I've been trying to find a fix for these issues for around 6 years, doing all kinds of troubleshooting and controlled experiments.
I recently came across something unusual that seems to reduce desync in my experience. I'm not claiming this is a fix, and I don't know why it might work. I want to see if other people can reproduce it and whether it could lead to further investigation.
What I'm doing
Before launching the game, I open OCCT → Stability test → Combined, enable VRAM and 3D Adaptive, set both intensities to around custom→40%, and start the test (continuously running it in the bg). I also run Intel PresentMon in the background with its overlay enabled. Make sure Intel PresentMon is set to the 'GPU Focus' preset. PresenMon make a bigger impact than simply running OCCT alone imo.
I've tried this in Valorant and Overwatch. The effect isn't 100% consistent, and the required intensity varies depending on the game and its settings. Valorant sometimes needs 50–60%. In Overwatch, low graphics settings might require around 60%, while high settings might need only 40% : Depends on the hardware load ig.
The idea is to find a load level where the FPS starts taking a noticeable hit and then check whether the desync-related symptoms change.
Things I've noticed (What needs to be met for this to work)
- GPU utilisation: Needs to stay around 99% consistently.
- PresentMon graph: GPU Busy and frame-time/presentation traces appear to line up closely, almost like a single line.
- FPS metrics: The 1% low FPS under the Presents metric is significantly lower than the average Presents FPS.
- Trade-off: FPS takes a hit, and frame pacing can become uneven, with occasional spikes.
The symptoms I'm referring to are things like dying behind walls, taking damage after reaching cover, inconsistent hit registration, and similar desync-related behaviour. I'm not claiming this improves input lag, smoothness, or frame pacing.
This might sound somewhat similar to the iron box experiment people have discussed. I'm wondering whether introducing a particular system load changes something related to power behaviour, CPU/GPU scheduling, or frame delivery. This is just speculation, though.
Would anyone be willing to test this with me?
I'd like to compare the game running normally against the game running with OCCT and PresentMon, keeping other settings as consistent as possible.
If you participate, please share your CPU, GPU, OCCT settings, GPU utilisation, PresentMon graphs, and whether you notice any difference in the desync-related symptoms.
I know this isn't an ideal solution, especially with the FPS loss. The goal isn't to claim we've found a fix, but to establish whether this behaviour is reproducible across different systems and whether it gives us a starting point to investigate the underlying cause.
My PC :
CPU: AMD Ryzen 7 5800X
Motherboard: MSI MAG B550 TOMAHAWK (MS-7C91)
GPU: NVIDIA GeForce RTX 3060 Ti
RAM: 32 GB (2 × 16 GB) BIWIN Predator Pallas DDR4
SSD: Acer Predator GM7000 1 TB NVMe
PSU: DeepCool PN850M 850 W 80 PLUS Gold
CPU Cooler: DeepCool GAMMAXX L240 240 mm AIO
Cabinet: ANT ESPORTS 690MT AIR
Monitor: BenQ MOBIUZ EX2510S — 1080p, 165 Hz
Operating System: Windows 11 (Build 26100)
