Is ACC VR going to be a hot lap simulator ?

I have a GTX 1080Ti / 7700k @ 4.9 / 16GB 3200 RAM system and have been doing some testing .

At 1920x1200 by myself and sunny i get 150 odd FPS with GPU using close to 99%.
If i add 20 cars i now have 113 FPS and 65% GPU usage even if i sit on the start line and let the field disappear Now if i go and make only one car visible in settings my GPU usage jumps back to 99% . Then set back to All visible usage drops to 65% again even though no other cars are anywhere in view (they are on the other side of track)
This worries me for VR as even with Low settings at 1920x1200 with my 7700k at 4.9GHz im not near 180FPS with only me visible !
Why do the Ai need to have the same physics as the player ? they don't care if they feel Sim of Arcade as long as they react to the player accordingly what does it matter what physics they use other than to cripple my CPU ?

This is only very early testing from me and i know things should improve but still i'm a little concerned ACC wont be great for VR users unless the Intel 28 core becomes mainstream and cheap !

 

It might be my eyes but in AC1 I get 45fps and I find that perfectly smooth. I only start to notice if it drops below 35fps. So I would be very happy with 113FPS.

 

I asked that in AC already: why that huge share of CPU resources used on full AI car physics if a simplier model for them hardly, if at all, would be noticable to the player. Its imo too much effort and focus and resources spend on something that makes too little, if any, difference. Other sims do not do it like this, but use simplified physics - and I am unable to tell a difference. AC never offered any advantage there. Much more important for the visible AI car behavior is the quality of the AI. And here, other titles lead.

Differences that can be measured in the lab, but are not apparent to me in practical everyday reality, I do not care for, honestely said. I did not get a satisfying reply back then for AC on this question, and I expect to get none for ACC as well.

The 0.1.2 version by no means is optimized, and the game is far from finished, but I am slightly worried about frames in VR and resources as well - and my rig is even slightly more powerful than yours, I use an OC version of the 1080Ti and have a i7 8700K. The rendering distance of cars is not satisfying currently. Crashes haunt me when pairing full car field, night, and heavy weather. Reducing car numbers seems to cure it, but I did not test it systematically, the game currently is shelved in wait for 0.2.

With a hardware-critical sim like this, resources should not be "wasted" for "gains" and differences that are not really apparent to the eye. It would be indeed a waste if said resources are desperately needed for other things.

I do not buy the argument that AC's AI cars are doing better due to their more complicated way of calculating physics (=using the full physics). That would only be an argument if indeed the resulting AI car behaviour would be superior to that in sims not using this method. But it isn't. And some others indeed offer the better AI car behavior. So the way AC makes its AI cars driving, is not superior at all. In RR, the AI cars use simplified physics. I cannot perceive it, however. And so do not care one bit.Why should I? For the theoretical knowledge of them using the same physics engin e loike me?

In the end, a simulator always is a bag of tricks, cheats and shoretcuts. The trick is to not let the auiedince realise it. As long as the audience does not realise it, anything goes, and every cheat is acceptable. Even more so if it is more economic with the resources available

 

I really dont like AI racing, its all about multiplayer for me, even when practicing I will just do it on my own or in a practice lobby with other people, so with no AI running on the same physics engine might help me in that regard.

 

In VR, you want 90 fps stable. With that both-displays-combined resultion of today's headsets. 45 already is a compromise, and you can see it.

On a monitor, you can see the difference between 30 and 60 fps when watching angle speeds (your view rotates, the envrionment around you turns). When driving straight, it is indeed less prominent. Some pro players report increased latency times of the steering input when frames are low, but I am not sure I understand the background for that, and I never had that issue, at least not in a way that I took notice of it.

 

If lowering the visible cars limit improves performance, it's not the physics which is causing problems. The AI and physics is still being calculated for all other cars, they're just not being drawn.

I just checked and confirm what @Stains found. With a full grid and all opponents visible, I saw GPU use of 56% and about 74fps. Lowering the visible opponents to 1 changed that to 84% GPU use and 110FPS, which is a huge difference.

