The ACC CPU Benchmark Thread

Should do, not hugely slower than 9th or 10th gen. Maybe an OC in order.

Not seen any GPU scaling vids yet from Hardware Unboxed, usually good for working out what will pair well.

 

Schould SMT on Ryzen 3600x be ON or OFF?
Does Steam Game Launch Option (-USEALLAVAILABLECORES) work or does it make no sense?

 

I noticed no real difference between 6C/12T & 6C/6T on a 3600.
With AutoOC and PBO set to 115w/tdc75/edc90 with SMT off it was boosting up to 50mhz more on single cores and up to 25mhz on all core.

In game I failed to notice any tangible difference. That wont necessarily mean the 3600x will be the same though.

 

I tried a little different benchmark, maximize AI, low graphic settings (epic view distance + all cars visible)
Setting is Spa @ 23:00 with 49 AI, starting in position 50, including short formation and then following the train of AI for a full lap.

5900x with a GTX960:

AC2-Win64-Shipping.exe benchmark completed, 38992 frames rendered in 191.156 s
Average framerate : 203.9 FPS
Minimum framerate : 164.5 FPS
Maximum framerate : 221.6 FPS
1% low framerate : 158.2 FPS

Basically 200+ after green, before it is a little lower around 170-190.


@Krassmus It is January now

 

made a mistake , but cant edit the post anymore. these are the real numbers

 

Thanks, my reminder in the calendar was set to next Sunday

So here are 3 thread and 4 thread Cinebench R15 results (closest to simracing titles as R15 doesn't use avx and most simracing titles use 2-3 threads but as my benchmarks for ACC show, 5 cores are needed for the best performance):

(if the 10600k would just hit the "magic" 90 fps in VR, the 5600x without OC would have a headroom of 15-17 fps)

Here's the source of these numbers over at Racedepartment:
https://www.racedepartment.com/threads/amd-ryzen-for-simracing.162761/page-43#post-3319872


And to end my post:
I told you so

(prices are currently way too high due to the supply shortage, msrp of the 5600x = 299$, 3600x = 249$, 10600k = 262$)

 

i7-9700k, 16GB Ram 3200CL16, RTX 2080ti

2021-01-12 23:33:29 - AC2-Win64-Shipping
Frames: 8350 - Time: 60000ms - Avg: 139.167 - Min: 121 - Max: 167

 

i7-9700k, 32GB Ram 3600CL14, RTX 2080ti

Frames: 8844 - Time: 60000ms - Avg: 147.400 - Min: 128 - Max: 181

 

Amazing that is just the RAM CL speed doing it. And maybe the doubled size?

 

to be honest I was surprised as well - increase in CPU & GPU Tests! Actually the previous 16Gb 32ooMhz/CL16 Memory was only planned as temporary solution due to delivery issues when I build the system initially. And of course I did not seat the new modules correctly when installing so had a nice panic phase when the system did not boot following the memory installation....

 

I´ve changed my

32 GB CL 18 20 20 40 3600 Mhz to
16 GB CL 17 17 17 35 4200 Mhz.

I wonder if I´ll get more out of it when I downclock it to 3600 CL 14.

 

Frequency over latency. Most application gain more from raw memory speed and not from faster response.

The gain could also come from the updates to the game.
I bought a 3070 (became available a week before the 3060ti launch) and a 3060ti and compared them in acc.
Had more fps with the 3060ti without touching any settings!

But some optimizations were made and I think some new setting for the mirrors...

Put the 3070 back in and had more fps.

Maybe the cpu performance changed too?

 

game and driver was identical - tests performed within 24h with no update in between

 

Ah lol I didn't that you posted 2x... Sorry!
Was on the phone and the site jump-scrolled a bit while it was still loading.

So about 7 fps more now. At your high fps that's about 6%.
From my benchmark, the difference between 3200 cl16 and 3466 cl17 was 94,7 to 97,4 fps.
That's already 2,9%.

Good to see that our benchmarks add up to make sense!

At least for acc, higher memory frequency seems to make a difference!

Are you tempted to try to go higher than 3600 and putting the timings to auto? This should make it kinda stable when trying a higher clock...
Would be nice to see a test with 3800 or more!

 

Here you go, 4200 Mhz CL 17:

2021-01-14 22:03:47 - AC2-Win64-Shipping
Frames: 11190 - Time: 60000ms - Avg: 186.500 - Min: 148 - Max: 224

 

Addition:

It´s the benchmark on page one: 720p, Nürburgring 21 cars, low, low, low

I tried 3700 Mhz CL 14 14 14 30 yesterday (Couldn´t get 3800 Mhz to work quickly). Gives me the same latency as 4200 Mhz CL 17 in AIDA64 but about 6 GB less write/read speed. So I´ll stick with the 4200 Mhz then.

 

Would've been interesting to see the fps in acc

Btw what cpu and clockspeed do you have? Don't see you mentioning it?

 

Hi, I suppose you can´t see my signature via mobile?

9900KF 5 Ghz all cores. I was really suprised about the high numbers I admit. I don´t think CL 3700 CL 14 would give me more FPS because the latency is equal and the processed data per second is way lower.

 

That seems to be it, yep! Now at the PC and I can see it!

Yep, don't think so either. Would've been interesting to see how big the loss is though. But don't feel bothered, I hate changing memory stuff...

 

Memory OC is hateful, need to climb behind the rig to get at that clear CMOS switch lol

Playing about with a nice sample of a ryzen 5 3600 with SMT off and trying to get it stable at 4.45 ghz for assetto, has anyone worked out a way to set the game to run on a particular CCX? Windows prefers CCX0 but I could push 4.5 on CCX1 but the main game thread is always on the weaker CCX.