Gtx 1080 TI and VR performance

i like it very much and have no reasons to go further over it to other dd-wheels like osw or bodnar

 

What wheel did you have before? Was the accuforce a big jump in fidelity like they say it is?

 

have answered to pm - lets not clunk the thread with irrelevant discussion

 

Np .Now back to your scheduled program.

 

I have been doing some more testing using FCAT and the results are quite revealing about what the challenge is in keeping high frame rates.

These tests are all run with the Lotus Exege S at Imola using my base setting detailed in my signature below. The test involved starting position on the grid and what I did at the start.

This graph presents the results from three different runs.

a. Start 24 and don't move off grid.
b. Start 1 and maintain a 1st place position.
c. Start 24 and maintain a close 24 position (that is stay at the back of the pack
.
The results are most easily seen at the of the bottom chart showing synth frames in yellow and real frames in green.

Result A: At a start position of 24 the system is unable to provide frames on time and operates in ASW mode even before the race begins. As I sit still and the other cars move into the distance the system catches up and has no further problems.

Result B: At a start position of 1 there are no other cars in my view. I maintain a leading position throughout this test. The system operates in 90 fps and produces 0 synth frames during the 60 sec timeframe.

Result C: At a start position of 24 I drive closely at the back of the pack. There are other cars close in front of me throughout. The system operates in ASW modes 95% of the run.

I conclude that the number and closeness of cars IN VIEW has a very big impact on the ability of the system to maintain a full 90 fps.

It is very important to consider this when testing. Random starts like I was running in previous test will produce large amounts variability in results making drawing conclusions on other variable difficult. Running 1st places the least stress on the system and may hide import other factors. Running last is the most demanding on the system and will be the most discriminating in regard to things that impact performance although some variability will be caused by things that happen on track and effect the "view".

 

Since ac is still being updated I wonder if ac will be the first sim to use vrworks. Maybe one of the devs could comment on this if it really is as easy to add as NVidia says it is. I can only speak for myself but id guess many would take advantage of it and buy a second gpu if it allowed a premium vr experience.

 

@K_Soze
you have probably got this advice already but did you try to disable/enable windows game mode? You can also try to disable the xbox app.

 

I have tried this :
- disabling windows game mode
- disabling xbox app
- disabling DVR
- enable windows high performance power plan
- enable nVidia high performance
- remove GPU memory overclock
- stop all diagnostic services
- changing color range (not using full HD range)
- stopping Intel Turbo Boost

But the only thing that resolves the micro stutter problem is to revert to Windows 10 1607.

 

iRacing is using SMP (Simultaneous Mult Projection) for a while, so AC will definitely not be the first sim to use VRWorks...

 

In my last post I concluded that the most demanding test of the system is to start last and stay to the back. With 24 cars my system was not capable of producing 90 fps during an entire 60 second run. I therefore decided that to appropriately evaluate quality settings that impact system performance and display quality I needed to start in the back and stay in the back with fewer cars.

I selected 20 total as a good first step. Here is an FCAT graph representing the results. Once again Lotus Exege S at Imola. I collect 60 seconds worth of data which on Imola covers the turns up through Tosa. I am not totally consistent with when I start the data collection but within a few seconds each time.



In these graphs there is a consistent pattern of challenge to the system. At the start a lot of cars are bunched up in the display window, as they string out the system handles them better. As they bunch up in the turns the system is again challenged.

From a variable testing point of view this is an ideal frame work with which to test. The system can respond either positively or negatively to the variable by showing more or less synthesized frames.

I will be run my next test of test using 20 cars, starting at the back and stay at the back. These conditions will be my new base for testing.

 

More testing results using these base conditions: Lotus Exege X, Imola, Start 20, Stay 20

Base video settings in signature.

In these results I compare the base to the base except AA 2X from AA 4X and then base except OS at 1.3 from 1.4.

Results are average of four runs

Base-- Synth =499 (9%)
AA 2X-- Synth = 584 (11%)
OS 1.3-- Synth = 992 (18%)

The results show no benefit from making either of these changes (possibly worse on the OS change).

I conclude that what is limiting the ability of the system to deliver better performance under these condition is not related to either the AA or OS being set too high.

I did notice an increase in the jaggies at AA 2X.

Clearly it makes sense to leave these settings as is.

I don't know enough to guess what is going on in the video processing stream so I will just look for other variables that are impacting performance.

 

I am experimenting with 8X AA(1.4 PD) and post processing on the lowest. Image is very good. I can maintain 90FPS. I haven't tried a grid higher than 9, but this might be the way to go. Text is very clear.

Sent from my Nexus 5X using Tapatalk

 

Try the Nvidia 382.05 driver. My tests showed a clear improvement in performance.

The impact of grid size on performance is big, particularly if you start at the back of the grid. That being said there is no reason not to identify what the limits are with your system.

We all have to recognize that there are compromises we need to make between GPU cost, image quality settings , number of cars we run, and fps. We all have to choose based on what's important to us.

I am trying to learn how these variables interact so I can make intelligent choices for me.

 

More results. These are run on the impact of the setting World Detail on performance.

My base level is set at "very high". I tested "high" and "maximum" in comparison to my base.

Results-- Average of 4 runs

High-- Synch= 439 (8%)
Very High-- Synch = 499 (9%)
Maximum-- Synch= 871 (16%)

Graph of results


The system shows degradation in performance as the system works with high world detail.

I also noticed a distinct loss in detail on the "high" setting. There is a less obvious improvement in detail on "maximum". I would like to move to "maximum" but for now I will continue on "very high" and see if I can pick up extra performance with other settings.

 

With my setup I can't really go above SS 1.3, World detail very high (not maximum), shadow high (not ultra), reduce smoke and reflection low quality ultra refresh rate. But it is still amazing...
Almost the same setup by the way

i7 6700 4Ghx (OC 4.4 GHz), GTX 1080 TI (ASUS Rog Strix GTX 1080Ti OC) and 16GB ram (3000 MHz) with HTC Vive. Win 10 pro version 1607.
Fanatec CSW V2.0 wheel, CSP V2 pedals

 

I know are systems are just about identical except you run Vive and I run Rift.

Did you solve your problems with Chasing The Wind? Can you run it acceptably now?

 

It is still one of the more demanding Special Event but after the two first corners it runs very well. I may invest in 7invensun aGlass, it could improve by a factor 2 the performance in a few months

 

we can

Well lets hope they take advantage of it. Have you tried it in I racing?

 

Not recently. I remember that iRacing was one of the first (if not the first) game to implement SMP (20% fps increase) but it seems their implementation is only for triple screen setup and not VR.

 

I did an FCAT run on the Challenge, Chasing the Wind. K_Soze reported it to very demanding on the system.

Here is a graph of my results running at my base settings detailed below.

Dropped frames = 47
Synth frames = 1215 (23%)



It is very demanding on the video processing system.

Slightly worse system performance than the Lotus Exige S at Imola starting 20 and staying 20. The grid position is 13 of 15 with mixed competitors. Earlier testing showed that various cars created more burden on the system than others. It would be interesting to see what causes this situation to be so demanding.