GTX 980 Ti SLI Severe Performance Loss with G-sync Monitor

[gridironcpj]

I posted about this almost a year ago when I had two GTX Titans. We never got to the bottom of the issue, so I just ignored it and took the performance hit. I was under the impression that maybe the GTX Titans in SLI just had issues with g-sync. However, as a new owner of two GTX 980 Ti's, I can confirm that the issue is not limited to Titans and seems to exist for any card.

So here's the issue. When I enable SLI, set single display performance and maximum performance, and turn off g-sync with my PG278Q as my only display active, I take a 17-18 percent performance hit in 3Dmark in my graphics score. If I use my PB278Q (my other monitor) as the only display, but keep my PG278Q plugged into my primary 980 Ti, the problem still persists. However, when I unplug the PG 278Q from my PC, my performance skyrockets 17-18%. Even more strangely, this issue does not persist when I disable SLI. It's only an SLI issue. I Google searched the issue and nobody seems to have the same issue. What in the world could this be? Should I RMA my PG 278Q? The monitor works fine, it's just that damn performance hit in SLI..

[aeliusg]

I posted about this almost a year ago when I had two GTX Titans. We never got to the bottom of the issue, so I just ignored it and took the performance hit. I was under the impression that maybe the GTX Titans in SLI just had issues with g-sync. However, as a new owner of two GTX 980 Ti's, I can confirm that the issue is not limited to Titans and seems to exist for any card.

So here's the issue. When I enable SLI, set single display performance and maximum performance, and turn off g-sync with my PG278Q as my only display active, I take a 17-18 percent performance hit in 3Dmark in my graphics score. If I use my PB278Q (my other monitor) as the only display, but keep my PG278Q plugged into my primary 980 Ti, the problem still persists. However, when I unplug the PG 278Q from my PC, my performance skyrockets 17-18%. Even more strangely, this issue does not persist when I disable SLI. It's only an SLI issue. I Google searched the issue and nobody seems to have the same issue. What in the world could this be? Should I RMA my PG 278Q? The monitor works fine, it's just that damn performance hit in SLI..

Is all your PCI-E stuff sorted out? Do you get the same thing in games or is it just benches? Sounds annoying, good luck.

[Bishi]

Possibly a stupid question but are you sure you don't have vsync on with gsync? That means it will limit your frames when > monitor refresh. You should use vsync off for benchmarking (even if you have GSYNC on). The other thing I would suggest is checking to see if there is anything else using the GPU that could be causing this (chrome, logitech software, steam, etc)

[Jarnis]

Remember: Any benchmark that is based on framerate must be run at g- and v-sync disabled, or otherwise you can end up capped by the maximum refresh rate of your monitor.

So for example, with 3DMark, if you have 144Hz screen with g-sync on, and reach more than ~140fps at any point during your benchmark, your score will be lower than if g-sync was disabled. Nothing wrong, just a fundamental incompatibility of g-sync and the value the benchmark is measuring.

Also there appears to be a small performance penalty (few percent) when g-sync is enabled even if your framerate never reaches your display refresh rate.

(also some older versions of 3DMark capped the tests to 60hz, making this issue much easier to run into. Most recent version uses your desktop refresh rate, so the "cap" with g-sync on equals to that)

Edit: Looks like recent drivers introduced the feature to disable v-sync alone, leading to tearing once you go past the refresh rate of the monitor. So in theory if you have v-sync off but g-sync on, you should still get a proper result from a fps-based benchmark but you will see tearing if framerate goes beyond your display refresh rate maximum. Haven't personally tried this yet.