Is Assetto Corsa CPU heavy?
- Thread starter fatal4you
- Start date
didn't tryed that but I noticed the difference between my old I7 860 and my I9 9900K. Before I could race on Nordschleife with max 10 opponents with huge fps drop. Now it looses 15 fps but is still around 130 to 140 what I have seen with a full grid.
Maybe the ai overbrakes there a lot and the tyre smoke is killing your performance. that was also the case with my I7
Maybe the ai overbrakes there a lot and the tyre smoke is killing your performance. that was also the case with my I7
Heavy on the cpu regarding overall usage, temperatures and power consumption?
Nope, definitely not.
Do you need the fastest single thread cpu that you can get to have stable 90 fps in vr with a 30 car grid?
Yes, absolutely!
Many simracers get a 9900k or 10900k or lately the amd 5xxx CPUs.
Not because they need the amount of cores.
They need the higher clocks due to the binning process.
My 10600k can only run on 4.9 GHz with reasonable voltage.
A 10900k boosts to 5.3 GHz on one core at default!
Assetto corsa uses 2 big cpu threads, one not so heavy thread and then multiple small threads.
You can see that with process explorer when going into the properties of the ac exe.
One thread can only maximize one core at the time so if you have 8 cpu cores, one cpu thread has a maximum of:
100% / 8 = 12.5% cpu load.
This means when you have a 8 core cpu and Taskmanager shows:
2x 12.5% + some little threads = 30-40% overall cpu load
Your cpu is still running at the limit! Some core simply do nothing for short moments.
Taskmanager will show an average though.
If you go down to single clock cycles, you woukd see 2 cores running at full load for a short time, then the next 2 cores running at full load for a short time.
Windows spreads the load across all cores. It's more efficient since the other cores can start caching etc.
This sadly also means that you can not see the cpu limit.
You can however look at your graphics card load. When the graphics card load is 95% or higher, the graphics card is limiting.
If the graphics card load is below 95%, either the cpu is limiting or an fps limiter (fps limiters, vsync etc).
I'll post my assetto settings later today when I'm at the pc
Nope, definitely not.
Do you need the fastest single thread cpu that you can get to have stable 90 fps in vr with a 30 car grid?
Yes, absolutely!
Many simracers get a 9900k or 10900k or lately the amd 5xxx CPUs.
Not because they need the amount of cores.
They need the higher clocks due to the binning process.
My 10600k can only run on 4.9 GHz with reasonable voltage.
A 10900k boosts to 5.3 GHz on one core at default!
Assetto corsa uses 2 big cpu threads, one not so heavy thread and then multiple small threads.
You can see that with process explorer when going into the properties of the ac exe.
One thread can only maximize one core at the time so if you have 8 cpu cores, one cpu thread has a maximum of:
100% / 8 = 12.5% cpu load.
This means when you have a 8 core cpu and Taskmanager shows:
2x 12.5% + some little threads = 30-40% overall cpu load
Your cpu is still running at the limit! Some core simply do nothing for short moments.
Taskmanager will show an average though.
If you go down to single clock cycles, you woukd see 2 cores running at full load for a short time, then the next 2 cores running at full load for a short time.
Windows spreads the load across all cores. It's more efficient since the other cores can start caching etc.
This sadly also means that you can not see the cpu limit.
You can however look at your graphics card load. When the graphics card load is 95% or higher, the graphics card is limiting.
If the graphics card load is below 95%, either the cpu is limiting or an fps limiter (fps limiters, vsync etc).
I'll post my assetto settings later today when I'm at the pc
Heavy on the cpu regarding overall usage, temperatures and power consumption?
Nope, definitely not.
Do you need the fastest single thread cpu that you can get to have stable 90 fps in vr with a 30 car grid?
Yes, absolutely!
Many simracers get a 9900k or 10900k or lately the amd 5xxx CPUs.
Not because they need the amount of cores.
They need the higher clocks due to the binning process.
My 10600k can only run on 4.9 GHz with reasonable voltage.
A 10900k boosts to 5.3 GHz on one core at default!
Assetto corsa uses 2 big cpu threads, one not so heavy thread and then multiple small threads.
You can see that with process explorer when going into the properties of the ac exe.
