Triple setup and SOL - struggling with 60fps, is that normal?

[Lizard Of Bodom]

Hi all
I finally finished setting up my triple monitor setup and to my disappointment I noticed benchmark tool is struggling to get anything close to 60fps, even with graphic options turned down pretty drastically. I have Ryzen 3700X + GTX980(SLI) so didnt really expect it 3x1080p setup would be an overkill for this machine. I have read a lot about triple setups, most people reporting very posotive results on single cards like 1070 etc so I am wondering if thats because I use SOL vs "naked" AC? Or most of reports are with graphic options toned max down to achive best fps?
Its my first multi-monitor build so maybe I just got my hopes too high, or I am doing something wrong on graphic setup - not sure if SOL has some specific settings I might have missed, except of disabling AA, reducing shadows/reflections etc etc...?
any help appreciated

3700X + 16GB RAM + 2xGTX980OC
3x1080p screen in Nvidia Surround

 

[Kek700]

I can give you what I have with my triple monitor setup.

i5 6600k at 4.6Ghz
GTX 1080
important, 8GB 3600mhz ram
3x1080p 144hz monitors
16x
4x
high
med
no post processing
reflections high, no mirror reflections
med
static
22 oponents all gtr gt3's
brands hatch gp
10.00am,20 deg C ,optimum, 98% strenght.
bench mark
points 26932
avg fps 183

in game play at brands hatch
start 105 fps after 10minuits of driving 135 fps
gpu temp 52 deg C
cpu 48 deg C 70% usage at 4.6ghz

 

[Billy Pilgrim]

Sol isn’t the problem. I barely noticed any FPS hit going from naked AC to AC plus Sol.
Your weak point could be the CPU
I have triples, i7 at 4.3 approx., 1080ti, all my graphics settings on very high or max or whatever. I average around 100 FPS.
I used to have gtx 780s in SLI and got very good FPS with thatSLI too

 

[Lizard Of Bodom]

Thanks For the replies.
Kek700 - much obliged,thats a great info.
if you are getting 186 on benchmark, it means something is seriously not right on mine. In theory 980SLI should be much more powerfull than 1080, makes no sense to struggle with 60fps... need to do a test with single 980 and compare how it goes

Billy Pilgrim - i would argue the CPU point, its 3rd gen Ryzen 8c/16t with 4.2Ghz, it barely notice 20% of usage when working. I would suspect SLI issue at this point... WIll do more testing over weekend and report back, thanks all!

 

[Deleted member 197115]

Does AC support SLI, and if CPU is bottleneck, GPU doesn"t matter really.
Check Main T utilization in render stat app.

 

[Lizard Of Bodom]

Does AC support SLI, and if CPU is bottleneck, GPU doesn"t matter really.
Check Main T utilization in render stat app.

Yea, AC supports SLI and apparently pretty well
elaborate please on "main T" and "render stat" app? withing AC you mean?

 

[Kek700]

Post your benchmarks, may help quantify your system with AC.

 

[Jens Roos]

SLI is well supported in AC (have 1080TI SLI), does the 980 support HB bridge or just single connector? Check that your cards is running in PCI express 3.0 x16

 

[Deleted member 197115]

Yea, AC supports SLI and apparently pretty well
elaborate please on "main T" and "render stat" app? withing AC you mean?

Yes,

Spoiler: AC Render Stats App

Render Stats App
---------------------------
FPS - Frames Per Second, your framerate
FT - Frame Time, time in milliseconds for how long a frame takes to render
Main_T - see below
DIP - Number of draw calls (not really important to the average player) DIP the better the performance in general, Indicating Optimized game assets like tracks, cars)
TRI - Total number of triangle polygons
B.Size - Average size of draw call batches (again, not important to the player but the lower the better)
SCENE - Tri count of the scene (after culling) on screen
PHY - see below
PHY_LATE - same as above, shoulderstand indicate the number of times physics calculation Took over 3 ms per frame
FMOD USAGE - FMOD is the sound engine used in AC. Not sure what the percentage exactly to but high values have impact on performance.

Main_T value is for the graphics thread. It shows the relationship between the time spent in the application (CPU) and the time spent to complete a frame (FRAME TIME). Values of CPU near 100% indicate a CPU-bound system graphically.
EX ... if a frame took 16ms to complete and CPU is 50% it means that AC was able to prepare the frame in 8ms .. and the GPU took 16ms to render it. (GPU bound) .. if CPU is 100% then it took the CPU 16ms to prepare the frame and the GPU is probably capable of faster (CPU bound).
No relationship whatsoever with the value read in task manager.

