1. Do you need support for Assetto Corsa Competizione? Please use the proper forum below and ALWAYS zip and attach the WHOLE "Logs" folder in your c:\users\*youruser*\AppData\Local\AC2\Saved. The "AppData" folder is hidden by default, check "Hidden items" in your Windows view properties. If you report a crash, ALWAYS zip and attach the WHOLE "Crashes" folder in the same directory. Do not post "I have the same issue" in an existing thread with a game crash, always open your own thread. Do not PM developers and staff members for personal troubleshooting and support.
  2. As part of our continuous maintenance and improvements to Assetto Corsa Competizione we will be releasing small updates on a regular basis during the esports season which might not go through the usual announcement process detailing the changes until a later version update where these changes will be listed retrospectively.
  3. If ACC doesn't start with an error or the executable is missing, please add your entire Steam directory to the exceptions in your antivirus software, run a Steam integrity check or reinstall the game altogether. Make sure you add the User/Documents/Assetto Corsa Competizione folder to your antivirus/Defender exceptions and exclude it from any file sharing app (GDrive, OneDrive or Dropbox)! The Corsair iCue software is also known to conflict with Input Device initialization, if the game does not start up and you have such devices, please try disabling the iCue software and try again. [file:unknown] [line: 95] secure crt: invalid error is a sign of antivirus interference, while [Pak chunk signing mismatch on chunk] indicates a corrupted installation that requires game file verification.
  4. When reporting an issue with saved games, please always zip and attach your entire User/Documents/Assetto Corsa Competizione/Savegame folder, along with the logs and the crash folder (when reporting related to a crash).

The ACC CPU Benchmark Thread

Discussion in 'ACC Hardware Discussions' started by Deatroy, Dec 7, 2018.

  1. Krassmus

    Krassmus Gamer

    Best bang for the buck:
    9600k or 10600k. I'd definitely recommend the 9600k though.
    Big problem with amd is that it looks good on the single core cinebenches but once you put load on 3 cores, the amd cpu's won't boost as high as they do on 1 core and you'll lose performance!

    ACC will gain fps up to about 5 cores. So a 6 core cpu without ht or a 4 core cpu with ht are the sweetspot, price vs performance wise.
    At Racedepartment there's a mega benchmark thread where someone tested this by disabling cores and ht step by step on his 9900k.

    Going from 4 cores to 6 cores gives a nice boost, but no massive boost. Enabling ht if you have more than 4 cores will actually be negative for your minimum fps!
    But with only 4 cores it will give you a decent boost, still.

    End of Story:
    Get a cpu with 6 or 8 cores and without hyperthreading. Or disable ht for acc if you get a cpu that has it.

    Absolute best bang for the buck currently is the i5 9600k with 6 cores. But only if you overclock it.
    In most comparisons/benchmarks you'll see the 9700k being a lot better than the 9600k.
    That's due to a big gap in stock clock speeds!

    There are some benchmarks with both at 5 GHz on all cores and then, depending on the game, the performance is identical.

    A game like the latest tomb raider or assassin's creed in dx12 will be able to make use of 8 threads or more but ac, rF2, r3e, pcars2, acc won't!
    At least not more than let's say 5% fps boost.
    For 50-80% higher costs!

    The amd ryzen are sadly still behind for games that only make use of 4-6 cores.
    Simracing titles need single thread performance and Intel still has the edge as soon as you put load on more than 3 cores (amd will throttle then).

    I just upgraded from a i7 2600k @ 4.4 GHz, 4c/8t to the new i5 10600K, overclocked to 4.9 GHz on all cores.
    I did some tests and the truth for simracing is pretty sad.

    My best buddy got his 9600k a year ago and it runs easily 5.1 GHz on all cores. My 10600k, even if I disable ht is only stable at 4.9 GHz within reasonable vcore range.

    He gets higher fps in acc...

    And I paid 70€ more for the z490 Mainboard (we both have the gigabyte gaming X) and 90€ more for the 10600k compared to the 9600k..

    My minimum fps are slightly better without ht. So 160€ basically wasted if I'd only do simracing.

    Which is quite a lot in relation when a 9600k setup here in Germany costs:
    125€ z390 gaming X mobo
    65€ 16gb 3200 CL16 ddr4
    195€ i5 9600kf

    But I wanted to be future proof and I also play assassin's creed. At least that game makes use of ht...
     
  2. Krassmus

    Krassmus Gamer

    Little addition:
    "overclocking" these modern i5's will work the following:
    Go into bios, put the turbo for all cores to 49 (for 4.9 GHz) and go into windows.
    Download and run Hwinfo64, "sensors only", keep an eye on the vcore, especially the maximum.
    And on "max core °c" to get the value of the currently hottest core inside the cpu.

