Terrible GPU utilisation?

Hey guys.
I'm having a problem with my GPU only being utilized between 50-70% in game, giving me only 55-65 FPS max on most maps with the noticeable exception being the testing grid map. I've had this game across multiple PC's and it seems to stay stuck on that refresh rate. Is this game engine issue or is there something I can fix.

For reference my PC specs are as follows:
Ryzen 2700X 4.3GHZ
16GB RAM @ 2933
Vega 64

 

u sure you dont have the FPS limiter on 60 FPS?

 

Am sure, yes. I've checked all that.

 

interesting. im not sure what going on then. i hope you can get it resolved tho

 

@Nadeox1 I wonder what could be going on here

 

Yeah I hope it can be fixed. High refresh rates really help with the sense of speed on games that don't support VR.

 

I would try turning off dynamic reflections and/or SSAO. Beamng is very CPU heavy and you are being CPU limited. Even a 2700x likely can't keep up fully. Dynamic reflections and SSAO tend to add a lot of CPU load, so try disabling them first.

 

I can give that a shot but it surprises me that one of the highest end processors isn't enough when they recommend an i5, which at the time was 4c/4t.

 

An old i5 and a decent GPU will run the game just fine on lower graphics. The problem is dynamic reflects can (more than?) Double the CPU load if turned up to max, and while physics can properly multithread, graphics is mostly single core limited, so more cores doesn't help.

 

even a 9900k with exotic sub-zero cooling for a 6ghz+ 24/7 overclock wouldn't be able to handle truly ultra graphics under all circumstances

 

Settings are pretty well maxed, except for dynamic reflections which are about 50%. Running at 1080p

 

Turn faces per updates to minimum and distance of dynamic reflections to 80m turn off SSAO if enabled.

It is all about single core performance really, you can test that with banana bench, 1 car result left most column tells roughly how well your single core performance is.

I have seen really wide range for that model Ryzen, some have been really good result, some have been really poor results, not really sure why.

Maybe memory speed affects to that a lot in Ryzen, but not sure of that.

Anyway, on maps like Italy you get still be limited by single core performance, on other maps too if keeping dynamic reflections pretty much anything that minimum.

Of course if you run more cars, more easier it is to be limited by single core performance, it is just something most don't realize.

Light rays etc. also cause some CPU load.

Then another possibility is that you have mod that spams lot of errors to console ~ key to open, that can slow down a lot.

Also some software like Defender used to cause massive SC load making banana bench results lot slower some 6 months ago, not sure if they have fixed that yet.

Pretty much always if your GPU usage drops and you don't have vsync or fps limiter on, you will have either single core limit by CPU or so many cars your CPU is struggling, but last one is very unlikely to occur with Ryzen, single core limitation is first one to hit with ton of cars, even with fastest 6 core intel.

Also check out banana bench thread, there was oddly low result from one 2700X user and impressive high from another, so something to check out for sure:
https://www.beamng.com/threads/banana-benchmark-results-comparison.8595/page-46#post-977294
https://www.beamng.com/threads/banana-benchmark-results-comparison.8595/page-46#post-977531

 

By far the most helpful answer so far, you are a godsend.

 

It might be something to do with turbo clocks, at least with 6 core intel turbo drops off annoyingly fast with default settings, so one needs to turn settings around a bit in bios to get best out of it, maybe same for Ryzen?

Improving single car result in Banana bench should yield better fps, but as I don't have such system I don't know what causes lower scores, quite interested of what you can find out though.

 

Nah mine is 4.3ghz OC, so clocks don't change.

 

The game doesn't rely on the GPU as much as it should. This is due to poor optimizations as far as I can see, I have brought this up to the devs and I don't think they have any plans to change it. I'd suggest lowering GPU clock speeds (if possible) to increase fps. This, in theory, will make the system more well rounded for the game as it is CPU intensive. I can't say for sure it will give you a lot more fps but it may make it more consistent. Love to hear the results.

 

wtf? in what possible way would lowering gpu clock speeds increase the FPS?

you want both CPU and GPU tacked on as high as possible, so when it fluctuates between being cpu/gpu bound, you don't get the microstutter from the GPU or CPU ramping up its clock speeds. AMD made it fairly difficult to impossible to disable powerplay like 18 months ago in a driver update, but there's a fantastic little doodad out there called "clockblocker" which will fix you right up.

 

In games where the CPU is used more like Beamng, having an equally strong GPU can decrease fps. There are videos that explain this concept. I took some measurements to prove what I said is true. This is what I got with the standard default truck on the grid map with game version 15 something. Everything is measured with EVGA Precision and Task Manager open to control and view load on both CPU and GPU. Game was not restarted during testing and no other application was opened. Highest GPU clocks refer to the highest stable clocks my 1050ti 4gb from EVGA can achieve ( 170 core and 550 memory) Lowered GPU clocks refer to -200 on core and -502 on the memory. FPS measured with Fraps. Max voltage and fan speed active on both GPU and CPU for consistent results. CPU locked at 4.4Ghz on all cores. This is what I saw. This was while being driven and crashing. (Min-max) 2 minute test Period.

1 car, highest GPU clocks= 80-102 fps

1 car, lowered GPU clocks= 90-108 fps

2 of the same cars, highest GPU clocks= 55-67 fps

2 of the same cars, lowered GPU clocks= 54-62 fps

It appered that with more cars the higher the GPU clocks the better but with 1 car the opposite is true. I'd say it comes down to system hardware and vehicle and maps used. However, I still think its worth trying lowering the GPU core to see what happens however results will depend on multiple variables including but not limited to specs, game version, etc.

 

care to cite your sources and provide actual performance graphs m'guy

also 1050 ti vs vega 64 and whatever you have vs 2700x heehee