Oct 13, 2023
4
1
15
Description
Hi, I need assistance with my recent problem. In some games (especially Dying Light 2, Detroit Become Human) I have a problem with sudden frametime spikes (?). In the case of Detroit, the problem occurs at specific times, while in Dying Light 2 simply along with movement, the same thing happens in the built-in benchmark. From the indicators I can't read any unknown aberrations. When frametime goes up then GPU clocking and GPU memory clocking go down, while CPU goes up. All temperatures, voltages are normal. The computer passed stability tests in furmark, 3D Mark, Memtest86. No problems detected. Benchmark scores are also within the norm: AS SSD, 3D Mark (Time Spy Extreme, Storage Benchmark), Cinebench R23. LatencyMon did not indicate anything. All test sequences were executed with minimal background applications.

I started noticing the problem around the time I replaced the drives - from an Adata Gammix S11 Pro to a Lexar NM790 and using a Thermalright contact frame. Then I installed a clean Windows 11 Pro - stable channel. Before that, I was running Windows 11 Pro on the beta channel.

Detroit Become Human problems preview
What is weird is that after reloading the scene, this moment runs smoothly. The problem will recur after restarting the game. This repetitiveness and the lack of a problem the second time the scene was loaded suggested to me a problem with the shader cache (??).

Please help. I have run out of ideas. :(

First troubleshooting attempts

On this clean installation of Windows 11 Pro, depending on several consecutive Windows "quality" updates, the problem worsened or waned. It never went away. I control-installed Windows 10 Pro (disabling E cores due to lack of communication with Thread Director) and the problem was then the least, but it still existed. I checked to see if the contact frame was causing a lack of contact with the pins. I checked if the problem occurs on new, moderately old and very old AMD Radeon drivers.

The rest of the things I tried
-UEFI update/downgrade;
-re-seating GPU;
-clean GPU driver install;
-RBAR On/Off;
-XMP On/Off;
-CPU OC On/Off;
-Freesync On/Off;
-MPO On/Off
-Game Bar Off;
-Game mode On/Off;
-Elevating games priority via Task Manager;
-etc...
I don't remember more than that. In total, I troubleshot for ~~30h on my days off per month.


Full specification
i5-12600K (Enhanced Multi Core Performance enabled, stock)
Gigabyte Z690 Gaming X DDR4 (F27, newest UEFI)
Powercolor RedDevil RX 6900XT (OC bios, RBAR enabled, stock)
Corsair 32GB (2x16GB) 3600MHz CL18 Vengeance RGB PRO SL (XMP enabled, Gear 1, dual channel)
Lexar NM790 1TB (running on pcie 4.0x4, system)
Lexar NM790 2TB (running on pcie 4.0x4, rest)
Corsair RM1000X
Arctic Liquid Freezer II 240mm
Windows 11 Pro (22H2, 22621.2428, Windows Feature Experience Pack 1000.22674.1000.0)
Monitor LG 27GP850
 
Hey there,

Have you cleared CMOS each time you did a bios update? If not, it might be worth a shot now, to be sure.

Does the odd behaviour coincide with the Windows 11 install? Given they are beta/channel, it's very likely it's a bug. I guess you'd need to install the current consumer version to be sure.

Was the old drive connected when you did the new install on the new drive?
 
Oct 13, 2023
4
1
15
Hey there,

Have you cleared CMOS each time you did a bios update? If not, it might be worth a shot now, to be sure.

Does the odd behaviour coincide with the Windows 11 install? Given they are beta/channel, it's very likely it's a bug. I guess you'd need to install the current consumer version to be sure.

Was the old drive connected when you did the new install on the new drive?
Hi, thank you for your reply.
When updating/downgrading UEFI it performs wipe automatically. I've manually reset it and still no changes.

No. This setup (on Gammix S11 Pro drive) was working smooth for 2 years on Windows 11 Beta. Switching to W10 was only for testing purposes. Currently working on W11 stable. No changes.

The old disk is not mounted. Only Lexars were connected during the installation.

Normally, I would say that it is the fault of the drives, since there is a coincidence between the occurrence of the problem and the upgrade. On the other hand, crystal disk info does not confirm any errors, in AS SSD the drives do the results as declared by the manufacturer, in 3D Mark storage response time is typical. It's hard for me to accuse these drives, especially since everything is fine in Windows file operations.
 
Oct 13, 2023
4
1
15
Rather than explain why (I'm lazy and don't want to type :) ).

This is why you should clear CMOS : https://forums.tomshardware.com/thr...faults-within-the-bios.3730428/#post-22493048

Look out for @drea.drechsler answer.
Hi, partially good news!

After clearing the CMOS (in the meantime the GPU drivers were updated to 23.10.2), a few changes appeared.

1. Detroit Become Human:"Helicopter sequence" lag reduced to ~~0,1s - still noticable but real less disturbing.

2. Scenes are loading at least twice as fast now.

3. CPU utilization when playing DBH dropped to around 40/50% rather than middle 80's.

At this point in this game i consider that minor problem as driver shader cache issue (game is basing on Vulkan).

Unfortunetely in Dying Light 2 no changes so far.
Everything is tested in borderless. I guess that the game would have better performance on fullscreen mode (Windows optimizations etc.), but it is broken(?) in my case. When I select fullscreen mode, the monitor goes blank and after a second it turns on at 2560x1440, but only 60Hz and I can't do anything about it. Tests in such settings will not reflect the scale. In frameless the game runs normally at 2560x1440 165Hz with freesync.

Only in this game I have a problem, the rest of the games work in fullscreen 2560x1440 165Hz freesync.
 
Hey,

Excellent. That's a result in my book :) Glad it's sorted for the most part.

Asides from doing another DDU on the GPU driver, it it's still persistant in Dying Light, then it's possibly some bug that may be solved with a game update.

In the meantime, good luck and happy gaming :)

Let us know if you've any more issues.

One other thing to try: If you use an overlay, like steam/MSI or whatever, then try without. They do sometimes cause minor stuttering. You can also try changing the mouse polling rate. That may also help.
 
Oct 13, 2023
4
1
15
Hey,

Excellent. That's a result in my book :) Glad it's sorted for the most part.

Asides from doing another DDU on the GPU driver, it it's still persistant in Dying Light, then it's possibly some bug that may be solved with a game update.

In the meantime, good luck and happy gaming :)

Let us know if you've any more issues.

One other thing to try: If you use an overlay, like steam/MSI or whatever, then try without. They do sometimes cause minor stuttering. You can also try changing the mouse polling rate. That may also help.
Hi. The problem has reappeared. It's practically back to the initial effect, except that now the CPU is running at 100% all the time (DL2), while the graphics card fluctuates between 60-98%. Bottleneck is impossible because I specifically selected the processor and card two years ago so that there would be no bottleneck. Two years I played usually so that it was the card that was maxed out and the CPU was within 60%.

In view of this, I checked whether disabling virtualization and C-States would change anything. No change.

In the next few days I will check if all CPU pins are straight and I will go back to the stock ILM to see if it will change anything. If nothing helps, the idea is to eliminate the possibility of a Windows-side error - I'll try to fire up the games on some reasonable Linux distro, but that will be quite an invasive step for my data and its migration, so I'll wait with that for now.

I will keep this thread updated. Maybe someone who also has these problems will benefit from my experience.
 
  • Like
Reactions: Roland Of Gilead