    Do a cinebench multi run and check:
    - max vcore below 1.35v? Good
    - max core temp below 85°c? Good

    If vcore (or temp) are too high, set the vcore mode either to fixed or adaptive and put in 1.30v, then repeat the cinebench run.
    Lower the vcore until you get a blue screen or until temperatures are below 85°c. Or until the max vcore is below 1.35v, if Temps are already fine.

    Really... Overclocking these i5's is easier than tweaking acc graphics settings.

    Improvements:
    Stock 9600k runs:
    1 core = 4.6 GHz
    2 cores = 4.5 GHz
    4 cores = 4.4 GHz
    6 cores = 4.3 GHz.

    From my experience in acc this means 4.3 GHz average with some jumps to 4.5 GHz from time to time on one of the cores.
    This means for calculations:
    5x 4.3 + 1x 4.4 = 26

    10 minute easy peasy overclock:
    6x 4.9 = 29.4

    Thats 13.1% more performance. Or for the magic 90 fps in VR:
    If you would have 79.6 fps with stock clocks, you would have nice and fluid 90 fps with the easy overclock.

    If temperatures and vcore/stability are fine you can go even higher.

    My mentioned best buddy is running 5.1 GHz @ 1.28v and barely scratches 70°c.
    He has a mesh case, 2x 140mm be quiet silent wings 3 and a thermalright macho Rev. B.

    I run my 10600k at 4.9 GHz @ 1.344v and hit 83°c during cinebench multi.
    Hovering between 60-70 in acc though.
    I have bad airflow in my damped silent case, 2x 120mm silent wings and the thermalright le grand macho rt
     
  3. Stains

    Stains Simracer

    That was also done using a replay to do the CPU benchmarks and also been replied to by Stefano stating that it isn't a correct way to test a CPU using a replay in ACC because it simplifies the physics during replay!
     
    D1visor, Krassmus and sissydriver like this.
  4. Krassmus

    Krassmus Gamer

    I know. But then you open process explorer, go into the properties of acc-win64 or whatever the real game running exe is called, navigate to the "threads" tab and look how many threads there are with high % of load.
    Note: the % can only reach 100% divided by the cpu core/thread count.
    For my 2600k this was 8, so 12.5% maximum.

    When you set up acc to be cpu limited (low resolution etc like in this thread), you'll see 2-3 big threads.
    Mostly one maxing out close to 12.5%, one at about 11% and a third one, probably running sounds or background stuff at 5-9%.

    This doesn't change when you run a replay or when you actually do a race with ai.

    I have to say I also barely see my fps changing between replay, online race or ai race.
    As someone already found out earlier in here: the biggest cpu hit is the rendering of opponent cars.

    Number of visible cars and mirror settings have a massive impact on cpu performance.
    Whether these cars are live calculated, physics wise, or just a replay barely seems to matter.

    So I would say:
    Yes, I 100% think Stefano said the truth, ai cpu load is awesomely spread across multiple cpu cores.
    But it barely makes a difference due to the rendering still being the big performance hog.

    Here are Screenshots and the video to "prove it":
    However my testing was a bit quick and dirty... Normally I should do a full formation lap and measure avg fps...

    Sadly I screwed up and kept "temporal upsampling" on... Which seems to already produce a graphics card limit.
    I'll run the replay again without it and provide the additional numbers, sorry!



    For the screenshots I used the exact frame where I shift into first gear. Look at the process explorer thread loads and the fps.
    My 10600k now has 12 cpu threads so 100% / 12 = 8.33% maximum single thread limit!
    The numbers from the screenshots are:

    AI live race:
    Threads: 8.2% / 8.05% / 4.92% / 4.28%
    CPU limited by the first 2 threads
    FPS: 91.5 fps at 76% load on the GTX 1070

    Replay of the same race: (sadly with upsampling so slight GPU limitation...)
    Threads: 8.06% / 4.53% / 3.6%
    CPU limited by the first single thread
    FPS: 88.5 fps at 92% load on the GTX 1070

    Replay without temporal upsampling to make sure it really wasn't a graphics card limit:
    Threads: 8.00% / 4.38% / 3.39%
    CPU limited by the first single thread
    FPS: 96.8 fps at 78% load on the GTX 1070

    Conclusion: AI vs Replay means one additional CPU thread is used, the fps are still limited by the first big thread though.

    Screenshots:
    AI live:
    AI_Live.jpg

    Replay Upsampling:
    Replay.jpg

    Replay without Upsampling:
    Replay_noUpsampling.jpg
     
    Stains likes this.
  5. Guidofoc

    Guidofoc Alien

    Thanks for the details. I eventually opted for a Ryzen R7 3700X (8 cores/8 threads) as it was available and at a reasonable price. I understand from your posts that the Intel CPU per-thread performance is higher than the AMD CPU. As I am just running ACC on a single 21:9 2560x1080 screen I suppose that even an R7 should be quick enough. I will report back once the other parts arrive and I can assemble my PC. I will wait before getting a new graphics card so my performance will be limited by the Radeon RX580 more than anything else.
     