 

Yep, i asked that question to the dev's in AC, when AI first appeared.

Specifically... could we set the AI to 'half' physics rate.
The answer iirc was, AI driver animations were tied to the physics(along those lines).

At least we've got the option this time around to set the number of visible AI.
However, i would still like the option to set AI to a lower physics rate, if that at all is/will be possible in ACC.

 

It's a worry for sure how this will turn out for VR, I know i will certainly need to upgrade my PC soon , but I'm in favour of full physics for the AI.
Endurance racing will be my main form of racing and I don't really like the idea that the AI can survive longer because of simplified physics.
They should be vulnerable to the same issues and wear and tear as I am. If It was only short stint racing i was interested in I'd probably just race online.
I guess an option as was stated would be ideal.

I'm as dumb as a box of frogs when it comes to programming and such but it would be nice to have the ability to have another PC handle the AI somehow. I'd be quite happy to invest in a simpler PC to run alongside my main PC if it would help with such matters.

 

That 113 fps will translate to 35 in VR and that is totally not acceptable !
We need approximately 180fps in single screen to be close to 90 for VR .

@Skybird i agree , the tricks from developers to make us believe the Ai are real is what makes the other games so good for Ai driving !
What about MP ? Are we having to calculate all the other players cars on our rigs when we race MP , lol
As it stands Ai racing needs work and Kunos could maybe compromise their Ai must use player physics so we (the actual player) can use a full field and not be totally crippling our CPU !
Please @Kunos cut the heavy physics for Ai and just make them believable !

 

When I changed from full grid visable (even though they where on the other side of track , so not actually visable to me) my GPU load stayed just as low and around the same as when the grid was all just in front of me ! So i came to the conclusion that the Ai being calculated is indeed heavy for the CPU !

 

I can't stand to have cars disappear and pop in around me ! That totally kills immersion ! We have massive rigs to help eliminate pop in yet we have to make it an option so our rigs can hanble the Ai physics ?

 

lol , I like that idea but i as you have know idea about coding !
I tried racing the Ai (albiet poorly with a DS4 on flat screen) and found the Ai to not handle at all like my car and they were able to hold massive pushes from me while on the grass ! Not at all realistic

 

You're misunderstanding what simplified physics means.
It has nothing to do with damage and wear and tear, it simply means less calculations = less cpu/resources.

The AI will still take the same amount damage they would as if they were on full rate physics.

 

I didn't mean damage, what i mean is if they're hopping over curbs and colliding with each other that would normally upset my car and cause a spin but they just continue on as if nothing happened because of simpler suspension and tyre physics , I wouldn't be a fan of that. PCars 2 is a good example, those AI are nuts.

 

As I said they already drive well and truly easier than the player car ! They are able to be pushed along the grass and maintain control as though they are on the track almost like they are magnetically being pulled back on !

 

Yeah hopefully early days though and they're going to get better, certainly don't want them dumbed down even more.

 

Could you explain how you know definitively that 'That '113 fps will translate to 35 in VR'? I assume you have tested the final release version, if you could publish your testing methodology that would be great.

I hope and expect it to be a great hot lap simulator (whatever that means)and a great race simulator

 

Maybe if they used lesser physics they may actually be able to drive better than they already do in AC1 !

Here is a quick and dirty video demonstrating my PC with AC2

 

lol , that 35 fps in VR was just a guesstimate all i know is if i only get 100 odd fps in 2D it won't translate well to VR which needs 90 fps to be a great experience !
But im sure you already know that and are just being a smart ar$3
Oh and yeah mate I'm actially playing AC3 atm as i can't be arsed playing old AC2 anymore

 

I think your conclusion is incorrect. Reducing the opponent visibility setting to the minimum produces a *very* significant increase in performance (over 30%), even though the physics and AI calculations are still happening for 20 cars.

When opponent visibility is set to 'All', expected behaviour would be for the frame rates to equal the above when no other cars are in view. The reality is that there is only a small improvement, them being 'visible' but not on-screen is obviously hindering performance.