The ACC Benchmark Thread

This is my attempt to generate somewhat reproducable benchmark results for ACC.

You'll need to download FRAPS - it's free, lean and easy to use. You can get it here, for example:

https://www.chip.de/downloads/FRAPS_13010723.html

Set up a hotkey for benchmarking and tick the checkbox at "Stop benchmark after 60 seconds".



IMPORTANT: Deactivate FPS cap and VSYNC for the benchmark run! It can potentially make your system look worse than it actually is! Adaptive sync like FreeSync and GSync can remain active.

Now, set up a quick race like this:



Here's a zoom-in on the left red box:



In "realism", set the formation lap to "full lap":



Start the race. As soon as you're in the cockpit, hit the benchmark hotkey. Keep a steady position until the benchmark stops (no weaving, bumping etc.). You will notice the end of the benchmark by the FPS counter reappearing in the corner of the screen. Quit the race.

Open this thread, post a picture of your in-game GFX settings and the view (camera) and FOV you used (this is important, don't miss it!). Add the FRAPS result from the file FRAPSLOG.txt.

Please write down your CPU, GPU and RAM and make sure your signature is up to date.

Attached to this post there's a result table of the old benchmark method, where a "last corner" formation followed by one race lap was recorded. A new table will be uploaded soon with results according to the new method explained above.
In the new table, there will be no more multiple entries per user but only one - the favourite setting of the user. Should the user change his settings and upload new test results, the table will be updated.

Here's a result "preview" of the table:



This thread shall be used for a friendly discussion and direct comparisons between users.

Looking forward to seeing your results

 

And I'll start right away:



2018-11-18 14:41:36 - AC2-Win64-Shipping
Frames: 9344 - Time: 189687ms - Avg: 49.260 - Min: 34 - Max: 68

 

It will be good to compare with original AC.

 

Personally I'd recommend using something like MSI Afterburner which also has a benchmark function built in and records 1% lows and 0.1% lows.

But overall test results seems highly variable, especially with manual car control and in the rain.

 

AB can potentially take control over your GPU...fans, clocks etc. A tool that can only monitor was the goal.
And variance is less than you might think. Just do 5 runs, always stay just behind the entire field and you'll see: AVG results vary by 1, 2 FPS or so.

In general: please let's not question the method here or this thread will derail. Post your results or move on. Thanks.

 

Not to my experience, but fine. If you think discussing a proper benchmarking methodology is derailing your thread I won't bother...

 

Proc: FX8350
GPU: GTX 1070
16GB ram

2018-11-18 15:19:44 - AC2-Win64-Shipping
Frames: 6430 - Time: 178000ms - Avg: 36.124 - Min: 27 - Max: 57

 

Thanks for joining

This time it was not relevant but next time, remove the FPS cap please ;-)

 

proc: i7-2600K
gpu: gtx 1060 16GB



2018-11-18 19:20:50 - AC2-Win64-Shipping
Frames: 7141 - Time: 173672ms - Avg: 41.118 - Min: 29 - Max: 61

 

Without changing anything you can have the same kind of results @Epistolarius talks about by simply reading the Fraps frametimes file ("ACC date frametimes.csv") with Frafs.
Best of both worlds and with a graph included.

 

Clarification.
I use Vsync (because I output to a TV). Do you prefer that off? Or is it ok with it on?
I also use EVGA Precision XOC and NZXT CAM software. This is to manage my fan profiles. Otherwise my system heats up PDQ and that can then cause some throttling. So I will run it still with those on, hopefully doesn't impact the overall result too much.

 

It doesn't matter what tools you use in order to set up your GPU. I use WattMan for example in order to keep my GPU's temps in check. You want a realistic result so you should run whatever you use when gaming.