One thread can only maximize one core at the time so if you have 8 cpu cores, one cpu thread has a maximum of:
100% / 8 = 12.5% cpu load.
This means when you have a 8 core cpu and Taskmanager shows:
2x 12.5% + some little threads = 30-40% overall cpu load
Your cpu is still running at the limit! Some core simply do nothing for short moments.
Taskmanager will show an average though.
If you go down to single clock cycles, you woukd see 2 cores running at full load for a short time, then the next 2 cores running at full load for a short time.
Windows spreads the load across all cores. It's more efficient since the other cores can start caching etc.
This sadly also means that you can not see the cpu limit.
You can however look at your graphics card load. When the graphics card load is 95% or higher, the graphics card is limiting.
If the graphics card load is below 95%, either the cpu is limiting or an fps limiter (fps limiters, vsync etc).
I'll post my assetto settings later today when I'm at the pc
RasmusP
, I should have stated that I'm using a i7 920 from 2009/2010. I was using an i7 2600k, overclocked to 4.4 GHz from 2011-2020.RasmusP
I could run 99% of all games completely fine. Witcher 3, need for speed, dragon age, assassins creed etc etc.
The 3 games I struggled to have 60 fps:
- Gothic 3 (it was patched to only use 1 core due to crashes when using more. It's from 2006 and my fps would drop to 20-30 in cities...)
- assetto corsa with more than 20 cars during race starts
- rF2 when driving past the pits with more than 10 drivers on the server
Now with my 10600k, I upped my fps limit to 90 fps and I still get some drops here and there into the 80's. But only when racing with 30 cars on the server.
The single core performance on Intels basically stagnated from the 6000 series to the 10000 series. Just higher clockspeeds and some optimizations.
Now Ryzen 5000 put 20% on top of that! These CPUs are shredding through the simracing titles with easy
I'm hoping that real benchmarks of the 11600k will show 15-20% single core boost over my 10600k.
I can just put that cpu into my z490 motherboard and put my cooler back on.
Anyway, some random cpu talk hehe.
I'll post my optimized AC+CSP settings
So my settings are:
Note: I'm on 0.1.71. It's stable for me, looks great. 0.1.72 and 73 have a new interiors shader (Reflections FX) and I dislike how shadows are drawn on the sunscreen part of windscreens.
However this new Reflections FX version has some reflection smoothing which allows to use only "1-face-per-frame" in the Assetto graphics settings. So it's slightly better on the CPU! (I'm currently using 1 face per frame anyway, but the reflections are stuttering a little bit...
AC settings:
- CPU performance: World details, reflection frequency and distance. Shadows only matter if you disable them completely in CSP. The resolutions are almost purely on the graphics card, not the cpu!
In general, CPU performance is mostly influenced by the amount of details, not the quality of these details. So more reflections = lower fps, more world details = lower fps, more elements from your HUD apps = lower fps.
But resolution of the shadows or reflections? Doesn't really matter.. These matter for the graphics card though!
CSP settings: Here are the active extensions. What I don't show is simply the default!
I'm using 2x MSAA since it costs barely anything and helps with aliasing on rough geometries. Then I disabled post-process-anti-aliasing like fxaa, smaa, mlaa etc and instead activated the super awesome TAA!
It's miles ahead of the TAA from other games imo! Since I'm using this, AC looks clear and smooth with barely any pixel crawling or aliasing-shimmering
The most important CPU related settings:
- CPU optimizations
- Custom Fonts Rendering!!!! Check your fps when you disable ALL HUD apps! They will skyrocket. The moment you enable your HUD apps again, the fps will go down a LOT.
Custom Fonts Rendering will bring you a lot of these lost fps back.
Note: I'm on 0.1.71. It's stable for me, looks great. 0.1.72 and 73 have a new interiors shader (Reflections FX) and I dislike how shadows are drawn on the sunscreen part of windscreens.
However this new Reflections FX version has some reflection smoothing which allows to use only "1-face-per-frame" in the Assetto graphics settings. So it's slightly better on the CPU! (I'm currently using 1 face per frame anyway, but the reflections are stuttering a little bit...
AC settings:
- CPU performance: World details, reflection frequency and distance. Shadows only matter if you disable them completely in CSP. The resolutions are almost purely on the graphics card, not the cpu!