    Krassmus likes this.
  6. Krassmus

    Krassmus Gamer

    So guys I had some spare time and thought I'd knock myself out. I couldn't be bothered to adjust all the settings but I deactivated temporal upsampling and the GPU load was below 80%.

    Procedure:
    - quickrace with AI
    - 720p with 50% resolution scale
    - my normal race settings. Shouldn't matter for the scaling stuff, process explorer/cpu load always looked identical but fps varied ofc.
    - Monza, dry, 13:00, 28 cars, starting 20th
    - Benchmarking from going into the car until the car in front lifts the brake. About 10 seconds duration

    - you can throw away the 0.1% and 1% low due to the short duration. However the min fps in afterburner/rtss seem to be something like 10% low?

    - did 3 runs for each configuration

    I uploaded a video with the results at the beginning and end. My graphic settings, race settings and the bench start/end points are shown.

    Before I show the results I'd like to re-post the multithread usage of ACC.

    AI racing: 2+2 Threads, 8.2% / 8.05% / 4.92% / 4.28%

    Replay of AI racing: 1+2 Threads, 8.06% / 4.53% / 3.6%

    Online racing: 1+2+1 Threads, 8.28% / 7.83% / 6.39% / 4.31%

    In my opinion online racing and AI racing will show very similar scaling results with different amounts of real or virtual cores.

    Finally, the results:
    ACC_CoreScaling+OC+MemOC_10600k.jpg

    And a little graph vor overview, although you need to keep in mind that it really stretches the differences in fps. Look at the true numbers!
    Attention: all memory scaling is done with 6 cores + HT + setting affinity to only real cores!
    ACC_Benching_Graph.jpg

    Conclusion:
    - you can take the results from @Daniel Costa with his 9900k and adjust the table by 1 core, since live racing (ai or online) adds one big thread
    - ACC gains a lot going from 4 to 5 cores and a little bit going to 6 cores
    - ACC loses average fps with HT. However the min fps seem to be very consistent or better (looking at 1% low)
    - with 4 cores, HT gives a bigger boost to min fps than it gives to avg fps. As experienced by many users
    - at 5+ cores, putting the game to only the real cores gains back the avg fps

    - 2600 to 3200 MHz memory increases fps quite a lot, at least when you have enough cores. Sorry, I couldn't be bothered to do the bench again with 4 cores...
    - Thinking in percentages, with 4c + HT my CPU should only have 81.9 fps avg with 2600 MHz memory.
    -> about 9 fps could be gained, in theory! I'd say in reality the difference will only be 5 fps

    And the video:
     
    grandepunto, Ole77 and Guidofoc like this.
  7. TXC_IJOJOI

    TXC_IJOJOI Racer

    I will upgrade from a 4770K @ 4,6 Ghz to a Ryzen 3600 shortly.
    For anyone thinking about upgrading, I will post the results in here :)
     
    Krassmus likes this.
  8. hozes

    hozes Rookie

    I upgrade to i7 3770s ti i5 9400f
    Its woth it totally, before update, i can race with max 15 ai with 8 visible, mirrors off and locked to 62fps, gou usage max 75, if i increase one of it, i get dips
    After update, with 25ai, 15 visible and mirros on high, 120 meters, i can easily play 62fps, get gpu bottleneck, gpu hits 99 percent if i increase resolution scale (1080p with %120 resolution scale)
     
    Krassmus and TXC_IJOJOI like this.
  9. hozes

    hozes Rookie

    Sorry, upgrade to i5 9400f from i7 3770s
     
    Krassmus likes this.
  10. Stains

    Stains Simracer

    What settings in game and what gpu?
     
  11. hozes

    hozes Rookie

    I choose epic, then only lower shadows to high ( shadows distance on epic) and visible opponent to 15
    At 1080p ( resolution scale on 100) gpu usage max 84 on nurb and spa i test, locked 62 fps
    Gpu is msi vega56 @1650mhz core: 925mhz memory
     
  12. TXC_IJOJOI

    TXC_IJOJOI Racer

    So here are the numbers:
    4770K @ 4,7 Ghz, 16GB DDR3 @1600Mhz, 1080Ti @ 2000Mhz 720p
    min: 56, max: 109, avg: 98

    Ryzen 3600 @Stock, 16 GB DDR4@3200Mhz, 1080Ti@ 2000Mhz 720p
    min: 79, max: 147, avg: 129

    So a BIG improvement, especially the min FPS are way better.
     
