I use VR so I require constant 90 fps and on some parts of newer tracks (like the infield section of Daytona) fps drops to 70-80 fps and this is under ideal conditions (sun, 15 AI cars, 8 visible cars) and with lowest graphics settings. Completely impossible to run non-ideal conditions (sunset, rain, night, large grids of 32 cars). I think "track details" on "low" displays still far to many objects, I don't see a big difference between "low" and "full", visually and fps wise. The video below shows how Daytona looks on iracing with lowest detail. It is basically just the track and nothing else (this is what some top drivers run, to ensure max. performance in all situations). IMO it should be like this: "low" removes all unnecessary objects like grandstands, spectators, buildings, trackside objects and displays just the track like on iracing. "medium" should display grandstands (but no spectators), some buildings, some trackside objects and "high" should be what is currently called "low", which isn't low at all. This would remove many performance issues. https://www.twitch.tv/videos/1988105822
Beside the fact that you are right (I also noticed this on most tracks) have you tried the usual tweaks in e.g. json files like lowering distance of rendered objects in the rear mirror, texture override for opponent (and player itself maybe); also it helps a lot to increase performance when post fx and texture filtering are off or lowered and two major performance demaning options (to me) are road and environment reflections.
Hm ... I have to say sorry, but as mentioned here and there as a VR user you need to test and benchmark a lot as well as rF2 is not VR optimized. What GPU do you use?
RTX 2080 I don't know how much work it is or if there is some reason that I am not aware, but I really hope that someone form Studio 397 could reduce the number of track objects for newer, more demanding tracks, to give users who are struggling to get 90 fps a performance optimized version of those tracks like on iracing. There is no point for me buying tracks if I cannot use them.
I understand and basically agree, but this is one aspect / part of a never ending story concerning rFactor 2 and VR - there are a lot of sims out there which are much more optimized for VR (and even some which are worse) and a couple of games / sims let you reduce number of objects. I do not know what the devs had in mind concerning "track details", but I think the majority of players are taking this as "number of track environment objects", like me. BTW Daytona, Nürburgring GP and some other are the most performance demanding tracks (Daytona at the infield, Nürburgring at e.g. the wheel) Sorry, but I think I cannot help :-/
Yes, G2, full resolution. It tried lowering res but it looks blurry in the distance, so not an option. More surprisingly, lowering res, using fixed foveated rendering or reducing vertical FOV "work" visually, but don't have any effect on fps. I can run 640x480, so it looks like a game from 1990, but get the same fps as full res.
That's effectively what it means, yes. I suppose technically you could use it to swap out to lower detail versions of the same objects, but given that would mean adding more alternative objects to the track it probably isn't done that much. An interesting test would be modifying the SCN file via an upgrade and hiding many of the off-track objects via visgroups, to actually test what difference it makes.
Sounds interesting to me but I am not very familiar with modding - what would it make much more interesting, if someone could use the modified scn-file track version while joining a server with an unmodified one.
Yes, it is as if all pixels from full res are computed anyway, but only 640x480 of them are displayed.
Track details are controlled by each track's individual SCN file. Unfortunately the moddability of rF2 goes against it a bit here as a track maker can either choose to have very carefully assigned visibility groups for their track or make a track with for example zero visibility groups. In the latter case your track would look and perform identical on full or low track detail setting. I have seen many modded tracks that are poorly made in this sense and I'm not sure how much effort was put into Daytona even though it's official content. SCN also controls other things like level of detail, so a track maker can have very much control of how their track performs in rF2.
Hm ... I need to think it over because this sounds weird, but I am not sure if we are really able to lower resolution on a VR headset. FPS should increase if you are lowering SteamVR oversampling (I do not remember how it is called exactly ... if you set SteamVR to 50% or so resolution). There are a lot of people reporting major issues in performance while Windows hardware acceleration scheduler is switched on while using VR. I have switched that off after I encountered problems with tracking (lighthouse) and jitter ingame and that solved it. But if other applications are running fine in VR then it must be something pretty close to rF2 settings.
If you are CPU limited, which often is the case in VR, then lowering resolution will make next to no difference.
I don't use SteamVR, I use OpenXR Toolkit. Windows hardware acceleration scheduler is switched off. I really think the problem are the track details. In the oval part of Daytona I have constant 90 fps, only in the infield I get 70-80 fps and stuttering. If these track objects could be removed this would solve my problem I think.
@blagger I watched a bit of your stream now. I think you have too much details visible in mirrors like all the track distance markers. Go to player.JSON file and check that "rearview cull" is enabled. If not it will show everything in mirrors, which is unnecessary and wastes FPS. I reckon you could gain some FPS from this change.
You mean the twitch stream I linked in my first post? That is not me, it is just an example of iracing at Daytona at lowest track detail settings. I don't stream. Regarding "rearview cull", I set it to 100 and I also lower the resolution of the UI which also influences the resolutioin of the virtual mirror to 1024x768.
That's wrong. Sorry, I gave you the wrong value as well. The options for rearview cull must be either true or false, so like this: "rearview cull":true (no quotation mark set around true). If it's set to anything else it will probably not work at all.
