BANANA Benchmark Results Comparison

Not much, usually 5 minutes.

 

Hi everyone!
I just did the banana test, and I was wondering if the result is only influenced by the cpu or by the overall performances?
So anyway here is the resluts for my AMD Ryzen 7 1800X with no overclocking (3600MHz) : 248.664 Mbeams/s

 

How do the input, audio and graphics run on the CPU? Do they run in separate threads or all run on 1 thread? for example in a Quad core system 4 cores / 8 threads, do they take up a single thread? I noticed a pattern that typically you can run 2 less vehicles then threads available. For example 4 cores / 8 threads means you can run 6 vehicles 60fps, the 7th vehicles can drop it down to 20fps

I'm under the impression that the audio, input, graphics and all AI run on a single thread, meaning no matter how many cores you have, the thread running those things can get bogged down, more so if running multiple vehicle AI's, it can slow everything down sort of like a wall/limiter (similar to how a slow HDD can slow down an otherwise fast computer)

 

Are there any 3950X benchmarks out there yet?

 

I haven’t seen any 3950x benchmarks yet, but I assume it won’t be the golden chip that can run 30+ cars in BeamNG it may run 16 vehicles well because of its 16 cores but adding more Vehicles then that would get hyperthreading involved Which seems to bog things down, I noticed on the different systems I’ve run, that you can run 1 or 2 less vehicles than cores and stay around 60fps playable with AI turned on, high speed chips can match cores/cars but once you get above the count of your cores the performance drops drastically despite hyperthreading. So a 3950x will get awesome results I’m sure in banana, but in real world should be able to run 14 cars well with AI, more than 14 and things go south.

General Rule: As many cars as cores @60fps

EDIT: my hypothesis: I’m suspecting there is some limiting factor, for example a threadripper 64 cores at 5ghz each and a Nvidia RTX Video card (Pick which ever fastest one is out there) you won’t be able to even run 30 cars @60fps because there are other processes that come more into play with so many cars loaded, those processes slow down the game despite having 64 cores at 5ghz each. Maybe a Dev can set the record straight though this is just a guess.

 

@Justy4WDTURBO @ThreeDTech21
There is one

 

For fun I ran the test with only 8 cores enabled. (7980xe)

 

5GHz! That’s an amazon processor, can you hit 5Ghz on all cores?

 

Yeah not a problem at all. I used to run 5.0 on all cores 24/7. I changed cases due to size and now I run 5.0 on most cores and a little less on the ones that run warmer.

 

8c16t or 8c8t?

 

8c16t

 

As previously posted, 18c36t is not limiting fps if under 36 vehicles. I would imagine as long as you have 1 thread for the game plus one thread per vehicle it will run fine.

 

Makes sense, when I used the Ryzen 2700x I was able to load 14 vehicles with decent FPS, so 1 thread seemed to be running the game, while another maybe ran windows 10. I once heard about the vehicles possibly being able to run on GPU cuda cores / stream processors in the future but don't know where I saw it.

EDIT
here: https://www.beamng.com/threads/cuda-opencl.500/
but its from 2013

 

Thanks for clearing that up! Also I was wondering is it possible to call other vehicles using the jbeam in a vehicle folder? For example I want to modify the roamer jbeam so that it can spawn in a trailer when the roamer is spawned, almost like adding a new part/slot but instead of activating a jbeam in the folder it can get jbeams in other folders. I think that is how the "common" folder currently works.

 

Ok, I just ordered a new ryzen 5 3600 and a new RAM, next week will be here, but meanwhile I opdated BIOS and overclocked a bit the actual 1600 and I get a 29,5% of inprovement!

 

OK, I just changed my old Ryzen 5 1600 to the 3600, the results with no OC:


And with the next level of OC called Precision boost overdrive:

 

Interesting, my score actually went up with 2 trucks on my 1600 :/