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.
    Dismiss Notice

ACC - 0.5.2 - Performance Review by GameGPU (29 GPUS & 27 CPUS)

Discussion in 'ACC General Discussions' started by Daniel Costa, Jan 22, 2019.

Tags:
  1. Daniel Costa

    Daniel Costa Racer


  2. Similar Threads
    Forum Title Date
    ACC General Discussions About Matchmaking feature in ACC Monday at 2:32 PM
    News and Announcements About Matchmaking feature in ACC Monday at 12:42 PM
    ACC General Discussions Real Life Vehicles have ACC ..... Sunday at 3:08 AM
    ACC General Discussions ACC v1 Launch Party May 14, 2019
    ACC Multiplayer Minimum requirements for the ACC dedicated server May 14, 2019
    ACC General Discussions ACC vector based Logo? May 13, 2019
    ACC General Discussions Future ACC development and DLC May 10, 2019
    ACC General Discussions ACC SAMSUNG WATCH S3 May 9, 2019
    ACC Troubleshooting ACC 07.2. crashes with "HUNG" ERROR May 9, 2019
    ACC General Discussions ACC logo fanmade OC, what do you think? May 7, 2019
    ACC Physics Who has changed their mind about ACC FFB? May 5, 2019
    ACC Physics Does ACC support ABS feedback through wheel? May 5, 2019
    ACC Troubleshooting Cant access PitStop Menu May 4, 2019
    ACC Multiplayer accweb - Assetto Corsa Competizione Server Management Tool via Web Interface May 3, 2019
    ACC Controls & Peripherals Fanatec DD1 for ACC? May 3, 2019

  3. Whitestar

    Whitestar Hardcore Simmer

    Somehow, in the top diagram, 81fps looks much much better than 80fps. :D
    Wonder if they put in the wrong number.
     
    LeDude83 likes this.
  4. LeDude83

    LeDude83 Alien

    I don't speak any Russian but it seems that the GPU benchmark was run with AI opponents. Can someone confirm this?

    If so: this will cause the CPU to bottleneck - not due to the physics calculation but the rendering of the cars, as shown here:

    https://www.assettocorsa.net/forum/...pu-benchmark-thread.53720/page-3#post-1013074

    If my assumption is true, the test wasn't a GPU test but more of a system test. The ACC GPU Benchmark (hotlap mode) achieves a lot better results and a wider spread between the GPUs. Here's the current table. The fairly big differences between some test runs of the same models are most likely down to NVIDIA/Radeon driver versions and settings, as the "performance" option (wherever available) gives a significant FPS boost.
    But it's quite clear that any RTX model released today can achieve 100 FPS with max details in Full HD in ACC.

    [​IMG]

    IMO the performance is still not good even with those figures.
     
  5. Daniel Costa

    Daniel Costa Racer

    You have to benchmark what you are trying to achieve or what matters to you in the game. I also dont agree much with benchmarking with AI's but this use case is and will be the most used function of the game, so it has relevance.

    The GPU results are still valid because physics are rendered in separate cores to the rendering one. And the added scene complexity helps to put more load in the GPU's and graphics tasks of the CPU. The finish off, the cpu was the fastest you can get and OC to 5.2ghz, which helps mitigate cpu bottleneck, which as you can see there is none with the 2080ti when tested at 1440P and 4K. These are almost 100% on the GPU and very little on the CPU.
     
    Raklödder and ledinguelevrai like this.
  6. yeah, I think it is more 90fps if we compare the gap
     
  7. LeDude83

    LeDude83 Alien

    I totally disagree with you. If something is labelled a "GPU Test" then I need to isolate the GPU as much as I can. Anything else is misleading and plain wrong. What we're looking at here is a system benchmark. The fact that such a system benchmark is much closer to practical application (I totally agree with that) doesn't justify this misleading labelling of the results.
     
  8. Daniel Costa

    Daniel Costa Racer

    At 1440p and 4K, you have GPU bottleneck over CPU bottleneck, proved by the scalability you see between all the cards. You can disagree all you want. Thats how things work.

    Like you saw me saying many times, I dont like having CPU intensive tasks when benching GPUs either (draw distance, shadow quality, mirror quality, reflection quality etc).
    But the data above seems valid because dropping CPU intensive tasks would not improve increase or decrease the relative performance between cards at 1440p or 4K.
     
    Last edited: Jan 22, 2019
    Raklödder likes this.
  9. LeDude83

    LeDude83 Alien

    What's bottlenecking what can be seen in trustworthy usage graphs of the system components. Your conclusions from "scalability between cards" are not valid. They might be a somehow solid speculation at high resolutions but at Full HD they can't be trusted at all.

    You seem to struggle a lot with systematic approach to testing tasks and try to cover that up by referencing to your expertise as a HW reviewer (you said that in another thread) and by quoting smart sources.

    Sorry about that rather personal note but maybe you should stop claiming that you're "still right" when so clearly you're wrong. I don't even understand why you defend these results so hard - are you somehow involved with that site?

