AC GT3 @ PORTIMAO, Sunday 2nd May 2021

[demetri]

It is a thing, but it would need developing for specifically and I guess VR games (or VR modes) are just not a big enough piece of the pie for developers to do this.
Some details on NVidia website about VR SLI:
https://developer.nvidia.com/vrworks/graphics/vrsli

I think it's dead in the water

 

[jackdyson31]

T1 comparison.

RD version

"Other" version

Both clips are taken from about the same spot and the same scale (based on the length of the Audi). The other track is significantly narrower which calls for a later entry at a lower speed. Pit exit is also much closer to T1

Sorry this is late Dmitry, and thank you for all the data and videos again, I checked the videos and the plan views you posted here and they are substantially different. I think in the one we raced the corners are totally blind with respect to the other one. From the real footage - I note they are more often blind than not, except for the crucial penultimate one as you go up the last rise.

Well at least this explains the lines I saw you take ...

 

[jackdyson31]

I think it's dead in the water

Shouldn't be on ACC according to the NVIDIA page as the technology does support Unreal engine - thus you would need a profile by NVIDIA to engage the parallelisation pipelines on SLI and AFR should turn on for ACC when run on Turing GPUs.

Here look this is interesting:

SLI Support?

Will we see SLI support, for multiple Nvidia graphic cards, now that RTX is available?

www.assettocorsa.net

(post #19)

and here are the full release notes :

http://us.download.nvidia.com/Windows/419.35/419.35-win10-win8-win7-desktop-release-notes.pdf

night

 

[demetri]

Sorry this is late Dmitry, and thank you for all the data and videos again, I checked the videos and the plan views you posted here and they are substantially different. I think in the one we raced the corners are totally blind with respect to the other one. From the real footage - I note they are more often blind than not, except for the crucial penultimate one as you go up the last rise.

Well at least this explains the lines I saw you take ...

I can only see one blind corner in the real-life video. It is T10 (uphill right-hander followed right away by a downhill right-hander). But that's a normal corner, not a hairpin like T13, and yes, they start braking before seeing the apex, but it becomes clearly visible by the moment they need to turn-in, so it's not a big deal. Also, there're several lines through T10-T11 (you can go deep outside between them, or try to stay closer to the inside curb all the way) which are all OK and making a small mistake on approaching T10 is not going to throw you off the track or make you lose too much time.

Both tracks are not 100% true to the real one but at least you can drive and enjoy the other one. RD version feels like an unpleasant job you have to do. Meh...

 

[demetri]

Shouldn't be on ACC according to the NVIDIA page as the technology does support Unreal engine - thus you would need a profile by NVIDIA to engage the parallelisation pipelines on SLI and AFR should turn on for ACC when run on Turing GPUs.

Here look this is interesting:

SLI Support?

Will we see SLI support, for multiple Nvidia graphic cards, now that RTX is available?

www.assettocorsa.net

(post #19)

and here are the full release notes :

http://us.download.nvidia.com/Windows/419.35/419.35-win10-win8-win7-desktop-release-notes.pdf

night

That's a different thing. AFR - is alternate frame rendering where one card renders all odd frames and the other one does all even frames. It is intended for the flat screens.

For VR stereo, there's a need to render two independent views for the same in-game timepoint and those views are quite similar as their cameras positions are just 6-7cm apart. You would think that asking one card to render left eye view and the other card to render right eye view should do the job and should be easy to implement, and there should be no major synchronization issues as the views are very similar and should take nearly the same time to render. But maybe there's something else that breaks the idea.

One thing that comes to mind is Asynchronous TimeWarp, which quickly transforms the rendered picture based on the latest headtracking data to compensate for the head movement that happened after the initial rendering started. This is done to prevent "input lag" that will occur otherwise when you move your head and will probably break immersion and make you sick. Maybe you need both left and right frames data in order to make such tansformations, so when each card does its own eye you would have to implement data exchange between them and that kills all the performance boost achieved.

 

[jackdyson31]

but it becomes clearly visible by the moment they need to turn-in, so it's not a big deal.

yeah and then thats no issue at all - but I found myself needing to see the track before it arrived if you see what I mean so it needed a bit more practice than 2 hours which is all I could manage last week - just so many other problems to deal with.

Also, there're several lines through T10-T11 (you can go deep outside between them, or try to stay closer to the inside curb all the way)

Right - I always tried to minimise the distance so I stayed inwards, but I saw some of the faster people on really wild lines.

Both tracks are not 100% true to the real one but at least you can drive and enjoy the other one.

right - I wish there was a laser scanned jobbie of this one ; wicked track ; no idea what max was whinging on about this week.

RD version feels like an unpleasant job you have to do. Meh...

lolololol .... so sound like a connoisseur who discovered a few insects in a 50 year old bottle of the old vino

 

[jackdyson31]

That's a different thing. AFR -

Ok right - didn't realise

and those views are quite similar as their cameras positions are just 6-7cm apart.

yeah exactly, a trig perturbation formula comes to mind...

. But maybe there's something else that breaks the idea.

ok - I won't pretend to understand because I have not delved into the API properly

will probably break immersion and make you sick.

Neck FX is murder on the eyes and brain so I can only imagine what getting this wrong does...

its own eye you would have to implement data exchange between them and that kills all the performance boost achieved.

Ok when each card locks the main/global memory bus this would massively slow everything down, but the NVIDIA Tesla cards and one of the older NVIDIA XP Pascals does support dual memory transfers to global RAM. so it might not be that bad. As you say though - worth it? Who runs ACC on a Tesla ? And the XP is long in the tooth so to speak ...
so yeah quite possible all this gets chucked into the "who cares" corner of the room.

 

[demetri]

yeah and then thats no issue at all - but I found myself needing to see the track before it arrived if you see what I mean so it needed a bit more practice than 2 hours which is all I could manage last week - just so many other problems to deal with.

I had driven about 2,200 kms on it while practicing before the event. Learning the track, optimizing the setup, getting more consistent for the long race. That's nearly 500 laps and more than 13 hours of driving.
Some people can get up to speed quickly, but not me, I have to work for it

 

[jackdyson31]

I had driven about 2,200 kms on it while practicing before the event. Learning the track, optimizing the setup, getting more consistent for the long race. That's nearly 500 laps and more than 13 hours of driving.
Some people can get up to speed quickly, but not me, I have to work for it

You are one of the fastest and most technically able men here - mighty respect from me and much for me to learn from your example Thanks a lot for that Dmitry, sincerely, I really admire it. I work very very hard too usually, not quite your level but still the maximum given mine, unfortunately, over the last two weeks health problems have meant very little intense work - but, reading what you have said is massively encouraging for me going forwards and of course "I'll be back"

 

[demetri]

You are one of the fastest and most technically able men here - mighty respect from me and much for me to learn from your example Thanks a lot for that Dmitry, sincerely, I really admire it. I work very very hard too usually, not quite your level but still the maximum given mine, unfortunately, over the last two weeks health problems have meant very little intense work - but, reading what you have said is massively encouraging for me going forwards and of course "I'll be back"

Thank you and I wish you a speedy recovery