM\CurrentControlSet\Control\Session Manager\Memory Management" /v "FeatureSettingsOverride" /t REG_DWORD /d "3" /f
reg add "HKLM\SYSTEM\CurrentControlSet\Control\Session Manager\Memory Management" /v "FeatureSettingsOverrideMask" /t REG_DWORD /d "3" /f

 

[erocker]

Has anyone tried using InSpectre? I remember it being posted here around when that totally legit 3rd party security firm alerted about these security issues. https://www.grc.com/inspectre.htm

 

[LifeOnMars]

Windows 11 working quicker on all APIs here with latest drivers, DX 9 no issues whatsoever. Suprised that some are having issues but Microsoft really should step up their game with making Windows an optimised gaming platform. Why have they still not released a gaming only variant of Windows? Surely in the long term they would have better results.

 

[HD64G]

From the tests conducted either for windows 10, either for Linux, the gaming difference for Ryzen CPUs shouldn't be over 2-3% when the spectre and meltdown patches are applied since Intel CPUs should be hit much harder and even those didn't lose more than 5%. I will run the same benchmarks with and without those using InSpectre and will report back later.

Some food for thought on this topic:

https://www.extremetech.com/computing/291649-intel-performance-amd-spectre-meltdown-mds-patches
https://www.anandtech.com/show/1365...ith-spectre-and-meltdown-hardware-mitigations
https://www.tomshardware.com/reviews/gaming-performance-meltdown-spectre-intel-amd,5457-6.html

UPDATE: Not difference at all (as expected) with those vulnerability patches disabled. So, some ms services or the cpu scheduler difference between ms windows 7 & 10 make the difference most probably.

 

[RJARRRPCGP]

Windows 11 working quicker on all APIs here with latest drivers, DX 9 no issues whatsoever.

I hope the same for Halo Custom Edition. (a.k.a. Halo CE) That's DX9.