2080Ti vs RTX 3090 for VR

[Gunnars Driver]

As I said. Would we improve in VR. ED gonna grab the engine by the Balls. Vulcan? Is this now in final stage? Or another game engine?

 

I think the lastest info is that they work on the move to Vulkan but it is not a small thing. Conlusion is that it won’t happen in [two weeks]. That is, be prepared to wait long for this.

 

So the average DCS gamer should calculate to need a CPU that handles single thread good.

[Icebeat]

:)

 

There’s only one way to sort out this forum issue —> ED go and get a quick fix for Vulkan :smilewink:

 

The problem is not Vulkan or how slow the engine render (actually it is not bad), my Index at 150% running on my 2080ti give me a stable 90 fps in single-player (install dogfight maps), but if I load some scripted map/mission or multiplayer, the game goes to 35-45, the gpu has nothing to do here because the units are not even visible. What is happening here is that the CPU is wasting time calculating whatever ED decided that needs to be calculated in the same thread. Why the game decided that the pathfinder of a unit miles away every cycle from me needs to be calculated in the same thread all the time, I don't know.

Edited October 7, 2020 by Icebeat

[Baldrick33]

The problem is not Vulkan or how slow the engine render (actually it is not bad), my Index at 150% running on my 2080ti give me a stable 90 fps in single-player (install dogfight maps), but if I load some scripted map/mission or multiplayer, the game goes to 35-45, the gpu has nothing to do here because the units are not even visible. What is happening here is that the CPU is wasting time calculating whatever ED decided that needs to be calculated in the same thread. Why the game decided that the pathfinder of a unit miles away from me needs to be calculated in the same thread all the time, I don't know.

VR is more CPU intensive than 2D though, which is why Vulkan should reap benefits for VR users.

[hansangb]

as a 3090 owner, i don't see very much improvement after all. i spend alot of time tweaking my settings. Iike expected, cpu issues. 4.9ghz 8600k, 32Gb ram at the moment. Rift S user.

 

some more 0.2 SS, shadows on medium, MSAA 2x. That's it. Don't expect too much. Dubai on PG looks good for the first time. Normandy.. still performance like the previous 2080 rtx, even without terrain shadows slide show.

 

That's my own experience for now..

 

 

You should follow up in the Rift thread. There is NO way 3090 struggles with Rift S when others have SS in Reverb.

 

I would take everything from the start and take it one at a time. Something is not right with your setup if the 3090 can't handle RiftS.

 

I"m sure the Rift dedicated thread will recommend deleting shaders, or clean installing GPU drivers, to making sure you're not double SS (PD and in Oculus), reducing shadows etc. But something is way off.

[imacken]

You should follow up in the Rift thread. There is NO way 3090 struggles with Rift S when others have SS in Reverb.

 

I would take everything from the start and take it one at a time. Something is not right with your setup if the 3090 can't handle RiftS.

 

I"m sure the Rift dedicated thread will recommend deleting shaders, or clean installing GPU drivers, to making sure you're not double SS (PD and in Oculus), reducing shadows etc. But something is way off.

 

To be fair, he's not saying it can't handle it, he's just saying he doesn't see a great improvement. As I said above, that is quite possible. He is able to have higher MSAA, SS, shadows etc., which is all good. I suspect - and I may very well be wrong - that he is measuring performance as fps. Now if he is using motion reprojection or whatever Oculus calls it, then he probably won't see an improvement, he'll be at the 40fps (or whatever is correct for the rift S) buffer.

[Csgo GE oh yeah]

Seems to fly in the face of anything else we've heard.

 

What SS rate are you running at?

 

I guess the Rift S only has a fraction of the pixels to push around compared to the Reverb - 3.6m compared to 9.3m, I think - and the benefits of the 3090 are in high pixel scenarios, e.g. 4K or Reverb at high SS rates.

 

Also, your 8600 is around 20-25% low compared to 10900K. That also may contribute.

 

Could all be part of the reason you're not seeing much difference. Who knows?

 

In gaming ? 25% ? I kind of doubt it especially if the 8600K is a little overclocked.

Look here :

[FritzTheGerman]

it's called AWS for oculus. I tried both. Locking to 40fps, AWS disabled and enabled. I have cleared fxo and shaders. There also no double SS. Don't get me wrong. As Imacken said, there is AN improvement. But not much or at least as i expected. Just want to share my experience with you, and especially to those who are not sure if willing to pay the extra money for a 3090 based on some pancake benchmarks on youtube (at the moment) . My rusty 8600k could be the reason. Who knows I'll keep you guys up to date and try to do some more of course

[imacken]

In gaming ? 25% ? I kind of doubt it especially if the 8600K is a little overclocked.

