Triple screen FPS is a lot higher than VR FPS!

[Bumpety Boo]

Hi everyone!
TLDR (mostly) under the __________ line
I Recently purchased everything I needed for VR;
Lenovo explorer VR headset (1440x1440 per eye) / GTX 1070 / Intel I3-8100 / 16GB ram

I'm running average to low display settings: Anisotropic filtering 8x | anti aliasing 2x | world detail low | shadow low | no PP effects (minimum preset does not really help much with brake lights for some reason. anyway,)
I'm getting 70-90 FPS during most races, except when the grid is packed (start of a race), or when racing Nordschleife, then it's getting to as low as 45 FPS.

I just figured I got three 19" screens so I can give triple screens a go (last time I tried it was on my previous GTX 660, but back then Nvidia demanded all your screens to be the same model! same old monitors, several years later - this is no longer an issue )
Entering the first race and I am greeted with 180 FPS and over!

__________


my screens are (1280 x 1024) x 3 = 3,932,160 pixels, this is quite close to the Lenovo explorer resolution (1440 x 1440) x 2 = 4,147,200. In order to get even closer I upped the triple screen resolution in assetto corsa to 4190x1024 = 4,290,560 (but I can't be sure if it made any difference). loading Nordschleife with 23 AIs (or running online) and it never drops from 85, usually it's at 90.
So basically, FPS is about doubled in triple screen than in VR.
does that make any sense? I began wondering if WMR calculating your head position from the two front facing cameras - is a CPU intensive process, because otherwise I can't imagine why there's such a huge difference! Am I missing something obvious here? Anybody else get similar results, or the opposite?

 

[Stereo]

1440x1440 per eye - meaning it's twice that resolution per frame (since it draws both eyes at 90fps). Explains half the fps pretty easily.

 

[Bumpety Boo]

Mind you that I initially wrote 1440 x 1440 = 4,147,200, but I should have been more specific
(1440 x 1440) x 2 = 4,147,200, so I did accounted for the VR's double screen overall pixel count. and it is very close to the triple screen resolution I ran.

 

[Insert Coin]

On a triple screen the world is rendered from one viewpoint (monoscopic), in VR from two viewpoints (stereoscopic: eye distance), so everything has to be rendered twice instead of once.

 

[RasmusP]

Get openhardwaremonitor (little free standalone exe), play assetto corsa in VR and with triples, check your graphics card load.
If the gpu load isn't higher than 95%, it's your cpu. If your gpu load is above 95%, it's your graphics card. Easy

 

[Emery]

Are you in Multiview (angled monitors) or treating the 3 screens as a single wide monitor?

 

[Kek700]

Something does not add up. I have a gtx980ti which i have always assumed is “about” equal to
a 1070 also a clocked i5 6600k 4.5Ghz not sure were that stands compared to late 8100 i3.
My setting are the same, but with a full field 24 AI on silverstone i get 100fps on start line.
Running triples 5760x1080 appro 6.25 M pixels. So something does not make sense.
Also once race has started it quite often runs at 144 hz
Maybe wrong in the above assumption.

 

[Bumpety Boo]

On a triple screen the world is rendered from one viewpoint (monoscopic), in VR from two viewpoints (stereoscopic: eye distance), so everything has to be rendered twice instead of once.

yeah, you do render two images that are skewed, but I'm not sure if there are more calculations involved in rendering two half-resolution dissimilar images to rendering one full resolution continuous images. GPU stores the same graphics dataset. maybe the CPU has to crunch more objects geometry?

Are you in Multiview (angled monitors) or treating the 3 screens as a single wide monitor?

Nvidia calls their technology PhysX, and they span 3 monitors as a single monitor. windows then sees one 3840 x 1024 display. assetto corsa lets you set the angle between monitors as well as other dimensions of your set up. (Are there other ways around it?)

Something does not add up. I have a gtx980ti which i have always assumed is “about” equal to
a 1070 also a clocked i5 6600k 4.5Ghz not sure were that stands compared to late 8100 i3.
My setting are the same, but with a full field 24 AI on silverstone i get 100fps on start line.
Running triples 5760x1080 appro 6.25 M pixels. So something does not make sense.
Also once race has started it quite often runs at 144 hz
Maybe wrong in the above assumption.

yes, the 980ti and 1070 are almost identical. the processors are pretty comparable in performance with yours scoring 9%-25% better in different tests. also your test results are pretty close to mine - I'm getting 180fps when I'm not on the huge Nordschleife racetrack, and you're getting 100 while using higher resolution. So, what doesn't add up?

 

[Emery]

assetto corsa lets you set the angle between monitors as well as other dimensions of your set up.

Right. However, if one doesn't tell AC that the monitors are angled, then only a single viewport is calculated by AC rather than 3 viewports and there's a difference in performance. Thus my question of which configuration was used.