Sorry to update an old thread but it was exactly my "issue", and I have done some tests, and it's not an issue, it's just the limit of the CPU, the famous CPU bottleneck ! I use i5 9600K, I have tried with an overclock at 5ghz, and gtx 1080TI, 1440p ultra, PostProcessing Ultra. Day, no rain, Sebring. 1) If i race alone, my gpu usage is at 100%, the CPU is at 30% 2) If i race with 9 AI, my gpu usage is at 100%, the CPU is at 30% 3) If i race with +9AI, i have tried 20, then 40, the GPU usage drops and the CPU usage is still at 30%. Even the overclock at 5ghz doesn't push the GPU usage, so my conclusion is rFactor2 is ver very CPU demanding with AI, I have only succeed to push the GPU at 100% at the start of the race with only 9 AI or maybe a little bit more cause i have directly tested with 19 AI after (but with 40 AI all visible, even with a GPU usage at 60-70%, I can have 80fps at start, I start last, and after some times with less AI around me the GPÜ usage can be pushed). So if you have a low GPU usage even if your MSIafterburn or anythng else monitoring software show that you have a low CPU usage, it's wrong in rFactor2, my tests have proven that the game is very very CPU demanding with a lot of AI. To make the test yourself, race alone, then race with 1,2,3, 4....X AI until your GPU is not maxed at 100%, and you will have your CPU limit. With the i5 9600K at 5ghz, I can't push the GPU at 100% (at the start of a race, last) with +9 AI. It means that we need probably a CPU at 8 ghz to push at 100% with 40 AI at the start of the race, so be patient it's not for tomorrow. In the night and with rain, of course the GPU is pushed. With all others games, Project cars, Forza, ACC, F1 2018, Ride 3, the GPU is pushed at 100%, so it's good to see that the rFactor2 tyre simulation is heavily coded and impact the CPU. I will test iRacing, Automobilista and Raceroom, the 3 others sim probably behind rF2 about the physics simulation.
When your i5 9600K is 6 core CPU and rF2 has dual core support how can you expect more CPU usage than 30% ? Pic below is I7 8700K running rF2 at 5GHz and as you can see CPU MAX usage is 15%. CPU#0 and CPU#2 are the physical core(s) whic rF2 is using and as you can see Max core loads are 64 and 65%. There is no CPU bottleneck in in rF2
you are 100% wrong. if the GPU is not maxed with AI, it means the CPU has reached the limit = CPU bottleneck, just read what i have written and try to test a race alone, with 1 AI, with 2 AI, until your GPU is not maxed at the start, you will find the CPU limit. I have found that with 0 AI, the GPU is at 100%, then with 14 AI, the GPU usage decreases, it is only pushed at 85-90% , the CPU bottleneck is already present, with more AI (29 or 39), the GPU is at 60-70% at the start of the race, the CPU bottleneck is pretty massive. Please, you need to inform yourself more about the CPU limit. The cpu limit (or CPU bottleneck) is when your GPU should be maxed but it's not. tips : to see a CPU bottleneck, you have to reduce the resolution and the graphics, and if the framerate don't increase = CPU Bottleneck. For racing games, try to put 50 AI, and lower your resolution and graphics, if your framerate is the same, there is a CPU bottleneck edit1: Raceroom is the same, with AI, there is a massive CPU bottleneck, but with AC, with the max AI, the GPU is pushed at 100%, I will test iRacing but i'm pretty sure there will be a CPU bottleneck. The new 3d engine now like Project cars 2, are more GPU demanding, so the GPU is pushed more, but you have the same CPU bottleneck, with AI, the GPU usage drops, and the framerate drops....
@Alkoly Yes of course I am wrong as usual. GPU loads are depending a lot which track is used and how good it is optimized. I made test at S397 Le Mans with S397 GTE cars. 60 AI, 22 visible, all game settings max, I7 8700K >5GHz, 1080Ti >2063 MHz, Memory clock 6003 MHz. Min GPU load 92%, Max 100%
maybe with 60 AI test 60 AI visible and start last and show me your FPS at the start of the race with the GPU load in game with msiafterburn for example and at 1440P, not at 4K (4K always pushes the GPU at max). And all my tests were on Sebring with all AI visible of course....you don't really know and understand what is a CPU bottleneck, don't you ?
SEBRING, 144hz, 1440p ultra, Postprocess ultra, ALL AI visible i5 9600K at 4.8ghz (I've forgotten to up to 5ghz), 1080ti, (2.00PM, timeprogression x60, roadscale x15) 1 AI, GPU should be maxed, I have forgotten to remove the 144fps limit by the Vsync... so 144fps here (143) 14 AI, GPU is still maxed but no more 144fps but close 30 AI, GPU load decreases, the CPU bottleneck is present 60 AI, GPU load is 66%, the CPU bottleneck is massive Now I think you understand that the GPU load depends of the CPU, and you can see what a "CPU limit" or bottleneck is.
