5800x to the 5900x/5950x: Can BeamNG make use of it?

Currently, I have been using the 5800x since 2020 and it has served me well. Framerates in BeamNG hasn't been stellar while being paired with a GTX 970, but this year I had the opportunity to upgrade it to one of NVidia's Super cards and the performance became so much better.

But now that I've upgraded my GPU to a much better one... My 5800x has become the new bottleneck.

Performance is very "par". In DirectX mode, the average with traffic seems to be around 60FPS in West Coast, while in Vulkan it's 90FPS. All in all, fine framerates... but what I want are even more cars!

From my research, it seems that BeamNG uses "one thread per car" which means that the higher core count of the 5900x and the 5950x should mean immediate improvements in vehicular volume and potentially framerates too.

But this is all in paper. A post from Gabester back in 2013 says that "in theory you could have a lot more cars at a playable framerate if you had enough threads." But that's just a theory.

In the era of consumer 12-16 core CPUs, I now wonder if this is truly the case.
To anyone with a 5900x or 5950x, do you see this reflected on your system? Because if it is, I may decide to upgrade to the 5900x or even the 5950x if the improvements are really that obvious.



P.S.I would upgrade to AM5 7900x if I could, but that would require starting off a fresh install of Windows which I don't wanna deal with (Windows doesn't like moving to completely new motherboards/chipsets, see here)

 

this has been a question ive been wondering since forever...
until recently... i snatched a 5950x on a nice sale.
i sadly dont have a bananabench test from my 3700x cpu, but i could add one later to compare if needed.
i also recently build a pc for my dad with a 5800x3d, i tested bananabench on it as well.

between the 5800x3d and the 5950x, its very close.

5950x


5800x3d



with more cars the 5950x works a bit harder, but im gonna be honest, im a bit disappointed in the performance of the 5950x in beamng.

but so many things factor in the performance.
first bananabench i did with my setup, 3000mhz ram instead of 3600mhz in the pic above


from 626 to even short of 600 ive seen in some worse test. testing can be very inconsistent

all in all, i was very surprised at how well the 5800x3d handled beamng for an 8 core.
i havent been able to test beam more on his 5800x3d setup, but i can finally handle 32 simple traffic vehicles at around 20-30 fps ish.
or 16 cars at 60 fps with my 5950x. but keep in mind its simplified traffic. with the full complex vehicles it will lower it. dont remember the numbers id have to go test it but i was not having 30 fps with 20 cars if i remember well... even for a 5950x...

idk what i was expecting, i dont have a 4090 either, im running beamng with a vega FE. beamng loves its 16gb of vram. but its a limited gpu if you want to add lots of traffic.. its showing its age.


honestly, id love to hear more about this. to my knowledge it has always been a rule of thumb, 1 car per core, so more cores = better performance. but i feel like there is a bottleneck somewhere.

 

When the 3D cache helps as much as it does here, the 5950X is probably RAM limited, because it has to re-read the same code/data more often than the 5800X3D. Like in Factorio lategame where it trumps CPUs with much more threads

 

Interesting! Thank you for the deep insight!
What's that script you're running by the by? I never knew such a thing existed and would like to try it out myself!

Regarding the bit about a bottleneck, have you tried looking the Performance Graph within BeamNG (Ctrl + Shift + F)? It can indicate what component is currently bottlenecking your performance, though it has a hard time differentiating between CPU and GameEngine bottlenecks.
I would recommend testing it in Vulkan, as DirectX itself is a severe bottleneck for BeamNG performance - has to do with BeamNG exceeding DX's draw calls.

But if running Vulkan is something you did try, then perhaps there does exist a bottleneck somewhere else.

 

interesting info, i've been tied between upgrading to a 5800x3d or a 5950x. this is very helpful

 

bananabench is included with beamng, you can find it by going in your beamng installation folder, in Bin64 there is a banana-bench.x64.bat that will run a script to benchmark your cpu.

its been a very useful quick way to test a cpu performance without actually opening the game
im glad the devs made that.

i do try vulkan once in a while and test things out. i think it would be a good idea to try it out again at this point.
could provide more useful info. havent looked into the performance graph since i got the 5950x.
thanks for the ideas!

 

I use 5700x and honestly, you don’t need anything extra. I can have beamMP servers with up to 25 players and keep a 30fps. In free roam 40 ai cars at 25 fps

 

ryzen 1700, geforce 1060 6гб, 32gb(2400) = 250 beams в банана бенч.

 

I would be getting a amd cpueventually but have to see whats the plan. then maybe I can do more testing
Anyways for the intel side: (copied from hardware forum, is my data)
my specs, for reference: i5-13600KF, Asus TUF RTX 4080, 2x16GB DDR5 6400Mhz CL32, 1440p 165Hz
DX11, WCUSA, 100 - 120 fps avg on 1 Vehicle, 78.4 fps (cpu limited) with 16 simplified vehicle + 1 vehicle, 32 fps avg (heavily cpu limited, gpu was at 50w, cpu at 122w) with 16 normal vehicle + 1 vehicle
Vulkan, WCUSA, 194.6 fps avg on 1 vehicle,107.4 fps avg with 16 simplified vehicle + 1 vehicle, 84.9 fps avg (cpu limited, gpu was at 180w, cpu at 114w) with 16 normal vehicle + 1 vehicle