    Krassmus likes this.
  13. Kris.May

    Kris.May Gamer

    Any suggestion ryzen cpu to pair with my 1070ti?

    Ryzen 3 3300X
    Ryzen 5 3500
    Ryzen 3 2600
     
  14. Krassmus

    Krassmus Gamer

    2600 is "old" and has bad single thread speed. Not great for simracing.
    The 3500 and 3300x a difficult thing. The 3300x has better single thread performance but the 3500 has 6 cores and no smt that could cause issues if you don't know how to solve them.

    To be honest if you can somehow afford it, get a 3600. It has 6 cores and higher boost. Better than all 3 listed cpu's.
    If you feel like smt might cause issues for you, disable it and you still have the sweetspot of 6 cores (look at my tests for acc above in this thread).
     
    TXC_IJOJOI likes this.
  15. Matt_ZRS

    Matt_ZRS Racer

    Not done a proper bench mark but I can recommend my setup.
    3700x
    RTX 2080 super
    16gb 3600 Ram

    Runs in 4k most setting epic, maximum possible AI and all visible roughly 70-90fps when I last checked. I do lock with V-sync 60hz and its smooth.
    CPU total usage is 50% max when spikes but normally a lot less.
    Runs between 4400mhz-4290mhz at all times on all cores with stock cooler and default setup 65 deg C max or less.
    GPU is overclocked but also never goes above 65c ever.
     
    Last edited: Jul 3, 2020
  16. Kris.May

    Kris.May Gamer

    Ryzen 5 2600 Stock
    GTX 1070 ti (Undervolt +500Mhz Memory)
    16Gb DDR4 2800

    2020-07-05 02:43:49 - AC2-Win64-Shipping
    Frames: 5311 - Time: 60000ms - Avg: 88.517 - Min: 81 - Max: 95

    2020-07-05 02:46:01 - AC2-Win64-Shipping
    Frames: 5303 - Time: 60000ms - Avg: 88.383 - Min: 79 - Max: 96
     
    Last edited: Jul 4, 2020
  17. giustino

    giustino Gamer

    Hi to all,
    I don't know if this is the correct thread but I have a doubt. Is it possible that my I5 9600k CPU work at 90% with 26 cars and sometimes goes at 100% with big stuttering? obviously offline. If I reduce the cars from 26 to 23 the CPU never reach the 100% and it seems go well (about 90fps with medium video setting with RTX2060 with resolution of 3840x1080).
     
  18. Stains

    Stains Simracer

    90fps at 4k on a 2060! Can you please share your exact settings?
     
  19. Thomas Fun

    Thomas Fun Simracer

    It is not 4k, it is 1080p superwide 32:9, only half the resolution of uhd (4k).
     
    Stains likes this.
  20. Stains

    Stains Simracer

    Oops sorry, I just saw 3840 and thought 4k :oops:
     
Similar Threads
Forum Title Date
ACC Hardware Discussions Can I get more performance for ACC out of my CPU and RAM? Dec 11, 2022
ACC Troubleshooting ACC somehow CPU Spike background task dependant? Nov 22, 2022
ACC Virtual Reality What is the best cpu/gpu combo to run ACC at max settings Sep 2, 2021
ACC Troubleshooting ACC heavy CPU usage after update 1.5 Jul 19, 2020
ACC Hardware Discussions 2 x Radeon RX580 (Crossfire): does it work in ACC? What CPU goes well with them? Mar 31, 2020
ACC Hardware Discussions Test: CPU Core count and RAM scaling in ACC, AC1 and R3E Aug 20, 2019
ACC Hardware Discussions ACC - 0.5.2 - Performance Review by GameGPU (29 GPUS & 27 CPUS) Jan 22, 2019
ACC Troubleshooting gamebreaking stuttering after massive ACC crash CPU Oct 14, 2018
ACC Troubleshooting I have CPU Thermal Throttling and BSOD in ACC Oct 12, 2018
Chit Chat Room AC/ACC & CPU's / Threads & Cores Sep 22, 2018
ACC Graphics - Tracks and Cars ACC - CPU Bottlenecking Sep 22, 2018
ACC PS4/PS5/XB1/XBX/S General Discussions ACC Will it get cross party not just cross play on consoles? Saturday at 3:02 PM
ACC Physics Does ACC not implement real life braking mechanics?(Need some insights on this recent youtube video) Mar 16, 2024
ACC Multiplayer Communicating with accServer.exe (previously acRemoteServerUDP) Mar 14, 2024
Multiplayer Bugs & Issues Time of Day adjustment on AC Server Maker stuck at 8:00-18:00 even though I have full version ACCM Mar 2, 2024

Share This Page

  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.
    Dismiss Notice