You don`t have to explain to me that a CPU bottleneck is when a CPU is incapable of keeping up with other hardware, generally the graphics card, in a certain task. It is very important to understand that no two games will perform identically. Each game generally uses a unique engine and the way the engine is coded or optimized is not always consistent or good. You first need to know how many cores the game can use and how CPU demanding in general it is. rFactor 2 uses 2 cores (central processing units > "die or chip") and is also fairly CPU demanding (AI) and the single core performance should be the king when it comes to overall gaming performance in rF2 also.The problem is that no mater if I7 8700k is at 5 GHz or 3.8 GHz single core performance is equal (GPU load 95-100% and equal FPS) when the same in-game settings are used. However you still need to pair an appropriate GPU with your CPU and what I can see from your pics i5 CPU limit is 14 AI and my i7 CPU limit is about 22 AI. When it comes to a CPU bottleneck it will always be subjective to the individual and there isn't really a right or wrong answer and some answers are more wrong than others.
Answer to the thread and your post : there is a CPU bottleneck in rF2. The low GPU usage is a CPU bottleneck. Thanks to confirm I'm right and you were wrong. My pics show 14AI and 30AI, not 22, I'm pretty sure your i7 8700K doesn't perfom better than the i5 9600K in rFactor2 at the same clock.
i7 8700k vs i5 9600k. Single core performance is the same when the same multiplayer (GHz) is used. rFactor 2 has dual core support. i5 9600K has 6 core and 6 thread > 2 core > 2 thread. i7 8700k has 6 core and 12 thread > 2 core > 4 thread. I am not sure if your i5 can handle 22 AI with 100% GPU load and perhaps you should test it with 22 AI as I have done with i7 8700k "14 AI, GPU is still maxed but no more 144fps but close"
22AI , I start last, the GPU is maxed, so your CPU is like mine in rF2, and it's with the 4.6ghz default boost, I'm not at 5ghz.... all you say from the begining is false, think twice before reply now. Your I7 8700k doesn't work better in rFactor 2 than the i5 9600K, I have told you.....
22AI, Imola 2018, I start last, the GPU is maxed too, and 4.6ghz only for the CPU... I'm pretty sure, even if I bought the i9 9900K oc at 5ghz, 8 cores 16 threads, the result in rFactor 2 will be the same as with your i7 8700K or the i5 9600k at the same clock speed.... You didn't know, (now you know) but I have known it, that's why I have made the short term choice to use the i5 9600K, with my racing sims , there is no difference in rFactor2 with the i7 8700K/9700K or the i9 9900K at the same clock speed (or 1%)...Next year I can easily upgrade to the i9 9900K if the racing games are heavily multithreaded, like BFV...but for rFactor2 it's a waste..... Anyway it was a pleasure to discuss with you, maybe with the new multiplayer system, we will meet us on a track.
It's not the tire simulation that is so much the issue, more the general CPU overhead of DX11. There are some rather knowledgeable people over at RaceDepartment who have measured CPU thread usage in every sim racing game. The dude who did this runs in 1440p and has a CPU bottleneck in every sim with a GTX 1070, so it doesn't surprise me that you have a CPU bottleneck with GTX 1080 Ti. Regarding how to measure a CPU bottleneck, there are basically two ways. First is to look at GPU load, like you have done. I have yet to see where this is not a valid method (when making sure you have vsync and frame limiters disabled). The second is to look at CPU load. There are two caveats here. The first one is the obvious, i.e., the total CPU load number is useless, since most sims are only able to load about 3 cores. The second problem is that Windows core scheduler likes to divide the load randomly to different cores. Even if a game is single threaded, Windows can divide it like this: Insruction 1 -> Core 2, Instruction 2 -> Core 3, Instruction 3 -> Core 4 This means you can still be CPU limited even if task manager and Afterburner show less than 100% load on any one core. The only way to see if a single thread is limiting is to use a tool like Process Explorer, which actually shows thread usage before Windows gets to decide which core it goes to. This is explained in another long post here. It was really eye-opening to me how much most games are limited by single core performance.
8 pages for the 9th Gen intel, thank you, I start now edit: very intersting, it confirms my thoughs. We have : - AC = 1.5 threads - AMS = 1.5* threads (thanks Martin) - Rf2 = 2 threads - Acc = 2.5 threads - Pcars 2 = 3.5 threads (stuck at 60% cpu load though and the fps drop to 50 in some occasions. Average about 80, goes up to 130+. GPU lowered to be below 60% so this is CPU limited) - F1 2018 = 6+ threads (lowering graphic settings = 160+ fps)
@stonec What surprises me is that i5 9600k seem to have higher CPU load (~30%) than i7 8700k (~15%) when max game settings are used. I can also see that i5 9600k core(s) loads rF2 is using are also higher than i7 8700k. I know that no game engine can claim 100% one core but still it would be nice know which bus load, VRAM and ram usage i5 9600k has but it is mission impossible to ask when @Alkoly discuss the level he does with me. Btw. my test done with res. 3440x1440, 34" singlescreen.
If i can help to explore more the CPU usage, it will be a pleasure. At 3440x1440 the GPU is more pushed, we should to all the test at 1080p. It becomes intersting.
Hi, I'm new on this forming but I have +- the same issue. I have a 2080ti and a Ryzen 2700x and my GPU is also around ~ 50%. I'm using a Valve Index without supersampling and I fail to maintain a stable 90fps. Track: VIR, car: Aston GTE (new one), 10 opponents. What tool do you use to monitor GPU usage? I'm using VR so ctrl-c isn't working for me. (only framerate with ctrl-t) Thanks!
