2080 and 2080 Ti

[Wulfen]

Flying over forest on the Kuban map and testing differing setting I`m finding GPU-Z giving a GPU load of 60% with 40% overhead to spare, while the Oculus debug performance tool is telling me that I`m 40-50% under-powered to achieve 90fps while being stuck at 45-50fps. The cpu during all this is running between 40-50% load. Regardless of my setting at the time, they can`t both be right and performance is being squandered in the process.
 

Edited October 13, 2018 by Wulfen

[TUS_Samuel]

8 hours ago, Wulfen said:

The software seems not able to max out the GPU

Because CPU is bottleneck. 

 

[JonRedcorn]

8 hours ago, Wulfen said:

Flying over forest on the Kuban map and testing differing setting I`m finding GPU-Z giving a GPU load of 60% with 40% overhead to spare, while the Oculus debug performance tool is telling me that I`m 40-50% under-powered to achieve 90fps while being stuck at 45-50fps. The cpu during all this is running between 40-50% load. Regardless of my setting at the time, they can`t both be right and performance is being squandered in the process.
 

Somethings wrong with your setup those usage numbers are extremely low. I'd try a fresh install of windows or at the least using ddu to reinstall the drivers. 

[Wulfen]

4 hours ago, TUS_Samuel said:

Because CPU is bottleneck. 

 

I don`t think so, further tinker around I had the GPU at 98% load where the CPU was at ~35% load. The 4790K will not be a bottleneck at high res for the foreseeable future or even at double the GPU power. Regardless off the settings, at 100,150,200% steam options I was getting ~45fps give or take over the tree covered mountains at low level on Kuban. It looks like the game engine and resource use is the issue.

1 hour ago, 15th_JonRedcorn said:

Somethings wrong with your setup those usage numbers are extremely low. I'd try a fresh install of windows or at the least using ddu to reinstall the drivers. 

DDU done on uninstall of the old GPU. It`s just that section of Kuban is all over the place imo and game engine management.. 

Edited October 13, 2018 by Wulfen

[TUS_Samuel]

14 minutes ago, Wulfen said:

CPU was at ~35% load

 

On 8/17/2017 at 10:07 PM, TUS_Samuel said:

That's why you see only 30-40% total CPU usage, while FPS is low

 

lol

[BeastyBaiter]

If you only want 45 fps, then yeah, you can crank super sampling to 5,000,000,000% and create a GPU bottleneck. But if you're chasing 90 fps, you're going to be capped pretty hard by the CPU unless running some kind of crazy LN2 setup with a 7 GHz sustained all core overclock.

[A_radek]

Wulfen cpu really is the bottleneck in the case of il2. It uses one core fully, less than half another and throws a few farts at a third. Thus the 35-45% usage people are reporting. If you had a cpu with more than 4 cores you’d see even lower cpu usage. The thread maximizing a core fully is what limits framerate so in short it’s single thread cpu performance that limits us.

 

In Windows it will look as all cores are equally loaded, that’s because these threads are constantly jumping around between cores as just hammering one would cause unnecessary heat.

 

Ss loads gpu much more than cpu. So you can crank ss up til you see 100% gpu usage.

 

 

 

 

[Wulfen]

14 minutes ago, TUS_Samuel said:

 

 

lol

Thanks for your intellectual comment, it`s probably early in your part of the world. 

 

 

2 minutes ago, SvAF/F16_radek said:

Wulfen cpu really is the bottleneck in the case of il2. It uses one core fully, less than half another and throws a few farts at a third. Thus the 35-45% usage people are reporting. If you had a cpu with more than 4 cores you’d see even lower cpu usage. The thread maximizing a core fully is what limits framerate so in short it’s single thread cpu performance that limits us.

 

In Windows it will look as all cores are equally loaded, that’s because these threads are constantly jumping around between cores as just hammering one would cause unnecessary heat.

 

Ss loads gpu much more than cpu. So you can crank ss up til you see 100% gpu usage.

 

 

 

 

Yes as I said the resources at hand cannot be fully utilized by the game engine so performance is being lost. IL2 likely needs to go down the Vulcan route along with other optimizations.

[Alonzo]

1 hour ago, Wulfen said:

The 4790K will not be a bottleneck at high res for the foreseeable future or even at double the GPU power.

 

That's a very general statement and is just plain wrong for IL2. As you go on to say, the CPU absolutely *is* a limiting factor in IL2, depending on your map and settings and all that.