Look here :

 

We could discuss this for ever. There are benchmarks all over YT showing all sorts of different things. In single core benchmarks, the 10900K is about 20% higher than the 8600K.

 

There are 2 facts, firstly, I noticed quite a difference when I went from a 8700K to my current 10900K and secondly, I'm just saying the 8600K is a contributing factor. How much? Well, that is up for debate.

[FoxTwo]

In single core benchmarks, the 10900K is about 20% higher than the 8600K.

 

Clock for clock the 10900k is identical to the 8600k in IPC.

 

With both overclocked to the same frequency, their performance will be identical in single threaded use cases.

 

The 10900k should overclock better as it generally has better binning.

[speed-of-heat]

15 to 20% sounds about right...https://cpu.userbenchmark.com/Compare/Intel-Core-i5-8600K-vs-Intel-Core-i9-10900K/3941vs4071

[*Rage*]

15 to 20% sounds about right...https://cpu.userbenchmark.com/Compare/Intel-Core-i5-8600K-vs-Intel-Core-i9-10900K/3941vs4071

 

That link shows a 5% increase in single threaded performance going from 8600k to 10900k when both are overclocked.

[speed-of-heat]

True enough but there are more than two cores at work sure much of it is the Dcs main core, so I always go for two core perf, and you also need to factor in steam vr/wmr etc as well..

 

Also bear in mind the Base clock performance is way faster

[imacken]

That link shows a 5% increase in single threaded performance going from 8600k to 10900k when both are overclocked.

 

This is one of these 'it'll go nowhere' discussions, but all I'll say is that I noticed a big difference in DCS (and elsewhere) when I upgraded my 8700K to the 10900K.

[FritzTheGerman]

Once again i`ve deinstalled my nvidia driver, checked nvidia graphic setting and compared with VR optimation guides . erased content in fox folder and metashaders 2 folder, to make sure.

So i did some research and stopped using Oculus tray tool/debug tool. SteamVR and fpsVR in order to avoid double SS/pixel density. switched to default

 

Instead of, i tried the ingame pixel density slider, which i usually don`t use, to the value 1.8

(Pls if someone can tell how much steam resolution in % is? I am just to clumsy) Cockpit,gauges are crystal clear.

 

 

As i already said in a previous reply, Dubai with all the fancy skycrapers on very low level looks pretty nice now in a single solo mission flight.

40fps when i fly right between the buildings and look left and right out of the cockpit. Before not possible with 2080 (NO ti) High altitude over dubai even more than 40frames. Solid!

You can test by your own.

 

These are my settings. feel free to give me some advice for improvement.

 

Texture high

Terrain Textures high

Civ. Traffic low

Water medium

Visib Range high (cant feel a difference to med)

Shadows medium (personal preference but tested low and flat)

4x MSAA (0.70mask size), (again my pers. pref. to get rid of shimmering)

 

(rest is disabled as it should be)

 

Clutter/ Grass 0%

Tree visibility 50%

Preload radius 45000

Chimney Smoke 0

 

Terrain Object Shadows - flat

anisotropic filtering 8x

 

rest unchecked

 

-----------------------------------------

 

To get to the point, as expected multiplayer missions like through the inferno on persian gulf with heavy cpu usuage, are not playable with the settings above. i checked the cpu frame time with fpsVR. terrible! i tried normandy single missions. just fine and an absolute improvment. later some charnwood campaign in the P51, once again, heavy cpu usuage. slideshow. cpu bottleneck once again. Same with Caucasus. Thats why i said "don`t expect too much" even a 3090 will not solve the cpu bottleneck as many people think so.

 

 

Not sure if a beefy 10900k will help. It seems it needs more than that in my opinion. Other will tell when they tested their 3090 with the 10th gen intels.

Edited October 7, 2020 by FritzTheGerman

[speed-of-heat]

Shadows and msaa are killing your cpu frame times

[FritzTheGerman]

i know. spend hours of testing. there is not much difference without msaa off and shadows to flat. You need to lower drastically to the point where everything looks like before without 3090. It`s not what it should be.

[imacken]

Once again i`ve deinstalled my nvidia driver, checked nvidia graphic setting and compared with VR optimation guides . erased content in fox folder and metashaders 2 folder, to make sure.

So i did some research and stopped using Oculus tray tool/debug tool. SteamVR and fpsVR in order to avoid double SS/pixel density. switched to default

 

Instead of, i tried the ingame pixel density slider, which i usually don`t use, to the value 1.8

(Pls if someone can tell how much steam resolution in % is? I am just to clumsy) Cockpit,gauges are crystal clear.

