Nvidia 5080 in VR - initial thoughts

[Mr_sukebe]

My 5080 arrived today and is now installed, having replaced a 3080Ti.  The rest of the PC includes a 7800x3d, 64GB ram, PCIE5 SSD.

My VR headset is an Oculus Pro, which does have eye tracking.  As such, I'm using that rather awesome Quad Views from Mbuchia, which most certainly does help.  My drivers are 572.16, using the latest DLSS (K?).  The DLSS setting is for quality upsampling and DLAA AA.

The hope was to achieve an FPS of 72, which is the native frame rate for the Oculus Pro, and in short, with the 5080, it nearly consistently hits it.

Settings included:

• High Textures
• High scenery textures
• Ultra Clouds (lower settings look a real mess with DLSS upsampling)
• scenery detail, forests around 80% across
• Flat shadows

In the Oculus, I have the "Home" disabled, and in DCS, I have the empty hangar, both of which save VRAM space.

I tried a variety of aircraft including the Viggen, Harrier, F4, F18, Apache and Chinook (not the mod). For terrains, I used Kola, Afghanistan and Falklands.

With most of the aircraft/terrain combinations, the GPU was running between 60-80% usage, even after upping the up-sampling in Quad Views to x1.5.  The frame did visibly bounce around a little, but in most combos, it was very smooth.  

VRAM usage was around 14GB+, so not much headroom.  The positive news was that with nearly every combo, once the shaders were built for all but the "odd" new texture, that the blips barely registered whilst it was loading.  My guess is that maybe data bus all being at PCIE5 "might" helped.

 

The caveats being:

• Add the Chinook showed up the implications of only 16GB of VRAM, which resulted in an absolute tanking of frame rate, whilst DCS was dragging in what are clearly "fairly large" textures for the aircraft.  As that was the only aircraft that I experienced issues with (and that was on the SA map), that's fairly positive.  Is it enough to justify the additional spondoolies for a 5090, well that's not for me to answer
• This was single player.  Haven't tried multi-player, so these findings "might" not be fully representative

In short, I'm pretty impressed.  I wasn't sure that it was going to deliver the hoped for 72fps, but in the majority of my tests, it happily managed it.

As for power consumption, turns out that it's very close to my old 3080Ti, meaning that I frankly don't really care.

Hope that helps.

 

[Qcumber]

Firstly, congratulations on managing to get a 5080. Secondly, it's encouraging to see that you get good results with the Q/OP. I have just upgraded to a 9800x3d and have a 5070ti arriving next week (hopefully).

Could you please share your QVFR settings? 

[Mr_sukebe]

Currently running with .22 * .22 and x1.5

[Qcumber]

8 hours ago, Mr_sukebe said:

Currently running with .22 * .22 and x1.5

What is your peripheral resolution? And what is your base setting in oculus link? Are you at the x1.3 setting (2745 pc)? That would make a centre res of equivalent to about 4100 px!?

[Mr_sukebe]

1 hour ago, Qcumber said:

What is your peripheral resolution? And what is your base setting in oculus link? Are you at the x1.3 setting (2745 pc)? That would make a centre res of equivalent to about 4100 px!?

Sorry, can't remember offhand, but I'll check later.

My assumption is that the default settings assume we'll be using fixed fov rendering.  As we have eye tracking, I've been a lot more aggressive, so the periphery multiplier is pretty low.  x1.5 is the upscaling for I assume the central area.

[RoboHackfeld]

Good to know, the 5080 gives us nearly what we've expected. I am stil waiting for a good offer for a 5080 but it will be definately my next GPU. 

 

It would be nice, if you could share some more Settings you are using. 

 

PD in Meta Link,

PD, Shadows and AA in DCS, DLSS (4.0?),

your QVF settings for Oculus 

[Oculus]
peripheral_multiplier=0.2
focus_multiplier=1.0
horizontal_focus_section=0.55
vertical_focus_section=0.55

and maybe how you  setup your Oculus Tray Tool/ Debug Tool

 

 

Thanks in advance and I wish you a lot of fun with your new GPU.

