New finds in micro-stutter issue

Everything about latency. Tips, testing methods, mouse lag, display lag, game engine lag, network lag, whole input lag chain, VSYNC OFF vs VSYNC ON, and more! Input Lag Articles on Blur Busters.
Post Reply
andrelip
Posts: 162
Joined: 21 Mar 2014, 17:50

New finds in micro-stutter issue

Post by andrelip » 28 Nov 2023, 08:59

While tweaking my overclocked system, I encountered micro-stutters and explored their cause.

Scenario A: When all P-cores are set at 55, the performance is smooth.
Scenario B: With 4 P-cores at 56 and 4 at 55, micro-stutters appear every a 2 seconds.

In both scenarios, the system successfully passes Prime95, Cinebench, and MemTest without issues. There's no thermal throttling or any other limitations according to HWInfo, and temperatures and power readings are comparable in both setups. Additionally, there's no Load-Line Calibration (LLC) or other OC adjustments between these configurations.

This observation leads me to two possible theories:

Theory 1: The issue might stem from timing inconsistencies. The asymmetrical frequencies of the cores could be causing glitches in the Time Stamp Counter (TSC) and multi-threading operations. Seems like the game is trying to force-resync when time drifts too much between cores or between what was expected by the game. Interestingly, varying speeds in E-cores don’t seem to cause similar issues, possibly due to the operating system or games treating them as separate groups, thus compensating accordingly. It's worth noting that while you might have all cores set equally and still encounter the same issue, in my case, the pronounced frequency gap may have made the problem more apparent. In your setup, factors like core-parking or C-states might be contributing to similar issues.

Theory 2: The issue might simply be a lack of system stability. Although the system passes all stress tests without errors, I wonder if there's a method to assess stability more precisely. Perhaps a technique to monitor error corrections could offer a more detailed insight into system stability.

System Setup:

Intel Core i7-13700K
Asus Prime Z790 Motherboard
G.Skill 6000CL30 Memory
Lian Li Galahad II 360 Performance AIO Cooler
Nvidia GeForce RTX 4090
Windows 11 Operating System

cursed-gamer
Posts: 80
Joined: 16 Aug 2023, 13:07

Re: New finds in micro-stutter issue

Post by cursed-gamer » 28 Nov 2023, 10:05

Theory 1 sounds interesting. There is definitely something wrong with timings because during every gaming session I have a feeling that my PC is "dragging an anchor" and "chasing" the action on a server. I can clearly see that enemies' position is being adjusted all the time.

My CPU: AMD Ryzen 7 5800X (no OC)

What do you suggest to do or how would you identify that something is going fishy with TSC?

Thatweirdinputlag
Posts: 308
Joined: 27 Aug 2021, 14:09

Re: New finds in micro-stutter issue

Post by Thatweirdinputlag » 30 Nov 2023, 01:48

andrelip wrote:
28 Nov 2023, 08:59
While tweaking my overclocked system, I encountered micro-stutters and explored their cause.

Scenario A: When all P-cores are set at 55, the performance is smooth.
Scenario B: With 4 P-cores at 56 and 4 at 55, micro-stutters appear every a 2 seconds.

In both scenarios, the system successfully passes Prime95, Cinebench, and MemTest without issues. There's no thermal throttling or any other limitations according to HWInfo, and temperatures and power readings are comparable in both setups. Additionally, there's no Load-Line Calibration (LLC) or other OC adjustments between these configurations.

This observation leads me to two possible theories:

Theory 1: The issue might stem from timing inconsistencies. The asymmetrical frequencies of the cores could be causing glitches in the Time Stamp Counter (TSC) and multi-threading operations. Seems like the game is trying to force-resync when time drifts too much between cores or between what was expected by the game. Interestingly, varying speeds in E-cores don’t seem to cause similar issues, possibly due to the operating system or games treating them as separate groups, thus compensating accordingly. It's worth noting that while you might have all cores set equally and still encounter the same issue, in my case, the pronounced frequency gap may have made the problem more apparent. In your setup, factors like core-parking or C-states might be contributing to similar issues.

Theory 2: The issue might simply be a lack of system stability. Although the system passes all stress tests without errors, I wonder if there's a method to assess stability more precisely. Perhaps a technique to monitor error corrections could offer a more detailed insight into system stability.

System Setup:

Intel Core i7-13700K
Asus Prime Z790 Motherboard
G.Skill 6000CL30 Memory
Lian Li Galahad II 360 Performance AIO Cooler
Nvidia GeForce RTX 4090
Windows 11 Operating System
I have an identical system and without touching the bios other than the usual stuff like XMP etc. Those micro stutters come and go, i'd say in a week i'd have 5-6 days of stuttering some of which are eye straining enough for me to quit playing after 10 minutes. And 1- 2 days the game is just butter smooth. Same exact scenario was happening with me previous build "9900KS, 2080ti". even though the 9900KS was constantly pushing 5GHz on all cores, I'd still have days where I just can't play due to the micro stutters and other days were okay. On both systems i've tried to have core-parking and C states "plus other energy saving features" be on and off and the stuttering was independent of their on/off states.