But when you use VSYNC or FPS cap you obviously never get FPS higher than your screens refresh rate/FPS cap value and that distorts the results - maybe your system can do better than the capped/synced FPS?!?
@MakeMeLaugh used FPS cap at 60 - similar situation. During his test it wasn't important because 60 FPS weren't reached (57 max). But should he reduce his settings now and his system could potentially hit 70 FPS, the benchmark result will be worse than what his system can do.

Is this kind of clear? Bottom line: just for the benchmark run, deactivate VSYNC and FPS cap.

BTW, @AlleyViper, @haxxx, @liakjim, @Slowheit: please chime in with your results - let's see some AMD power

 

I'm not near my AM4, but will do when I can. Given that this is UE4, which usually does better with Intel+NV, I wouldn't be surprised if I end up with 2/3 fps with a 2600x+OC'd RX480 when compared to a GTX1060 and a quad i3...

 

CPU i7 - 7700
GPU Asus GTX 1060 3G PH
16 gb ram



2018-11-19 18:09:54 - AC2-Win64-Shipping
Frames: 7923 - Time: 152203ms - Avg: 52.055 - Min: 33 - Max: 76

 

Formation Lap:
2018-11-19 18:28:30 - AC2-Win64-Shipping
Frames: 7959 - Time: 204094ms - Avg: 38.997 - Min: 29 - Max: 58

First Lap:
2018-11-19 18:31:56 - AC2-Win64-Shipping
Frames: 6425 - Time: 152703ms - Avg: 42.075 - Min: 27 - Max: 65

 

i7-8700k
evga gtx1080ti ftw3
16gb ram

Frames: 13946 - Time:199203ms - Avg:70.009 - Min:50 - Max:101

 

Not sig PC:
PhII 965Be 4x @4,1GHz/NB@2.64 12GB Ram@1600C7 2T
HD7870 2GB @1125/1350
W10 1809 x64

2018-11-19 19:06:12 - AC2-Win64-Shipping
Frames: 7406 - Time: 155109ms - Avg: 47.747 - Min: 32 - Max: 60

other runs:
Frames: 7183 - Time: 155578ms - Avg: 46.170 - Min: 33 - Max: 57
Frames: 7568 - Time: 154375ms - Avg: 49.023 - Min: 33 - Max: 64

On a single race under more decent day/clear conditions, lowering AI to 12 helps a lot to keep 60fps. It goes without asking, looks beyond terrible on low settings.
UE4 seems to be handling low vram/low to mid ingame texture settings in a way that seems to force exceedingly low resolution mip-maps when track/opponent objects are very close.

 

@LeDude83 Btw, I just noticed that some results have >150s, while others took >170s. I assumed that "once the race starts" was the green flag, to remove any weird cpu spikes due to AI misbehaving in early formation, but looking at your bench (with over 20s) you meant starting the benchmark right when in cockpit view.

 

Rain at night - you could have at least given me a chance!

Anyhoo...



2018-11-19 22:27:39 - AC2-Win64-Shipping
Frames: 9787 - Time: 152985ms - Avg: 63.974 - Min: 36 - Max: 84

Intel i5 4670K @ 4.2
Nvidia GTX 970
8 GB RAM

 

Thanks for paying attention @AlleyViper.

Guys, thanks for joining this thread
To make it very clear: the benchmark needs to be started right after clicking "start race", not when the green flag drops! The formation lap needs to be set to "last corner". This setting is found in options --> realism IIRW.
And you shouldn't overtake the field - stay in last or 2nd to last position, all the time. A deviation from those rules will render the result worthless.

The benchmark will report a time of something between 180 - 190 seconds.

If you did this differently the first time around, please come back and run the test again.

And something out of personal interest: @Martin Kolibal and @MakeMeLaugh could you kindly rerun the test with my settings? I'd like to know if my GPU performs as expected or if there's an AMD or NVidia bias in ACC.

My settings are here:

https://www.assettocorsa.net/forum/index.php?threads/the-acc-benchmark-thread.53257/#post-1009256

Thank you!