PHY is relative to the physics thread, values above 100% indicates that the system is not ready to calculate the physics in time (3ms per step) and that the game is now running slower than realtime (slowmo). Practically, there are too many AI and the CPU is not ready to handle it.

 

[Billy Pilgrim]

Thanks For the replies.
Kek700 - much obliged,thats a great info.
if you are getting 186 on benchmark, it means something is seriously not right on mine. In theory 980SLI should be much more powerfull than 1080, makes no sense to struggle with 60fps... need to do a test with single 980 and compare how it goes

Billy Pilgrim - i would argue the CPU point, its 3rd gen Ryzen 8c/16t with 4.2Ghz, it barely notice 20% of usage when working. I would suspect SLI issue at this point... WIll do more testing over weekend and report back, thanks all!

I believe that AC doesn’t use multiple threading in CPUs so the 20% usage may just mean that all the threads aren’t being used - no surprise there though as AC won’t use all your CPU’s threads.
That said, your 4.2 ghz should be good enough (it’s similar to what I get from my CPU)
SLI works very well with AC.

 

[Kek700]

posted on wrong forum.

 

[Jens Roos]

I believe that AC doesn’t use multiple threading in CPUs so the 20% usage may just mean that all the threads aren’t being used - no surprise there though as AC won’t use all your CPU’s threads

It's something like that, if your main thread usage in-game fps app, goes above 90% you start loosing fps, it will not go above 95%. After installing the CSP I got an +20% load that got fixed after uninstall and verify files using steam, you could always try that and see if it helps.

My in-game benchmark is this (tested today):
AC VERSION: 1.16.3 (x64)
POINTS: 28062
FPS: AVG=191 MIN=50 MAX=270 VARIANCE=21 CPU=80%

LOADING TIME: 12s
GPU: NVIDIA GeForce GTX 1080 Ti (7680x1440)
OS-Version: 6.2.9200 () 0x100-0x1
CPU CORES: 8
FULLSCREEN: ON
AA:4X AF:16X SHDW:2048 BLUR:0
WORLD DETAIL: 4 SMOKE:2
PP: QLT:4 HDR:1 FXAA:1 GLR:4 DOF:4 RAYS:1 HEAT:0

One could expect similar result using 980 SLI and lower resolution, above gives solid 100 fps
The benchmark is a bit optimistic...

 

[Lizard Of Bodom]

Thanks all for the advices, it allowed me to test the setup a bit more and confirm that hardware-wise all was 100% correct. However problem still persisisted, so I started to dig deeper and finally nailed the issue.
SOL itself was not a problem, indeed, but CSP on the other hand - was. One setting to be precise - sparks. For some odd reason, sparks=active setting in Particles FX makes total mess on SLI and actually make FPS drop to 50% or less. deactivating it, made FPS go to the roof and even with all settings maxed out (4xMSAA, 16x, full shadows/world/smoke/reflections/mirrors) I am able to benchmark 110FPS average.
It must be some bug withing CSP precisely related to SLI as on 1 card that I had before, there was barely any penalty for using sparks.
Thats on CSP 1.25 buid 208 I think.

Problem sorted :}

 

[Jens Roos]

Yes, I noted this yesterday when I tested CSP Light that is included in the latest IER mod.

Stock Kunos Audi LMP1 at Brands, at the start, last position:

One thing to note, even on all stock content an extra 20% main thread load was added. The load are different depending on combination of car/track, when over 90% you start loosing FPS.

Then I got informed that this is because I was running CSP Light and not the full version, as full version is even better than stock. so installed the full version instead.

Well, the main thread load was less, around 45% but FPS was also reduced to around 80 instead with less GPU load... Might be an SLI issue, at least CSP Light utilize the GPU better and I can easily switch CSP on/off with two scripts (still on standard AC).

I try find and change that spark setting and see if it have any impact.

 

[Lizard Of Bodom]

I try find and change that spark setting and see if it have any impact.

spark is in PArticles FX settings of CSP, although I have freeling it may not be an issue for you if you have single card. My main_T when testing was around 40% with sparks deactivated...

 

[Jens Roos]

if you have single card

No, I'm on 1080TI SLI, i7-7700K 5GHz

 

[Lizard Of Bodom]

No, I'm on 1080TI SLI, i7-7700K 5GHz

Sorry, missed this fact. So did you try it, did it help?

 

[Jens Roos]

I'm currently back on CSP light, the setting is:

[BASIC]
ENABLED=0 ; Active (set to 0 to disable Particles FX); 1 or 0

[SPARKS]
ENABLED=0 ; Active; 1 or 0

I plan to re-install the full version this weekend and see what is possible, CSP do add some nice features.