 

While I'm still hoping for an optimization pass, we have to work with the tools we've been given. At anything other than 'low' settings, you need to give IL2 massive amounts of CPU clock speed. If you're truly trying to get the most out of that GPU I would suggest using Balanced settings, Medium shadows, 4x AA and a crapload of SteamVR supersampling. You might also want to try things like the OpenComposite mod that bypasses SteamVR and saves some CPU cycles (only works on Rift).

[BlackMambaMowTin]

On 10/12/2018 at 1:18 PM, Alonzo said:

I upgraded from the 1070 to a 2080. Here are a few findings:

• On balanced graphics with 1.0 SS and no AA, performance improved from 82.4 FPS average to 85.3 FPS average on Chili's spitfire bombers track. This might seem like the 2080 isn't giving much, but...

 

Chili's spitfire bomber track is awful for me. No matter how I lower the graphics, it will average 45 fps for me on a 2080 ti and i7 6700K @ 4.3 Ghz. 

 

What CPU are you using? Rift or Vive?

 

I'm pulling my hair out trying to figure out why that bomber attack runs so poorly. If I just play the game on a mission, I get much better performance. 

[dburne]

1 hour ago, BlackMambaMowTin said:

 

Chili's spitfire bomber track is awful for me. No matter how I lower the graphics, it will average 45 fps for me on a 2080 ti and i7 6700K @ 4.3 Ghz. 

 

What CPU are you using? Rift or Vive?

 

I'm pulling my hair out trying to figure out why that bomber attack runs so poorly. If I just play the game on a mission, I get much better performance. 

 

That bomber track is in no way representative on what one would see playing the game - at least that is my perception.

I have flown many bomber intercept missions with multiple bombers to attack , in both PWCG and the Career Mode, and get way better performance actually flying them than what is given playing that track file.

 

I would advise anyone not to get hung up on those numbers with that bench.

I will take what I get playing the game, versus benching that track file any day of the week and twice on Sundays.

Edited October 13, 2018 by dburne

[Wulfen]

7 hours ago, Alonzo_ said:

 

That's a very general statement and is just plain wrong for IL2. As you go on to say, the CPU absolutely *is* a limiting factor in IL2, depending on your map and settings and all that.

 

While I'm still hoping for an optimization pass, we have to work with the tools we've been given. At anything other than 'low' settings, you need to give IL2 massive amounts of CPU clock speed. If you're truly trying to get the most out of that GPU I would suggest using Balanced settings, Medium shadows, 4x AA and a crapload of SteamVR supersampling. You might also want to try things like the OpenComposite mod that bypasses SteamVR and saves some CPU cycles (only works on Rift).

The major limiting factor is the ability of the game engine to utilize the potential of the CPU, not CPU itself.  Sure you can fire in a 9900K but the problem is not resolved and it will be even more hampered by the software pipeline. It`s a case of not seeing the wood for the tress.

Edited October 13, 2018 by Wulfen

[dburne]

2 minutes ago, Wulfen said:

The limiting factor is the ability of the game engine to utilize the potential of the CPU, not CPU itself.  Sure you can fire in a 9900K but the problem is not resolved and it will be even more hampered by the software pipeline. It`s a case of not seeing the wood for the tress.

 

Personally I think it is about time for the gaming companies  to start coding for what we have available to us and have had for some time now, multicores and hyperthreading.

I am sure that is no easy task and probably a crap load of work making changes to an existing game engine, but seems to me at some point it really needs to be done.

 

We have seen the last few years gains in pure speed on these CPU's have been very minimal, certainly nothing like what we were seeing like 10 to 20 years ago.

But there have been some nice improvements in multicores and hyperthreading, and in GPU's.

I am no computer or software engineer by any means, so may be just talking out my arse anyway.

[Wulfen]

56 minutes ago, dburne said:

 

Personally I think it is about time for the gaming companies  to start coding for what we have available to us and have had for some time now, multicores and hyperthreading.

I am sure that is no easy task and probably a crap load of work making changes to an existing game engine, but seems to me at some point it really needs to be done.

 

We have seen the last few years gains in pure speed on these CPU's have been very minimal, certainly nothing like what we were seeing like 10 to 20 years ago.

But there have been some nice improvements in multicores and hyperthreading, and in GPU's.

I am no computer or software engineer by any means, so may be just talking out my arse anyway.

Yes Moore's law has kind of hit the wall. Single core speed has improved ~15% since the 4790k, it`s a bit mad to say such a cpu is not up to the job when the software is such a handicap. Multi-core cpu`s have been on the scene for over a decade and game engines are still trying to carry water up a hill with the ladle instead of the bucket.

[Alonzo]

2 hours ago, BlackMambaMowTin said:

I'm pulling my hair out trying to figure out why that bomber attack runs so poorly. If I just play the game on a mission, I get much better performance. 