    And don't get me wrong: I'm not hating on the site and their tests, I appreciate it very much. All I say is that they never really ran a "GPU test".
     
    mr Switchblade likes this.
  10. Daniel Costa

    Daniel Costa Racer

    I edited my post above before you posted yours. Its really my final opinion on the matter.

    I never mentioned 1080P.
     
    Last edited: Jan 22, 2019
  11. MakeMeLaugh

    MakeMeLaugh Racer

    That just shows what we all already knew, game is bad optimized lol.
     
  12. Getting Coverage optimizing a game that is still being implemented new features every new update is totally out of the question.

    The optimization will only come at the end, when almost everything is implemented satisfactorily.

    Only at the end will we know how much it will improve.
     
    mr Switchblade likes this.
  13. Daniel Costa

    Daniel Costa Racer

    Its early-access which means the intent is to use the many possible hardware combinations from their userbase to test everything. If someone doesn't tell them it performs badly in some specific hardware, it may be too late to rectify things down the line.

    We don't know if kunos has 29 different GPUs and 27 different CPU's to test and see how their engine is behaving. They might be targeting a minimum spec based on two or three machines at the office.

    What we know for sure is that throwing compliments in their direction instead of testing things is not how the game will get better.
     
    alex bonner and MakeMeLaugh like this.
  14. If Kunos is able to provide me a current mid range PC I'll offer my "machine" for the minimum spec tests.
    Win -Win :D
     
    Last edited: Jan 22, 2019
    sissydriver and Raklödder like this.
  15. baakstaff

    baakstaff Gamer

    Gotta remember they're sorting by minimums, so its likely just the difference of 81 vs 80 fps avg plus 68 vs 57 fps min.


    That said, the gpu test shows a bottleneck somewhere else in the test system at 1080p; we're not seeing the same scaling from higher end gpus we'd see from other titles until the resolution is at least 1440p, and even then its not quite perfect.

    Comparing the GPU and CPU tests (specifically the 9900k), we see 81 avg/68 min GPU vs 77 avg/63 min CPU, which also seems to be the 9900k overclocked to 5.2 ghz vs stock.

    Since the 2080ti doesnt seem to be fully utilized at 1080p and overclocking the 9900k to 5.2 ghz only slightly improved performance, either its ram speed limiting the games performance (since there was no mention of a ram overclock), or the game itself.

    So no, its not a perfect gpu test unless you look at the higher resolutions, but its a good indicator for how well the game currently utilizes existing hardware. I'm fine with "it's in early access" being an excuse for the lackluster performance for now, but some improvements would be more than welcome (at least after they release a more complete multiplayer).
     
    LeDude83 likes this.
  16. Shaddix

    Shaddix Alien

  17. exited22

    exited22 Gamer

    Little bit shocked about the CPU comparison... My plan was to upgrade my i5 4690 non K in summer with a brand new Ryzen 3600 and some speedy 16Gb RAM... But seeing that the Intel 9000 series is able to produce about 20fps more is rather worrying for me, as I'm on a 1080 Ti (so enough GPU power) and VR :-/

    Seems like single thread performance is still very very important for ACC, probably because of a single render thread (again)...

    Hopefully Team Red has some good single thread performance gains in their pocket by switching to 7nm, as I really wanna support their good development!
     
  18. Daniel Costa

    Daniel Costa Racer

    With all due respect, i'm just speachless at the rationale of some people. That comment you just made is like saying water is wet unless you stay out of the rain.

    If you want GPU differentiators, look at 4K, which the review includes. But because it also posted 1080P results, it somehow becomes inaccurate?

    what? you think the 50 results from 50 different computers on a excel spreadsheet here on the forum is more accurate than that review? lets get serious here.
     
    mESCx and Raklödder like this.
  19. baakstaff

    baakstaff Gamer

    You're right, my bad. I was only talking about the 1080p results, didn't mean to completely discount the rest of the article.
     
    Daniel Costa likes this.
  20. elchshooter

    elchshooter Simracer

    To me performance is THE major ACC problem. I could drive on my oculus, but i see absolutely no way to ever get 90 fps min, which i managed to get most of the time in AC. i could drive on my 3440x1440 Monitor, but i see absolutely no chance of ever getting acceptable framerates on it. i even would have tripple 1080p/60fps Monitors to play, but you know...
    I really like the game, but there is no way i´d go back to a single 1080p monitor to get really stable framerates. It´s an absolute fun-killer for me, and the reason, why i really think of going back to iracing.
    I trust in kunos, but i don´t believe in +100% performance gain because of optimization...
     
  21. Daniel Costa

    Daniel Costa Racer

    i understand and feel the same. I have the odyssey and before it I had the rift, but most of the time I just want to get home sit down and play without worrying about this or that in VR. I now have a 55" 4K TV that does 1440p @ 120hz 5cm behind my wheel. It still is very immersive, with the added benefit of every game running at 120fps on it. ACC, AMS, RF2 are games that I am now playing more because of this change. But when I only had VR or a 27" monitor, I always chose VR with AC1 and R3E. Now I play everything, even Forza7 in HDR which is amazing.
     
    Winston likes this.

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