[SOLVED] Diagnosing Gaming Performance Bottleneck: Stuttering Frames

[GammaDentist]

I'm troubleshooting my performance with Overwatch, where if the fps is high enough sometimes the screen stutters for ~0.3s.
I suspect it's simply my PC not being able to keep up with 240 fps, so I tried benchmarking.

I'm rather surprised.
Why are there crazy spikes in CPU temperatures?
Since my CPU load is <20% and my CPU temperature already peaks at ~70 °C, is this indicative of insufficient CPU cooling?
Or is the GPU hovering around 80% load the problem?

CPU: AMD Ryzen 7 5800X
GPU: MSI GTX 1060 Aero ITX OC/ARMOR OC
RAM: 32GB DDR4
MB: ASUS TUF Gaming B550M-Plus
Cooler: Scythe SCMG-5100T

 

[Lutfij]

Welcome to the forums, newcomer!

What BIOS version are you on for your motherboard? What is the make and model of your rams? Which version of Windows 10 are you on? What version of Nvidia's GeForce drivers are you on?

 

[GammaDentist]

Welcome to the forums, newcomer!

What BIOS version are you on for your motherboard? What is the make and model of your rams? Which version of Windows 10 are you on? What version of Nvidia's GeForce drivers are you on?

SMBIOSBIOSVersion 1401
Crucial BL16G32C16U4B.M16FE
GeForce 465.89 (just updated to 471.96, same test same situation)
Are these what you're looking for?

By the way, I'm rather curious, is this info generally useful to diagnose problems?
Or is there something suspicious from the graph?
I did some more testing with Overwatch and during the hiccups, there is a spike in

1. simulation delay (time for current frame to get computed and displayed)
2. ping, following the simulation delay spike
3. packets incoming/outgoing per second

My current hypothesis on these stats are that since there are no packet loss (loss in/out), they are being affected by the stutter,
not the other way around.
Latency varies in a jigsaw pattern, so it is clearly interpolated between packets.
After the client stutters, there will be a pileup of packets, and they get cleared within the next frame.
Since multiple packets gets processed within the same frame, this causes the latency to reach 0ms, which is obviously impossible.
Other net stats follows similar logic, but are calculated differently so will present themselves as spikes.

 

[GammaDentist]

I have a new lead.
I have a dual monitor setup, one is a really old ASUS VS229 (60Hz) and the other is a new ASUS VG279 (240Hz).
When I enable VS229, the lag spikes disappear, but VG279 gets locked at 60Hz, which is unacceptable (obviously).
When I disable VS229, the lag spikes appear.
So somehow, VG279 running as the lone monitor has something to do with the lag spikes.
Is there something obvious I should check or tweak first in this case?

I tried turning G-Sync off and the spikes persist.

Then I tried turning off ASUS adaptive sync and the spikes still persist.
However, apparently this piece of garbage technology is exasperating the lag spikes by many folds.
As a side note, it also adds an insane amount of input lag, whoever didn't turn it off is in for a treat.

 

[Phaaze88]

Is there something obvious..?

The gpu actually trying to sync 2 different refresh rates and tripping over itself.
This has been a problem for a long time - for example: https://www.reddit(dot)com/r/pcmasterrace/comments/7h9bnu/psafor_people_with_dual_monitors_with_different/ (replace dot in parenthesis)
There's a number of possible fixes there, but the most concrete one has been to just have monitors with the same native refresh rate - resolution does not matter.


As for the 5800X's thermals, there is a flaw that is not shared with the 5600X, 5900X, and 5950X.
At the very least, take a look at the first 2 lines of the 1st post here: https://www.techpowerup.com/forums/...-3-chips-pbo-settings-temperature-fix.283776/

 

[GammaDentist]

The gpu actually trying to sync 2 different refresh rates and tripping over itself.

Actually,

When I disable VS229, the lag spikes appear.

I disabled the 60Hz monitor through windows display settings, do I have to physically disconnect the monitor?

At the very least, take a look at the first 2 lines of the 1st post here

Thanks, I'll fiddle around with some settings there if I find the thermal problems eventually causing problems.