 

 

As i already said in a previous reply, Dubai with all the fancy skycrapers on very low level looks pretty nice now in a single solo mission flight.

40fps when i fly right between the buildings and look left and right out of the cockpit. Before not possible with 2080 (NO ti) High altitude over dubai even more than 40frames. Solid!

You can test by your own.

 

These are my settings. feel free to give me some advice for improvement.

 

Texture high

Terrain Textures high

Civ. Traffic low

Water medium

Visib Range high (cant feel a difference to med)

Shadows medium (personal preference but tested low and flat)

4x MSAA (0.70mask size), (again my pers. pref. to get rid of shimmering)

 

(rest is disabled as it should be)

 

Clutter/ Grass 0%

Tree visibility 50%

Preload radius 45000

Chimney Smoke 0

 

Terrain Object Shadows - flat

anisotropic filtering 8x

 

rest unchecked

 

-----------------------------------------

 

To get to the point, as expected multiplayer missions like through the inferno on persian gulf with heavy cpu usuage, are not playable with the settings above. i checked the cpu frame time with fpsVR. terrible! i tried normandy single missions. just fine and an absolute improvment. later some charnwood campaign in the P51, once again, heavy cpu usuage. slideshow. cpu bottleneck once again. Same with Caucasus. Thats why i said "don`t expect too much" even a 3090 will not solve the cpu bottleneck as many people think so.

 

 

Not sure if a beefy 10900k will help. It seems it needs more than that in my opinion. Other will tell when they tested their 3090 with the 10th gen intels.

Are you seriously running at 1.8 PD? That is equivalent to 324% SteamVR SS!

[FritzTheGerman]

yes PD to 1.8 correct. even tried 400% in steam. of course just for the results of testing and fun.

 

i used to have 1.3 PD in Oculus tool for quite a while with my previous 2080. Can't see much different level of improvememt further than px 1.5

 

in comparison i spend some time to set ingame back to 1.0 pd or even 0.5 and used oculus pixel density and steam resolution to get a similar result in always the same cockpit in the P51 and F15.

Actually pretty impressive how this card performs

[imacken]

yes PD to 1.8 correct. even tried 400% in steam. of course just for the results of testing and fun.

 

i used to have 1.3 PD in Oculus tool for quite a while with my previous 2080. Can't see much different level of improvememt further than px 1.5

 

in comparison i spend some time to set ingame back to 1.0 pd or even 0.5 and used oculus pixel density and steam resolution to get a similar result in always the same cockpit in the P51 and F15.

Actually pretty impressive how this card performs

 

Just leave the in-game PD at 1.0, and vary SteamVR SS, but honestly, you are way over the top at 324%!

[speed-of-heat]

I missed the pd score... Good spot

[Harlikwin]

yes PD to 1.8 correct. even tried 400% in steam. of course just for the results of testing and fun.

 

i used to have 1.3 PD in Oculus tool for quite a while with my previous 2080. Can't see much different level of improvememt further than px 1.5

 

in comparison i spend some time to set ingame back to 1.0 pd or even 0.5 and used oculus pixel density and steam resolution to get a similar result in always the same cockpit in the P51 and F15.

Actually pretty impressive how this card performs

 

I used to run like 1.8PD for my CV1... Don't miss those days.

[Icebeat]

Shadows and msaa are killing your cpu frame times

 

MSAA and flat shadows are hardware techniques and shouldn't tax the cpu. Given that the GPU is a freaking monster, he should have this for free.

[speed-of-heat]

without shadows on medium (which is what he posted , though having tested other settings) my cpu frame time is somewhere between 6-8 in free flight, with shadows enabled you get spikes every 5th to 1/2 a second taking the cpu above 12ms and msaa increases my average GPU frame time by 2ms, SAAA is only a GPU hit

[RogueSqdn]

2080Ti vs RTX 3090 for VR

 

I have to do some research on this, but FalconNW says the 5950x should be available in around a month... I may be delaying my system for the new CPU.

 

I’ve been on the road all day so have to play catch-up with AMD’s announcement.

 

 

Sent from my iPhone using Tapatalk Pro

[Delareon]

Dunno if this Vid is allready postet, the Thread is quite long but take a look at this:

 

Basically FPS are stuck at 45, well there is an improvement but it makes me cry to see all those other games with different complexity but way better VR support....

So according to this Benchmark i wouldnt buy a 3090 if i allready had a 2080. performance increase is marginal for DCS that is.

For everything else in the Computer World the improvement is massive but in DCS we reached the possibilities of Hardware acceleration and will only get improvements if ED does this on the Softwareside.