 

BG Robin
 

 

[Qcumber]

3 hours ago, Mr_sukebe said:

Sorry, can't remember offhand, but I'll check later.

My assumption is that the default settings assume we'll be using fixed fov rendering.  As we have eye tracking, I've been a lot more aggressive, so the periphery multiplier is pretty low.  x1.5 is the upscaling for I assume the central area.

Yes. I have used some aggressive settings too. With DLSS 4 the lower resolution areas are much less noticeable. Pushing the centre resolution high also helps reduce DLSS artifacts. 

[Mr_sukebe]

My settings (updated) are:

Oculus Headset: Via USB3 data cable, Anker powered USB hub to PC

OpenXR

Oculus HomePage - disabled

Oculus tray Tool: 

x1.3 supersampling

Res: 3905

Link Bandwidth: 960

Quad Views:

peripheral_multiplier=0.3
focus_multiplier=1
horizontal_focus_section=0.22
vertical_focus_section=0.22

DCS:

Using Empty VR Hangar mod to save VRAM

(see image attached)

 

Edited March 7 by Mr_sukebe

[diamond26]

6 minutes ago, Mr_sukebe said:

Oculus tray Tool:  Not used as Quad views makes it redundant

Why you say it’s redundant? OTT is just an overlay of ODT (with some helpful functions) which you use by default when you connect through Link

Also you mentioned 14gb+ of VRAM use. Which tool you use to get this report ? I haven’t seen more than 12gb in my 16gb 4090 mobile one (in CapframeX or build in metrics of DCS). Is that for all modules or specific cases ? 

[Mr_sukebe]

OTT - I only really used it for upsampling.  As this can be done in Quad views, for myself it was redundant.  If you wish to use, that’s cool

 

VRAM usage:

You can see my settings.  Maybe ours are different.  Whilst checking they were reasonably complex missions, but I’d expect that to mainly affect CPU usage.

Either way, if you’re using less, we’ll done

[Aapje]

@Mr_sukebe

Shouldn't you update or remove your signature? It looks like it is two upgrades out of date now.

[Mr_sukebe]

28 minutes ago, Aapje said:

@Mr_sukebe

Shouldn't you update or remove your signature? It looks like it is two upgrades out of date now.

Thanks.  I hadn't even realised, as I don't get shown my own sig.

[RoboHackfeld]

vor 7 Stunden schrieb Mr_sukebe:

OTT - I only really used it for upsampling.  As this can be done in Quad views, for myself it was redundant.  If you wish to use, that’s cool

 

VRAM usage:

You can see my settings.  Maybe ours are different.  Whilst checking they were reasonably complex missions, but I’d expect that to mainly affect CPU usage.

Either way, if you’re using less, we’ll done

U can use it to reduce your FoV to save some performance. 

[Mr_sukebe]

9 hours ago, diamond26 said:

Why you say it’s redundant? OTT is just an overlay of ODT (with some helpful functions) which you use by default when you connect through Link

Also you mentioned 14gb+ of VRAM use. Which tool you use to get this report ? I haven’t seen more than 12gb in my 16gb 4090 mobile one (in CapframeX or build in metrics of DCS). Is that for all modules or specific cases ? 

 

Having played more this evening, I'll very happily take back my comment about OTT being redundant.

I tried a variety of things tonight, and settled on using OTT to manage the Link Bandwidth, Encoding Res, and multiplier

Had a meander about in Normandy and over Dubai, solid 72 frame rates on both, despite the GPU not being maxed out.

 

For awareness, I'm still getting micro-stutters even with a solid 72fps.  From a video I watched recently, my hope is that the move to Vulkan will assist.

Edited March 7 by Mr_sukebe

[Qcumber]

5 hours ago, Mr_sukebe said:

For awareness, I'm still getting micro-stutters even with a solid 72fps.  From a video I watched recently, my hope is that the move to Vulkan will assist.

