Cars RSS GT1 Championship

[LilSki]

After_midnight sees a 10% difference, test it what do you see and show screens ?

I'm not at home right now as I am at work.

 

[After_Midnight]

Just want to add to this discussion:
It seems to work just if you set the Reflection Quality to high.
If you set the Reflection Quality to medium there is no noticeable impact at the CPU load with different RRF settings.

At least on my system.

It's likely because you are simply shifting the bottleneck from cpu to gpu. The cpu is waiting on the gpu at higher settings and thus lower cpu usage.

Sound legit

 

[CC]

I'm not at home right now as I am at work.

I find it interesting though that no matter what settings I run and no matter ULTRA on MIN,
the biggest difference only on the CPU is changing it from STATIC to ULTRA,
ULTRA gives a better reflection + better CPU, I was confused to why and kept searching what I had changed took almost 20 tests but realized it was that 1 settings, like I say either way it appears to be backwards and yes maybe some bottleneck, but I would have thought within 5 years of release it would have been picked up or mentioned or recommended as a + setup on graphics ect

 

[CC]

Just want to add to this discussion:
It seems to work just if you set the Reflection Quality to high.
If you set the Reflection Quality to medium there is no noticeable impact at the CPU load with different RRF settings.

At least on my system.


Sound legit

I have only tried ultra and static which appears to be 30% on my system.

 

[CC]

Either way,
Our RSS pipeline on content internal or not now will allow us to make 501,000 triangle cars with the same performance as official content and will be twice the details, I found issues and mistakes I have corrected with years of experience and again yes its more time consuming and something not everyone will follow but be assured all my cars from now on will be insane quality from me.

 

[AccAkut]

yes its more time consuming .

Lucky you working without timed milestones..

...but serioosly, the car will run great as is, but for now it's only running on your system... get that thing to sellfy and then move onto equalising/updating the whole lot All those promised physics corrections and things, not 3 more LOD stages for all released

 

[CC]

Lucky you working without timed milestones..

...but serioosly, the car will run great as is, but for now it's only running on your system... get that thing to sellfy and then move onto equalising/updating the whole lot All those promised physics corrections and things, not 3 more LOD stages for all released

I mean with 300 tests running ultra and zero settings almost, why you think it makes a difference what system your running, personally I see not much difference if any, maybe 10% on the tests in general,
with a 1080 the test are achieving 2x what I am achieving and that card is twice as powerful, so roughly like I say depending on the systems take away the difference in graphics and roughly 10% variation I am destroying kunos content why would that only work on a 1070 with 6 x LOD and a brand new system on creating LOD,s.

 

[AccAkut]

I don't think I got my point across

 

[CC]

I don't think I got my point across

I understand what you said, but my system is not magical gains are gains maybe on some systems there will be a % difference but from years of experience this difference is minimal

 

[Andreas Knöpke]

On- and offline are also 2 very different stories in AC.

Offline 400 fps, same scenario online 120 fps although nobody else on track or in sight.

 

[AccAkut]

On- and offline are also 2 very different stories in AC.

Offline 400 fps, same scenario online 120 fps although nobody else on track or in sight.

With 15 AI offline I have ~60. Full grid online 100+ is possible - systems are very different

 

[CC]

On- and offline are also 2 very different stories in AC.

Offline 400 fps, same scenario online 120 fps although nobody else on track or in sight.

I noticed this also thats why the CPU is important cains the fps with bad cpu I am seeing here, lets check anyways we will do some tests tomorrow

 

[Ghoults]

So reducing overdraw. This means no alphatested or alphablended materials on lower lods? Maybe less shadows as well?

 

[AccAkut]

So reducing overdraw. This means no alphatested or alphablended materials on lower lods? Maybe less shadows as well?

yes. On stuff like grills and logo overlays I would switch to alpha_test first, than on LOD_C or so I would not use it at all, rather make all glass black with some reflection. My old 996 did that on 30 meter Lod_B already - no one ever complained (LOD_A was only 150k, never bothered with more than one stage)

 

[CC]

yes. On stuff like grills and logo overlays I would switch to alpha_test first, than on LOD_C or so I would not use it at all, rather make all glass black with some reflection. My old 996 did that on 30 meter Lod_B already - no one ever complained (LOD_A was only 150k, never bothered with more than one stage)

Yeah that is the trick make content to previous complaints, but yet no one never complains personally I see,
but then when you mention some magical improvements people say "well both rss and kunos lod both have the same visual differences" but yet its never mentioned in the threads live only at a later date,
i find at least dropping objects and materials each LOD helps massively but also the distance, was Gergo who 1st gave RSS updated LOD settings and we carried it forwards, nothing is never perfect, even this 57 update changes the LOD so much so visually there is more differences compared to standard techniques but the improvements in quality and performance outweigh the losses.

 

[NightEye87]

It's likely because you are simply shifting the bottleneck from cpu to gpu. The cpu is waiting on the gpu at higher settings and thus lower cpu usage.

This. Watch one of Stefano's streams where he explains the CPU stats.

Also, are you testing with your single car vs Kunos's multiple car benchmark? Might change things a bit

Anyway, good news!

 

[CC]

This. Watch one of Stefano's streams where he explains the CPU stats.

Also, are you testing with your single car vs Kunos's multiple car benchmark? Might change things a bit

Anyway, good news!

Yes you can add at least 5% variation to those stats minimum, you are correct and yes his streams was awesome I watched them all while in max the guy is a genius

 

[CC]

1 more thing I find interesting again not useful just chat,
but launch AC with your controller turned off and go ingame and then turn it on, nothing happens,
but use CM launch the game with your controller off and turn it on its active again,
using xbox pad here, but these differences to me are massive

 

[Michael Hornbuckle]

I'm sure having thirteen LOD's will be great and all, but a word from David about the physics errors would be great to hear.

 

[Andreas Knöpke]

Subtle as always