I've owned both AMD and Intel processors over time. The thing I always noticed about intel's hyperthreading is how well it runs in BeamNG. The way BeamNG handles processing is 1 car per available core, so if you have 4 cores you can typically run 4 cars well, once you add more cars the frame rates start to drop and even more so with AI enabled on all cars. With AMD (Ryzen 2700 8 core 16 thread) I noticed that with 8 cars loaded the framerate stays high, upon adding the 9th-10th cars the framerates where high (60fps Vsync on), adding the 11-14th car the framerates went medium (45fps Vsync on) but it started to dip drastically when adding the 15th and 16th cars the framerates plummeted to sub 25fps Vsync on. This makes sense because BeamNG processes 1 car per core, not per logical core. But.. this is where intel comes in. Right now I have the intel i3 10100 4 core 8 thread CPU (lowest of the 10 series line) I fully expected to run 4 cars at high frames - 5 cars at high frames - 6 cores at lower frames and 7-8 cars at sub 20fps keeping in line with the results of the Ryzen 2700. But.. upon loading 8 cars the frame rate was still high (60fps Vsync on),.. upon loading the 9th car framerate was high (60fps Vsync on), and when I turned on all AI the framerate hovered around 35-45fps Vsync on. When adding 10+ cars only then did the framerates drop to sub 20fps Vsync on. It looks like Intel with its hyperthreading can run as many cars as logical cores without framerate drops. This confirms what I saw on my older 4790k (4 core 8 thread) - it too ran 8 cars with minimal framerate drop. It seems that intel's hyperthreading works well in BeamNG but AMD hyperthreading does not work as well. Has anyone else noticed similar results?
Maybe your i3 is better able to maintain clock speeds with the increasing load compared to the Ryzen counterpart? I don't think the CCX layout and greater latency core to core on AMD would impact Beam too badly as I would imagine most of the threads "do their own thing" and don't interact with each other a lot.
I think maintaining higher clock speed may have something to do with it. I thought of the CCX layout too but when I tested it in different configurations in BeamNG the CCX layout seem to have little effect when running BeamNG even when there was crosstalk through the infinity fabric. The 4790k being generations older showed similar behavior to the 10100, it looks like the way intel implements hyperthreading is the main factor for BeamNG vs Ryzen
If it was all down the hyperthreading wouldn't that cause FPS to be significantly reduced as soon as you passed the core count, so 8 cars? Do you mean BeamNG is one car per core, not 1 car per thread?
One of the Devs chimed in one of the foums threads and confirmed BeamNG runs 1 car per core, not 1 car per thread, as a thread. I think both AMD ands Intel handle the hyperthreading well so you can go past your core count without much issue. The scheduler can run 2 or more cars per core but its not ideal especially if a processor doesn't have hyperthreading at all. Core count over Thread count is the rule, and hyperthreading is better then none for BeamNG. It just seems that intel can hyperthread cars very efficiently making it possible to run as many cores as threads even though that's not entirely ideal operation, you still much rather have as many cores as cars that you wont to run if possible. For the AMD side though I think the 5000 series may change this since they drastically improved IPC and changed the CCX operation. I havent had a chance to test a 5000 series in BeamNG but I assume they have matched Intel with its hyperthreading performance in BeamNG now. AMD recently took the gaming crown from Intel im assuming they have matched or exceeded IPC operation. I changed the title since I haven't tested with either the 3000 or 5000 series Ryzens