Since upgrading my CPU (i9-9700KF) my GTX 1070Ti is maxed out for a lot of the time and the developer console is mostly reporting ‘Limited by GPU’.
I wanted to squeeze a bit more out if my GPU but I heard that MS recommend you don’t overclock the GPU and indeed when I tried overclocking by a relatively modest amount the sim wouldn’t even start.
So I was wondering if anyone’s found a settings that are stable and offer a noticeable performance benefit.
I’m not a fan of overclocking myself, so I keep everything at stock. I also have limited by GPU… But I’m guessing everyone will have them as long as it’s under the monitor refresh rate. Which I think everyone would be running 30 fps on a 60 Hz display anyway.
Open MSI Afterburner and increase the fan speeds across the curve, and increase the power limit. That’s the easiest way to gain more FPS with no risk of instability.
same on my 2070 was able to get max 220mhz on core clock any higher the sim will crash.
Also maxed the power limit slider
left the voltage alone as think this is the only real possible way of damaging the gpu other than overheating it.
to get my setting i run the afterburner oc scanner within the afterburner app and let it complete was able to up it slightly on the core clock from the scan result. Just trial and error to make sure the sim is stable.
The only time oc will be an issue with msfs is if you push it to far myself I have had very few crashes since release and been oc’d most of that time.
With a Power Limit of 110 the game runs fine, at 115 it won’t go past the home screen. At 110 and Core clock set to >200 (speed of about 2050MHz) it ran a bush trip for a while then crashed. So I can get away with just a small amount of OC, with core clock speed around 2000MHz. Maybe to increase the power limit I need to adjust the core voltage as well or something.
I ran the OC scan as suggested. After 40minutes it exported the curve, but with a message saying something like ‘overclock considered unstable’. The curve set the core clock to about 1975MHz, so a bit less than where I had it.
I was thinking the same thing, I’m sure I ran it while experimenting with OCing my CPU, ran it at 4.9GHz.
Can you explain what you mean by that? I thought Limited by GPU or Limited by Main Thread was
dependent on the relative GPU/CPU strength.
Well, someone can correct me if I’m wrong, but from what I’m experiencing is that the data flows from CPU into GPU then into your Display. If your CPU is underutilised, it can’t process as much data as it needs to, so by the time it’s delivering those assets into the GPU to render, the GPU is also underutilised making you getting low FPS on the display. Thus this shows as “Limited by Mainthread”
But when the CPU is fully utilised, it has processed everything that it needs to process and it’s ready for the GPU to render on the display. But now the amount of details and quality that needs to be rendered is dependant on the GPU itself. But it’s not a bottleneck. It’s either the GPU can render everything it needs to based on your details and graphics settings, or if it’s struggling. Either way, no matter how well or how bad your GPU is rendering the frames, if you have a 60 Hz display, the GPU would try to render 60 FPS to Vsync the display (if it’s turned on). But based on your graphics settings and details, if the GPU is rendering too many high detail and it can’t deliver 60 FPS. That’s when it’s technically “struggling” but while the reality is that it’s actually not. That’s just how much that the GPU can do for that graphics settings.
So since most displays are 60 Hz, but for flight simming 30 fps is an acceptable frame rate. If people are not having CPU bottleneck, everyone would be getting “Limited by GPU” regardless. So I don’t think it’s a bottleneck, that’s just how it is. Maybe by reducing the graphics to get the FPS to match the display refresh rate, that’s when the “limited by GPU” to stop showing since now it has enough headroom to render frames at the same rate as the display. Or you have a powerful enough GPU to push every frame in every detail level at your display’s native refresh rate.
Not really, if your underutilised CPU renders 50FPS and your GPU can only render 40FPS then the GPU will be maxed out and you’ll see Limited by GPU, no?
You mean it is processing everything it is able to?
The way I see it the data flows from the CPU to the GPU to the display. If the CPU is underutilised it’s because the game code isn’t optimised to use all cores or even one core fully.
If the CPU processes more data than the GPU can handle then you see Limited by GPU and the GPU is maxed out. Whether the rendered FPS is below 60FPS or over isn’t affected by the display. For the purpose of this topic let’s say the display is 200Hz, then surely it just comes down to which of the CPU or GPU can render more frames.
Which is why I was looking at overclocking my GPU, to be able to handle more of the frames the CPU is throwing at it.