I’m using UE5.1 on a high end Win10 system (RTX3090, i9-12900K, 128GB RAM), but even with a level that has only a single plane and a single directional light, unreal engine cannot reach FPS above 250. What is the bottleneck here? Is it possible to go beyond that? If so, how?
Btw, it didn’t go beyond 120FPS before I adjusted fixed frame rate to 999 in project settings. I turned VSync off in Nvidia settings. I also used r.vsync 0 in Unreal command console, which didn’t seem to change anything.
Go to Project Setting > Engine > General Settings > Framerate > enable Fixed Frame Rate and put the desired value up 999 fps
I already did that as I mentioned in the post 
I see now. Why do you need 250fps?
does it matter? i just need uncapped FPS as high as possible. Isn’t this a simple core issue for a game engine in terms of rendering?
Hi,
The question asked by Ayovchev is legitimate. Why would you go beyond 120 fps ?
In term of optimization, fps is not really something you can 100% rely on. You should look at the milliseconds it takes to render one frame (console command: stat unit)
If all you want is to maxed out the engine to test your hardware, well… Maybe there is limitation and the engine doesn’t reach over what its coded for. But you would need to have a very complex scene with heavy stuff in order to test your render time.
If you ever find out how to boost your FPS over 250, feel free to mention it
What I meant is that (in my humble opinion) the reason should not have anything to do with the solution to this core issue of a game engine.
But anyway, the long answer is i need to calculate precisely how the additional geometry i will add to the scene impacts the frame rate (FPS), which translates to total computation time of a frame as you put it. This is for my research on computational geometry. I need to understand the impact with as little added geometry as possible.
For example, when it was capped at 120 FPS (before I figured out how to achieve 250FPS), I couldn’t see any impact on 120FPS until I added a ton of geometry, i.e. it remained at 120FPS until the scene got really heavy. Now with 250FPS, I can see the impact of added geometry sooner than before but I still need to see it a lot sooner.
Plus, I saw videos on youtube where people achieve 300FPS on UE so I don’t understand why mine is capped at 250FPS with such a strong hardware.
Hi,
I believe that you might want to check the GPU profiler instead of relying on FPS only
Here some information that might help you: https://docs.unrealengine.com/4.27/en-US/TestingAndOptimization/PerformanceAndProfiling/GPU/
Cheers!
thanks for the recommendation. gpu profiler didn’t help though, because there wasn’t anything to discard. but I adjusted the project from Directx12 to Directx11. I think this did the trick. I can now see as little as 1.3ms in standalone mode
MS cost is a factor of DrawCall count. Drawcall count increases with resolution.
At 4k native, on a 3090, with an empty scene, you are probably lucky to hit 120fps.
Anyway, your point is more valid than theirs. There’s plenty of reasons to want to know what the maximum MS cost of an empty scene is in this pos of an engine.
What’s more is that by default, FPS adaptation is done by bloating/delaying the MS count of the buffer, so unless you prep the engine settings right you’d have no real way to see the difference when adding in geometry until you are way past the point where you already added in too much for most non 3090s to handle.
Second, the MS count of Pie(play in editor) means nothing.
The size of Pie drives the cost of rendering. and it’s not necessarily full screen, which means your most visible/common view is bound to be much different than the actual project running in standalone with the editor closed.
Third…
Editor windows drag down overall performance of everything - the engine renders 3d stuff even when you don’t expect it. The base scenes on many windows (mesh, persona, etc) have pretty high costs themselves. The more windows you have open, the more the drag on Pie or even on standalone can be.
Fourth.
While FPS usually changes, keep in mind that if you use something like FRAPS or any external utility, DX12 doesn’t play well with em.
The engine’s FPS counter is biased af, since they want you to think the engine is really good when it really, really isn’t all that great (particularly when you add in the new junk that sucks up the life out of your 3090 by heating it into oblivion).
The profiler tools are a much better way to get readings, but you probably shouldn’t look at FPS a anything much more than an “there’s a problem somewhere” indicator.
Fifth.
Gsync? Nvidia? There are many other places to look at that could be hindering the displayed FPS from being correct. Not just the engine…
How would you know?
Make the game resolution smaller, if the FPS doesn’t jump up, something else somewhere is capping it…
thanks. these are really useful info to keep in mind