In general, CPU performance is mostly influenced by the amount of details, not the quality of these details. So more reflections = lower fps, more world details = lower fps, more elements from your HUD apps = lower fps.
But resolution of the shadows or reflections? Doesn't really matter.. These matter for the graphics card though!
CSP settings: Here are the active extensions. What I don't show is simply the default!
I'm using 2x MSAA since it costs barely anything and helps with aliasing on rough geometries. Then I disabled post-process-anti-aliasing like fxaa, smaa, mlaa etc and instead activated the super awesome TAA!
It's miles ahead of the TAA from other games imo! Since I'm using this, AC looks clear and smooth with barely any pixel crawling or aliasing-shimmering
The most important CPU related settings:
- CPU optimizations
- Custom Fonts Rendering!!!! Check your fps when you disable ALL HUD apps! They will skyrocket. The moment you enable your HUD apps again, the fps will go down a LOT.
Custom Fonts Rendering will bring you a lot of these lost fps back.
Good advices! I’m also a user of MSAA with TAA, can’t beat this combo. In my setup, enabling PP AA is just wasted resources for a blurry image.So my settings are:
Note: I'm on 0.1.71. It's stable for me, looks great. 0.1.72 and 73 have a new interiors shader (Reflections FX) and I dislike how shadows are drawn on the sunscreen part of windscreens.
However this new Reflections FX version has some reflection smoothing which allows to use only "1-face-per-frame" in the Assetto graphics settings. So it's slightly better on the CPU! (I'm currently using 1 face per frame anyway, but the reflections are stuttering a little bit...
AC settings:
- CPU performance: World details, reflection frequency and distance. Shadows only matter if you disable them completely in CSP. The resolutions are almost purely on the graphics card, not the cpu!
In general, CPU performance is mostly influenced by the amount of details, not the quality of these details. So more reflections = lower fps, more world details = lower fps, more elements from your HUD apps = lower fps.
But resolution of the shadows or reflections? Doesn't really matter.. These matter for the graphics card though!
View attachment 455426
CSP settings: Here are the active extensions. What I don't show is simply the default!
I'm using 2x MSAA since it costs barely anything and helps with aliasing on rough geometries. Then I disabled post-process-anti-aliasing like fxaa, smaa, mlaa etc and instead activated the super awesome TAA!
It's miles ahead of the TAA from other games imo! Since I'm using this, AC looks clear and smooth with barely any pixel crawling or aliasing-shimmering
The most important CPU related settings:
- CPU optimizations
- Custom Fonts Rendering!!!! Check your fps when you disable ALL HUD apps! They will skyrocket. The moment you enable your HUD apps again, the fps will go down a LOT.
Custom Fonts Rendering will bring you a lot of these lost fps back.
View attachment 455429
View attachment 455427
View attachment 455428
View attachment 455430
so it is almost the same cpu like the one I had a bit stronger then mine was I guess.RasmusP
the only option you have next to upgrading, is overclocking but if you make it wrong you have to make an upgrade anyway..
I had on my I7 an Noctua Fan something with NH12 or 14. Later I run that cpu with 3900MHZ instead of 2800 MHZ which was stock.
Maybe your cpu has also so much potencial for overclocking, but if you only using an boxed fan it will run to hot in stock anyway.
Had huge improvment but after 1,5 years it started to get unstable, then I replaced my board, CPU and Ram.
Last edited:
so it is almost the same cpu like the one I had a bit stronger then mine was I guess.
the only option you have next to upgrading, is overclocking but if you make it wrong you have to make an upgrade anyway..
I had on my I7 an Noctua Fan something with NH12 or 14. Later I run that cpu with 3900MHZ instead of 2800 MHZ which was stock.
Maybe your cpu has also so much potencial for overclocking, but if you only using an boxed fan it will run to hot in stock anyway.
Had huge improvment but after 1,5 years it started to get unstable, then I replaced my board, CPU and Ram.
True. When I start overclocking I would have to replace the stock fan, water cooling maybe.
See, what I don't understand, sim racing is basically a physics simulator, a math engine. The most math in parallel can be done on the GPU as it has the most math cores and not on the CPU. Kind of surprises me how CPU becomes the bottleneck. Sure it would be more effort for the programmers to support GPU computing for NVIDIA and AMD each, but it would be the perfect use case.
maybe it depends on you GPU aswell. I have an shitty GTX 970, which has an huge problem if you using all off the 4 GB GPU Ram. max 3,5GB is ok if you go over the card gets even slower then it is.
