AC GT3 @ Bahrain - Sunday 28th March 2021

[RasmusP]

Not that helpful for your specific issues but I just thought I'd swiftly compare the performance of Bahrain vs Monza:
90 fps limited! CSP set for mostly optimizations, fancy stuff disabled apart from TAA, AC settings mostly high to best and single player practice session at the starting line.

What jumps to my attention:
- MAIN_T 43,5% vs 29,9%
- RENDering thread 3.166 vs 1.371 !!!!!!!!
- CPU 4.6 ms vs 3.4 ms

This is mostly due to the DIP ; TRI ; B.SIZE ; SCENE stats.
I'm not sure what they are but these are basically the geometry of the track. Track makers try to keep these as low as possible, while maintaining a good look.

You can see that Bahrain has around 3x the DIP, 2.5x the TRI and 2x the SCENE.
This explains the way higher REND stat.

My GPU load for both:
Bahrain ~73%
Monza ~65%

So a bit higher load on the GPU. If you were struggling at the start, my first bet would be the CPU but if you're at the limit with super sampling, the GPU might be the bottleneck.

But:
- CPU from 3.4 ms to 4.6 ms = 35% higher load for Bahrain
- GPU 65% to 73% = 12.3% higher load for Bahrain

 

[RasmusP]

I'm confused. fpsVR was showing me the same, that it was gpu related. Yet my gpu only runs around 50% (1080ti) & looking @ the fpsVR history viewer, my max cpu thread usage briefly reached 85% (bulk @ 55-75%).

Did you drop into ASW (or whatever it's called for your headset)?
If you reach the limit of your GPU, the fps will drop to 50%.
The headset/vr software then knows that your GPU load is too high for native fps and won't try to reach them, while your GPU load will hover around 50%.

50% is right at the edge though! If you'd see 60% GPU load it would be clearer...

 

[demetri]

Not that helpful for your specific issues but I just thought I'd swiftly compare the performance of Bahrain vs Monza:
90 fps limited! CSP set for mostly optimizations, fancy stuff disabled apart from TAA, AC settings mostly high to best and single player practice session at the starting line.

What jumps to my attention:
- MAIN_T 43,5% vs 29,9%
- RENDering thread 3.166 vs 1.371 !!!!!!!!
- CPU 4.6 ms vs 3.4 ms

This is mostly due to the DIP ; TRI ; B.SIZE ; SCENE stats.
I'm not sure what they are but these are basically the geometry of the track. Track makers try to keep these as low as possible, while maintaining a good look.

You can see that Bahrain has around 3x the DIP, 2.5x the TRI and 2x the SCENE.
This explains the way higher REND stat.

My GPU load for both:
Bahrain ~73%
Monza ~65%

So a bit higher load on the GPU. If you were struggling at the start, my first bet would be the CPU but if you're at the limit with super sampling, the GPU might be the bottleneck.

But:
- CPU from 3.4 ms to 4.6 ms = 35% higher load for Bahrain
- GPU 65% to 73% = 12.3% higher load for Bahrain
View attachment 459397View attachment 459398

This basically confirms my guess that it was due to too many objects to render. This would affect both CPU and GPU, in my case (9700K @4.9 / RTX2080) GPU was the limiting factor.
Among the stock Kunos tracks, RBR is somewhat similar when you descend from T1 down to T2 and can see almost the whole track

I guess we may simply state that Tilke just can't build good tracks. Even in virtual world they have some stupid issues

 

[pattikins]

Did you drop into ASW (or whatever it's called for your headset)?
If you reach the limit of your GPU, the fps will drop to 50%.
The headset/vr software then knows that your GPU load is too high for native fps and won't try to reach them, while your GPU load will hover around 50%.

50% is right at the edge though! If you'd see 60% GPU load it would be clearer...

My framerate dropped in to the 70s & 80s for just a matter of a few minutes - towards the end of the qualifying round & the start of the race. Once I'd turned off a few AC apps there was a noticeable improvement. I was more surprised than anything else. With a Ryzen 5600X, I thought such things were a thing of the past.


fpsVR Report:
App: acs HMD: Samsung Windows Mixed Reality 800ZBA0 (90.002 Hz, IPD 63.1)
GPU: NVIDIA GeForce GTX 1080 Ti (27.21.14.5671) CPU: AMD Ryzen 5 5600X 6-Core Processor
Delivered fps: 89.68 Duration: 98.1min. Headset was active: 98%
GPU Frametimes:
Median: 7.4 ms
99th percentile: 11.1 ms
99.9th percentile: 12.8 ms
frametime <11.1ms(vsync): 99%
CPU frametime:
Median: 3.3 ms
99th percentile: 5.7 ms
99.9th percentile: 6.8 ms
frametime <11.1ms(vsync): 100%
Reprojection Ratio: 0.4% (for Index/Vive/VivePro headsets only)
Dropped frames: 14 or 0.0% (for Index/Vive/VivePro headsets only)
Max. SteamVR SS: 179%
Render resolution per eye: 1932x2416(by SteamVR settings, Max.) (HMD driver recommended: 1446x1808)

 

[Fiberoptix]

Did you drop into ASW (or whatever it's called for your headset)?
If you reach the limit of your GPU, the fps will drop to 50%.
The headset/vr software then knows that your GPU load is too high for native fps and won't try to reach them, while your GPU load will hover around 50%.

50% is right at the edge though! If you'd see 60% GPU load it would be clearer...

Just my observations but I have found VR performance to be much poorer in the modded tracks than the original ones in the game. This one in particular (maybe it was the lighting effects) I had to turn down quite a few in game settings to run a solid 90frames even alone in practice. In then end I enabled reprojection in steam so I could turn everything back up again.

 

[demetri]

With my I9-9900K and 5700xt i havnt had any issues. I run almost every setting on max, except smoke, mirror resolution and shadows.

With 3x1080p screens I had no issues at max settings either:

It would maintain stable 90 fps on race start with ~80% GPU load. Very different story in VR though

 

[RasmusP]

Once I'd turned off a few AC apps there was a noticeable improvement

This always indicates a CPU bottleneck. Apps are done in the main cpu thread afaik. Sadly...

With a Ryzen 5600X, I thought such things were a thing of the past.

Are you using CSP and the optimizations?
Without the optimizations, many mod tracks will drop your fps a LOT if you have more than the bare minimum of essential apps open, no matter the CPU.

With CSP you should never have issues though! But it depends on the apps...
And of course if you activate all the fancy stuff from CSP, you might get issues too.

I've posted my fps and render stats in the other thread (mx-5 Road Atlanta I think?) where you could see the CPU frametime varying massively depending on CSP on/off and the CSP settings.

 

[demetri]

This always indicates a CPU bottleneck. Apps are done in the main cpu thread afaik. Sadly...

Are you using CSP and the optimizations?
Without the optimizations, many mod tracks will drop your fps a LOT if you have more than the bare minimum of essential apps open, no matter the CPU.

With CSP you should never have issues though! But it depends on the apps...
And of course if you activate all the fancy stuff from CSP, you might get issues too.

I've posted my fps and render stats in the other thread (mx-5 Road Atlanta I think?) where you could see the CPU frametime varying massively depending on CSP on/off and the CSP settings.

One of the most important CSP optimizations is this:

Helps a lot with those GUI apps that show text output, including even built-in ones like Realtime app.

 

[RasmusP]

One of the most important CSP optimizations is this:
View attachment 459421

Helps a lot with those GUI apps that show text output, including even built-in ones like Realtime app.

Yep!
I had posted the comparison of stock ac vs just that one with my normal hud (sidekick, tyres, track map, acrl fuel) and it's definitely the biggest improvement of it all!

 

[Interslice]

I'm confused. fpsVR was showing me the same, that it was gpu related. Yet my gpu only runs around 50% (1080ti) & looking @ the fpsVR history viewer, my max cpu thread usage briefly reached 85% (bulk @ 55-75%).

I checked task manager sitting in the pits mid qualifying. It was telling me 96-100% cpu usage. It's a great problem to have, that there are so many top quality tracks available for free that we can pick holes and park those that aren't 100%!

 

[pattikins]

This always indicates a CPU bottleneck. Apps are done in the main cpu thread afaik. Sadly...

Are you using CSP and the optimizations?
Without the optimizations, many mod tracks will drop your fps a LOT if you have more than the bare minimum of essential apps open, no matter the CPU.

With CSP you should never have issues though! But it depends on the apps...
And of course if you activate all the fancy stuff from CSP, you might get issues too.

I've posted my fps and render stats in the other thread (mx-5 Road Atlanta I think?) where you could see the CPU frametime varying massively depending on CSP on/off and the CSP settings.

I've seen the thread & I've added your recommended settings to my dropbox. AC, on the whole, seems to be running fine with the vanilla version, but I will look @ the csp & SOL @ some point in the future.

 

[RasmusP]

I've seen the thread & I've added your recommended settings to my dropbox. AC, on the whole, seems to be running fine with the vanilla version, but I will look @ the csp & SOL @ some point in the future.

No need for sol!!
My csp is set to "weather controller" = "standard ac" or something like this.
So no sol for me!

If you want we can hop into discord and get you just the optimizations installed without changing anything else.
It will give your cpu some massive breathing room with 3 clicks

(it's a bit tricky to get rid of the "track & cars VAO (fake ambient occlusion for more 3d feel) and "configurations".
If you know how it's a matter of 10 seconds)

 

[demetri]

(it's a bit tricky to get rid of the "track & cars VAO (fake ambient occlusion for more 3d feel) and "configurations".
If you know how it's a matter of 10 seconds)

Does getting rid of that come with any tangible performance benefits?

 

[pattikins]

No need for sol!!
My csp is set to "weather controller" = "standard ac" or something like this.
So no sol for me!

If you want we can hop into discord and get you just the optimizations installed without changing anything else.
It will give your cpu some massive breathing room with 3 clicks

(it's a bit tricky to get rid of the "track & cars VAO (fake ambient occlusion for more 3d feel) and "configurations".
If you know how it's a matter of 10 seconds)

Can I take a rain cheque on your kind offer.

 

[RasmusP]

Does getting rid of that come with any tangible performance benefits?

Good question...
I don't think that it's more than 1-5% performance decrease. Afaik it's a few objects here and there but overall it's just colour changes and config texts like which light posts should light up etc.
The vao is just a fixed darker coloring that isn't "calculated".

But it changes the whole look of AC quite massively and at first, I didn't like it.
Brands Hatch was super differently, that's why I wrote it.

Took me half an hour to figure out how to delete the files (and if something happens if you just delete the files...).
There's no "uninstall configs" button...

Can I take a rain check on your kind offer.

Yes sure! It wasn't really meant like "right now" (although I would have time hehe) and more like "drop me a PM so we set a date & time for it"

 

[random2k4]

Is that down to the Radeon?

It would maintain stable 90 fps on race start with ~80% GPU load. Very different story in VR though

I am running CSP with the latest SOL, but only on 1x 1920x1200 display so my hardware is still a little overkill for that. Maybe I also got drops in performance but they are just not as visible and prominent as in VR or a tripple screen setup.

 

[demetri]

A short clip from the race. Taking the damn corner in style:

Too bad, but tires were most likely a toast after that

 

[beginer]

Not my best effort. Qualified 15th, was hoping for better. Got squashed in the first corner mess and ended up in 20th place. Really tried to keep it clean, but my efforts did not pay off.

I had some great battles with Mark, Chris and Colin, i think I passed the group twice just to spin off and end up at the back.

My setup allowed for faster corner speeds and slower in the straight. Passing in the corners is risky and my results confirmed it.

In the end I just settled to stay back and not push my luck any further.

Sorry for the bumps guys. Better racing next time.

Congrates to the podium winners and thanks Chris for setting this up.

I had tons of fun driving against you guys, too! We kept switching places again and again. I kept getting passed in the corners and making it up on the straits. I tried to put up a good fight without being overly aggressive defending corners. I hope i gave appropriate room when battling through the corners. If ever i do something stupid or unsportsmanlike, please let me know.
In general my driving seems to be improving since running with this group (I hope anyway). There are some talented people here and ive picked up a lot of tips by racing with yall and watching the replays.
Ive said it before and ill say it again: much love to you guys!

 

[Goblin616]

Can anybody please put up the GT3 server with Road America so we can start to practice? T.I.A.

 

[Kek700]

I have assumed it is being used by the Mazda mx5 cup brigade.