That said, on the new rig Valorant didn't seem to phased by this, the game is always smooth as long as I set the FPS counter past 330, off or go with the Gsync + Vsync method. However, I don't really play Valorant anymore, the game that I play is still behaving like it used to on the old rig even while pushing beyond 250fps. Yet it was somehow butter smooth on my 2015-2018 4-core intel system that barely pushed 80-100 fps on it.
Rog Strix Z790i - Intel 13700K - 4090 OC ROG Strix - 7200 Trident G.Skill - 1TB SK Hynix Platinum P41 - 1000W ATX3.0 Asus Tuf - 34'' Odyssey OLED G8 - FinalMouse Tenz S/Pulsar Xlite V2 Mini - Wooting 60HE - Sennheiser HD 560s - Shure SM7b - GoXLR Mini

andrelip
Posts: 162
Joined: 21 Mar 2014, 17:50

Re: New finds in micro-stutter issue

Post by andrelip » 30 Nov 2023, 21:01

Thatweirdinputlag wrote:
30 Nov 2023, 01:48
[quote=andrelip post_id=99448 time=<a href="tel:1701179962">1701179962</a> user_id=528]
While tweaking my overclocked system, I encountered micro-stutters and explored their cause.

Scenario A: When all P-cores are set at 55, the performance is smooth.
Scenario B: With 4 P-cores at 56 and 4 at 55, micro-stutters appear every a 2 seconds.

In both scenarios, the system successfully passes Prime95, Cinebench, and MemTest without issues. There's no thermal throttling or any other limitations according to HWInfo, and temperatures and power readings are comparable in both setups. Additionally, there's no Load-Line Calibration (LLC) or other OC adjustments between these configurations.

This observation leads me to two possible theories:

Theory 1: The issue might stem from timing inconsistencies. The asymmetrical frequencies of the cores could be causing glitches in the Time Stamp Counter (TSC) and multi-threading operations. Seems like the game is trying to force-resync when time drifts too much between cores or between what was expected by the game. Interestingly, varying speeds in E-cores don’t seem to cause similar issues, possibly due to the operating system or games treating them as separate groups, thus compensating accordingly. It's worth noting that while you might have all cores set equally and still encounter the same issue, in my case, the pronounced frequency gap may have made the problem more apparent. In your setup, factors like core-parking or C-states might be contributing to similar issues.

Theory 2: The issue might simply be a lack of system stability. Although the system passes all stress tests without errors, I wonder if there's a method to assess stability more precisely. Perhaps a technique to monitor error corrections could offer a more detailed insight into system stability.

System Setup:

Intel Core i7-13700K
Asus Prime Z790 Motherboard
G.Skill 6000CL30 Memory
Lian Li Galahad II 360 Performance AIO Cooler
Nvidia GeForce RTX 4090
Windows 11 Operating System
I have an identical system and without touching the bios other than the usual stuff like XMP etc. Those micro stutters come and go, i'd say in a week i'd have 5-6 days of stuttering some of which are eye straining enough for me to quit playing after 10 minutes. And 1- 2 days the game is just butter smooth. Same exact scenario was happening with me previous build "9900KS, 2080ti". even though the 9900KS was constantly pushing 5GHz on all cores, I'd still have days where I just can't play due to the micro stutters and other days were okay. On both systems i've tried to have core-parking and C states "plus other energy saving features" be on and off and the stuttering was independent of their on/off states.

That said, on the new rig Valorant didn't seem to phased by this, the game is always smooth as long as I set the FPS counter past 330, off or go with the Gsync + Vsync method. However, I don't really play Valorant anymore, the game that I play is still behaving like it used to on the old rig even while pushing beyond 250fps. Yet it was somehow butter smooth on my 2015-2018 4-core intel system that barely pushed 80-100 fps on it.
[/quote]

I just built my old PC for my son. It has an Intel 9700k (5GHz) with a 3080. I used the Acer 390Hz monitor with that computer, and I was shocked by how much more it stutters compared to the 13700k. This weekend, I will double-check temp, stability and other things to confirm.

andrelip
Posts: 162
Joined: 21 Mar 2014, 17:50

Re: New finds in micro-stutter issue

Post by andrelip » 30 Nov 2023, 23:14

cursed-gamer wrote:
28 Nov 2023, 10:05
Theory 1 sounds interesting. There is definitely something wrong with timings because during every gaming session I have a feeling that my PC is "dragging an anchor" and "chasing" the action on a server. I can clearly see that enemies' position is being adjusted all the time.

My CPU: AMD Ryzen 7 5800X (no OC)

What do you suggest to do or how would you identify that something is going fishy with TSC?
If theory1 is correct, then I would ensure that all the cores have the same clock speed at all times and monitor this with HWinfo. If a core has an average speed that is different, then it can start to accumulate drift.

AyayronPlays
Posts: 19
Joined: 16 Nov 2022, 10:52

Re: New finds in micro-stutter issue

Post by AyayronPlays » 03 Dec 2023, 22:39

Have you tried disabling the iGPU?

Post Reply