The graphix gets more shiny in this titles and need much more power, sometimes I wish they would stay with less shiny graphics and focus more on proper physics.
water cooling isn't nesessary, it is better of course but if you want an good water cooling system you are easy at 500- 600€ only for that. if you runing an boxed fan you should monitor your temperatures from the cpu. my was cooking even on half load. with that big cooling block and fan I was down around 30°C on my CPU. If you running 80 or more it willstart to reduce power.
but an huge cooling block doesn't help much if you have an crammed tower you need to get the hot air out and this will bring you much more stable fps then many people expect
Last edited:
The problem is that you can't run the calculation chain in parallel.
You always have the inputs on the tyres and then the chain reaction.
It might be possible to optimize some parts of it but you'd need a big team of highly skilled people to do this.
Project cars is running one Tyre per cpu thread for example.
But ac, acc, raceroom, automobilista rfactor 2...
All barely gain any fps when you have more than 4 cores.
Acc gains quite a bit when having 5 cores (I tested this by disabling one core after another in the bios on my 6 core cpu).
But that's it.
I habe no idea about the physics code but apparently it's not really possible and needs to be done on the cpu, not the gpu.
Reading this makes me sad that this feature doesn't work on triples screen setups!instead activated the super awesome TAA!
It's miles ahead of the TAA from other games imo! Since I'm using this, AC looks clear and smooth with barely any pixel crawling or aliasing-shimmering
StefanoCasillo
Jaxx Vane Studio
A couple of obvious reasons:
1) GPU to CPU (or the other way) communication is slow and very unpredictable. So if you decide to pay the cost for that communication you'd better use it for really MASSIVE amount of parallel work. Ie.. a stack of 500 cubes.
2) There are problems that can be easily solved in parallel (ie.. a LOT of the same calculations with the same input and independent output). Graphics is a perfect example.. you could run 1 thread per pixel... every thread reads from the same inputs and write to a single pixel (thus independent) as output.
As soon you don't have this scenario, things get complex very quickly. In a racing game you really don't have "many" of the same things. Cars only have 4 tyres, 1 engine, 1 gearbox, 1 differential, 4 suspensions and so on.
AC does multi thread at the single car level.. every car is solved in its own thread. This can easily hit the required perf target for single player on a decent machine. Running car physics on the GPU considering the amount of data that has to come back to the CPU to make the rest of the game work makes no sense at all unless you start talking about hundreds of cars.
3) GPU is good at performing linear calculations.. they are not as good as CPU at general computing.. branches, if statements, and stuff like that.
Last edited:
Just because Stefano knows how things work, benchmarks don't lie.
5+ cores and maximum single thread performance you can achieve.
(4 cores perform significantly worse and 4c + virtual splitting (ht, smt) still do too)
So Intel 8600k/9600k/10600k (and maybe 11600k, that gen is a bit different) are what you want. Overclock them to the maximum (basically all to 4.9-5.0 GHz).
I7/i9 will only give you a little bit more single thread performance but not worth the massive price increase.
And then there's the amd 5xxx gen, which blows everything else out of the water.
5600x is enough. And like with Intel, the more expensive variants will give you a bit more single thread performance due to more Cache and core clock, but absolutely not worth the price increase.
If you want to buy something new right now, 5600x. Period.
If you want to buy Intel, 10600k/11600k, depending on whether or not the games you want to play suffer from the internal latency of the 11th gen or not.
Elliot Faggetter
Premium
Anyone know why my CPU and FPS is high % but in Task Manager, its not even peaking? Seems AC isnt using my full CPU or GPU?
True that: AC runs only a few tasks, and many modern CPUs have many cores.
Task Manager shows average utilization.
AC may max out one core, but Task Manager might show < 20%.
If a CPU core is maxed e.g. for calculating physics, then GPU most likely cannot be.
That is one reason why overclocking tends to be more popular among gamers,
to get more performance from games that do not effectively exploit multiple cores.
An Intel i5 running a higher clock speed can outperform i7 or i9 with lower clock speed
or even the same clock speed if that i7 or i9 runs into power or thermal limits.
Last edited:
