AC GT4 @ Road America - Wednesday 22nd April 2020

[Faceplant]

Instead of engines overheating like in "the old world", it's now the cpu's in our rigs before a race. Immersive

 

[stribbeck]

Afraid it didn't go well for me the pc wasn't enjoying it so had to quit, had a collision in the second race as a car just appeared inside mine on the straight so apologies to who ever that was

 

[Interslice]

I think its the trackside detail* The amount of trees alone could be the issue? I was briefly getting a similar cpu 99% thing the other night when on the shutoko revival project.

*Not a computernichian

Beautiful track though. Great combo again john. Was hoping for a stronger BOP help for the porsche but wasn't to be!

 

[GEO147]

Yeah the Cayman was a bullet in staright line.

Had a great time last night. The cars are so nice to drive. Had some very close and fair racing with Matt in race 1. Nice racing Matt. Then in race two Matt, me and Ronnie were 3 wide for the pit straight and into turn 1. I was on inside and Matt in middle, I left you loads of room Matt but maybe you didnt see me there and turned in? Pity it sent you off.
There was a lot of lag and cpu issues and at times I think there was a lot of netcode issues as there were touches happening even when there was room!

Thanks for organising John, great combo.

 

[Interslice]

There was crazy stuff going on ahead of me with cars getting taken out by a lagged out car that just didn't brake or turn for the corner... fast forward a second or two and the 'offender' reappears half way around the corner oblivious to the colision their car had caused in another dimension.

 

[HTIDRaver2k8]

I had a few issues, I think one with Steve (Mitchell) where he disappeared into the shadow realm, then reappeared mid corner and round we both went.

Didn't have any CPU warnings even whilst streaming.

I had more issues with the DTM race for FPS/CPU lag.

But it was a good night of racing and I finally feel like I am getting somewhat confident in racing with other people now.

 

[GEO147]

I had a few issues, I think one with Steve (Mitchell) where he disappeared into the shadow realm, then reappeared mid corner and round we both went.

Didn't have any CPU warnings even whilst streaming.

I had more issues with the DTM race for FPS/CPU lag.

But it was a good night of racing and I finally feel like I am getting somewhat confident in racing with other people now.

It was me that you had your first incident with in turn 1. You said in your stream not sure what happened there. I wasn't sure either as I thought there was room and pretty sure we didn't touch on your screen nor mine.
I managed to keep her going but you ended up going off

 

[kenz247]

sorry lads i was unable to race due to CPU warning

 

[HTIDRaver2k8]

It was me that you had your first incident with in turn 1. You said in your stream not sure what happened there. I wasn't sure either as I thought there was room and pretty sure we didn't touch on your screen nor mine.
I managed to keep her going but you ended up going off

I apologise George I could have sworn it was Steve.... :/

I let a Steve pass me and thought it was all concluded :/

Really sorry bud

 

[GEO147]

No apology required. I think it was netcode myself. We were both clean.

 

[pattikins]

I think its the trackside detail* The amount of trees alone could be the issue? I was briefly getting a similar cpu 99% thing the other night when on the shutoko revival project.

*Not a computernichian

Beautiful track though. Great combo again john. Was hoping for a stronger BOP help for the porsche but wasn't to be!

It probably didn't help, but I believe Kunos limited the number cars (24) per track for a good reason. AC only uses four cpu threads, with two treads generally carrying out the bulk of the work. Add in nearly 40% more cars beyond the limit set by Kunos & it's not surprising to find that cpu's start to tank. I went from a buttery smooth 90fps (VR) to sub-reprojection levels once the grid started to fill up. It was more driving by memory (sketchy @ the best of times ) than any feel I had for the car. I fear Sunday will be the same.

 

[Medilloni]

There was crazy stuff going on ahead of me with cars getting taken out by a lagged out car that just didn't brake or turn for the corner... fast forward a second or two and the 'offender' reappears half way around the corner oblivious to the colision their car had caused in another dimension.

Sadly, I think I was involved in that Brian - I had to leave the 1st race after a couple of laps, I got a cpu warning as the car slowed down, next thing I knew I'd crashed 100's yards up the road My apologies for all those affected.

After that I logged out, lowered all my settings (halved frame rate etc) and it as fine for the 2nd race. I dread to think what's going to happen on Sunday's race with 40 cars

Edit: Just seen Patrick's post... I'm wondering if Sunday might be better run on 2 servers, if you @Chris Down wanted a hand with that Chris I'll be free to fumble help this Sunday.

 

[Faceplant]

One day lag will be programmed as a tire blowout

 

[Chris Down]

Sadly, I think I was involved in that Brian - I had to leave the 1st race after a couple of laps, I got a cpu warning as the car slowed down, next thing I knew I'd crashed 100's yards up the road My apologies for all those affected.

After that I logged out, lowered all my settings (halved frame rate etc) and it as fine for the 2nd race. I dread to think what's going to happen on Sunday's race with 40 cars

Edit: Just seen Patrick's post... I'm wondering if Sunday might be better run on 2 servers, if you @Chris Down wanted a hand with that Chris I'll be free to fumble help this Sunday.

Hmm, we have run 36 GT3s before without much issue, so I don't think it should be a problem. One of the problems here is the G55 far-LOD version is not super low poly, Kunos cars are better in that regard.

Last time we ran 2 servers people seemed not super happy about it, so I'd ideally like to avoid it.

It probably didn't help, but I believe Kunos limited the number cars (24) per track for a good reason. AC only uses four cpu threads, with two treads generally carrying out the bulk of the work. Add in nearly 40% more cars beyond the limit set by Kunos & it's not surprising to find that cpu's start to tank. I went from a buttery smooth 90fps (VR) to sub-reprojection levels once the grid started to fill up. It was more driving by memory (sketchy @ the best of times ) than any feel I had for the car. I fear Sunday will be the same.

Heh, I anticipated this question after race 1, and the thing which made this worse last night is that the max clients for the server was 48, and 48 cars were loaded -- even if those cars aren't used, it still takes passive load on client CPU during racing. After race 1 I quickly profiled CPU time spent (and thankfully Kunos still includes the debugging symbols for acs.exe), and tons of it was just spent in updating car physics and presence for idle cars -- PhysicsEngine::step, PhysicsAvatar work, and Car::getPhysicsState attached to ostensibly idle Car objects.

 

[GEO147]

Fair play regarding the know how. Yeah I didn't like the 2 server split that time. I felt it really killed the excitement of it.

 

[HTIDRaver2k8]

I don't mind the server splits but I prefer the larger grids.

Interestingly I rarely run into CPU issues now I've adjusted my FFB settings.

Prior to doing so I'd get a CPU occupancy warning in a single player practice.

And it's not like I have a cutting edge chip, it's just a simple quad core (with only 1 logical thread per core).

James plays on his laptop (A Razerbook admittedly) but doesn't seem to have any issues.

 

[Chris Down]

Here's a small edit of one of the amazing battles last night with @Aksu and @Axel Moebius

 

[Prince-Matt]

Had a great time last night. The cars are so nice to drive. Had some very close and fair racing with Matt in race 1. Nice racing Matt. Then in race two Matt, me and Ronnie were 3 wide for the pit straight and into turn 1. I was on inside and Matt in middle, I left you loads of room Matt but maybe you didnt see me there and turned in? Pity it sent you off.

I would say it was more your fault because I think u had enough space to take the corner tight, but it doesn´t really matter. I would identify it as a race incident / misunderstanding. And I am not mad about it cause I had no fun on the stock setup on the G55 so i left Ofcorse more reaaaally close fighting that we had would be better but it is how it is

BTW Ronnie, I hope u are not angry about the draft bumping on the straight? It was necessary to overtake Geo, and it worked

 

[Medilloni]

.......Heh, I anticipated this question after race 1, and the thing which made this worse last night is that the max clients for the server was 48, and 48 cars were loaded -- even if those cars aren't used, it still takes passive load on client CPU during racing. After race 1 I quickly profiled CPU time spent (and thankfully Kunos still includes the debugging symbols for acs.exe), and tons of it was just spent in updating car physics and presence for idle cars -- PhysicsEngine::step, PhysicsAvatar work, and Car::getPhysicsState attached to ostensibly idle Car objects.

Thanks for that info Chris, numpty that I am didn't think unused cars would sap resources. The good news is I can (hopefully!) put my settings back to 'acceptable'

Whilst you're here I hope you don't mind me picking your brains... are Anisotropic and Anti-aliasing settings resource-hungry? I have Anisotropic at 4 and Anti-aliasing at 2, but would like to run them higher if poss, stops the headaches, and I can see braking points Frame rate is max 66 btw.

 

[Chris Down]

Thanks for that info Chris, numpty that I am didn't think unused cars would sap resources. The good news is I can (hopefully!) put my settings back to 'acceptable'

Whilst you're here I hope you don't mind me picking your brains... are Anisotropic and Anti-aliasing settings resource-hungry? I have Anisotropic at 4 and Anti-aliasing at 2, but would like to run them higher if poss, stops the headaches, and I can see braking points Frame rate is max 66 btw.

Anisotropic filtering is basically free nowadays (AF performance is primarily dependent on memory bandwidth, which is generally not a problem on reasonably modern computers) so you can run with 16x without worrying That said, if you're on a laptop with an integrated GPU, you might want to still keep it at 4x or 8x.

For MSAA, performance hit is exponential. So 2x is usually fairly negligible, but you're more likely to feel 4x or 8x+. FXAA is fast but can cause some objects to become more blurry, since it's significantly less precise in its evaluation (that is, it literally just evaluates something close to the final output, rather than looking at 3-dimensional objects and how textures apply and are being viewed).

Here are some settings that should be reasonably nice looking, without being too taxing even on an older card and CPU. Set your FPS limit to your refresh rate