When you say "solid 72fps" what do your frame times look like? With a good stable fps they should be essentially flat lines. If you see lots of variations or sudden spikes then this could be the issue. You say that your GPU is not maxed out but what about VRAM and CPU load? Are you getting an even load across all cores?  

[Mr_sukebe]

Not completely equal CPU use across all cores, but none maxed out.

VRAM varied. Whilst over Normandy in a 190A8, it was 12GB.  In the Gulf, over Dubai it was full.  To give me some breathing room on VRAM, I’m going to put textures back at medium.  In VR, the visual hit is not huge.

I’ll check frame times later today.

[RoboHackfeld]

Dont know exactly if it helps too free up some VRam but reducing the resolution for your Desktop DCS Window seems to be worth a try. 

null

[Mr_sukebe]

Just been for a meander around the Falklands in a harrier.  Textures high.  From the data displayed by DCS, it had reserved 15GB of VRAM, and was using under 7GB, that's promising, though being fair, it was a very simple scenario with not much there.

Frame time was mostly 13, sometimes going into 14.

Task manager showing GPU usage around 80% at a pretty much continuous 72fps.

 

Have just updated my Encode Resolution.  3905 was NOT correct.  Should have been 3840.  Changing that has improved definition.  Have also disabled DLSS upsampling, whilst using DLAA. Still getting a solid 72fps.  OK, think I'm good at that.

Edited March 8 by Mr_sukebe

[Qcumber]

7 hours ago, Mr_sukebe said:

Frame time was mostly 13, sometimes going into 14

The max frame time for 72 fps is 1000/72 = 13.9ms. 13 ms is just below what is needed to maintain 72 fps. 14 ms is just above. 

Even though your GPU usage is only 80% it looks like you might be at the borderline for maintaining 72 fps. 

 

Edited March 8 by Qcumber

[RoboHackfeld]

Am 8.3.2025 um 14:14 schrieb Qcumber:

The max frame time for 72 fps is 1000/72 = 13.9ms. 13 ms is just below what is needed to maintain 72 fps. 14 ms is just above. 

 

 

 

So just tu clarify, I am still figuring out how everything is connected.

If I have to maintain 80fps I need to get a frametime lower than 1000/80=12.5ms?

If I need to get 90 fps this one is like 1000/90=11.11ms. Correct? Right now I am playing with 80hz and SSW(40fps) so like 1000/40=25ms should be enough to get stable 40fps.

 

The thing is, I am always switching between SSW and NonSSW to see what FPS I get and if it is stable above my 40fps for SSW but if I know my frametime just needs to stay below 25ms, better 20ms I could easily work with that than always switching between SSW and nonSSW. 

 

[DCoffey]

Just got my 5080 and vr is super smooth 

[Blackhawk163]

Just got my 5080 ahead of the tariffs and it’s a middling upgrade for me in VR. Pretty much the same as the 4080 super I have. Might just send it back and just wait for the 6000 series or if AMD decides to step back into the premium space again. 
 

Really wanted a 5090, but it seems that Nvidia just isn’t interested in selling any to us consumers atm. 

[Tensorial_Architect]

It is not that Nvidia is not interested in selling 5090s, ... they are VERY happy to rape wealthy American consumers for their flagship; The problem is that TSMC cannot produce the GB202 in moderate to high yields. Sadly, Blackhawk, I would also describe the RTX 5090 as, ... middling. Go for an RTX 4090, ... still the best high-end sweet spot for DCS.

Edited April 5 by Tensorial_Architect

[Tensorial_Architect]

Mr Sukebe, when opportunity presents, would you mind showing us a few pics of your setup? I'd love to see your Moza, Jetseat, and general layout.

[Hempstead]

I have the Unobtainium, an RTX 5080 card. But I don't have a computer to put it in.... yet.  Still waiting to get my hands on a Ryzen 9950X3D... *sigh*.