 

Lots of trees, flying near the ground. That Kuban map is pretty hard on performance, which is why it's good for a benchmark. I agree it doesn't represent the average in-game case, but if you're on the deck trying to shoot someone or not get shot you want the performance to be as good as possible there.

[dburne]

2 minutes ago, Alonzo_ said:

 

Lots of trees, flying near the ground. That Kuban map is pretty hard on performance, which is why it's good for a benchmark. I agree it doesn't represent the average in-game case, but if you're on the deck trying to shoot someone or not get shot you want the performance to be as good as possible there.

 

That is just it though, all I have  been flying for the past several months is PWCG campaign in the Spit over Kuban , and the same in the Career Mode.

I get way better performance in bomber intercept missions in those.

[chiliwili69]

On 10/6/2018 at 2:03 PM, VBF-12_Stick-95 said:

3DMark VRMark Blue Room actual user benchmarks:

i7-8700K @ 4.7, MSI 2080Ti - benchmark 5249, avg fps 114.43, target fps 109
i7-8700K @ 5.0, MSI 1080Ti - benchmark 3600, avg fps 78.48, target fps 109

 

I  couldn't find someone running a 8700K at 5.0 with the 2080Ti, though it may be out there. This 2080Ti comparison reflects an increase of ~46%

 

We can not (and must not) extrapolate VRMArk test results to IL-2VR.

Why?

because IL-2VR is currently limited by CPU (if you have 1080 or above)

 

My prediction is: If you have a 1080Ti using most "common or recommended settings" you are not going to increase you fps with a 2080Ti.

Why?

Because your 1080Ti is only running at 60% load. So your 2080Ti will run at 40-50% load, but will not increase your fps.

On 10/10/2018 at 5:43 PM, Alonzo_ said:

I suspect that VR will get accidental improvements from developers improving the main game, rather than special love and attention, because it *is* still a niche market.

 

Yes, VR users are still a small minority in IL-2. But it is a growing minority as you can see in this poll. This give us hope:

https://forum.il2sturmovik.com/topic/27278-about-vr-and-vr-devices-used/

 

I would say that VR is less than 5% of the total IL-2 users. But has not facts, just a guess.

 

In any case, VR is a very much loved feature by Jason and developers. They are very much proud of that as shown in Las Vegas recently.

Adapting to VR a game which was not designed for VR is a major effort that they implemented very well overall.

The only problem is the performance, so we need to decrease our settings depending our CPU/GPU.

I believe that offloading the heavy thread of IL-2 is a complex thing which will take a considerable time from developers. But I wish I wold be wrong.

On 10/12/2018 at 7:18 PM, Alonzo_ said:

On balanced graphics with 1.0 SS and no AA, performance improved from 82.4 FPS average to 85.3 FPS average on Chili's spitfire bombers track. This might seem like the 2080 isn't giving much, but

 

Even if it is still only 3fps gain, why do you think you get that gain? , with SS=1.0 your 1070 should not be loaded at all.

On 10/12/2018 at 7:18 PM, Alonzo_ said:

Applying in-game 4xAA on the 2080 it matches the 1070 performance at 82.0 FPS

 

Who does the work of in-game AA? the CPU or the GPU? 

[chiliwili69]

On 10/12/2018 at 7:18 PM, Alonzo_ said:

This gives ~83.7 fps on the demo track at about 80% CPU usage

Do you mean 80%GPU or 80%CPU?

If you mean 80%CPU, what tool do you use to measure that?

On 10/12/2018 at 8:03 PM, dburne said:

it should allow me to increase my settings

According to the Alonzo post the only settings you will benefit going from 1080Ti to 2080Ti will be SS. And a bit of AA.

Not all the other settings of IL-2.

[dburne]

16 minutes ago, chiliwili69 said:

 

According to the Alonzo post the only settings you will benefit going from 1080Ti to 2080Ti will be SS. And a bit of AA.

Not all the other settings of IL-2.

 

I shall see...

[Alonzo]

17 minutes ago, chiliwili69 said:

Do you mean 80%GPU or 80%CPU?

If you mean 80%CPU, what tool do you use to measure that?

 

Oh sorry, it was GPU. I'm not really measuring CPU these days, as you've pointed out it doesn't really show much (less than 50% "usage" on CPU, but we know the CPU is the bottleneck).

 

I believe AA is applied by the GPU, and it just takes one simple call from the CPU to enable it. Other effects such as Shadows are much more complex for the CPU because they are geometry/textures/calculations. AA is just a filter that applies to the whole frame.

50 minutes ago, chiliwili69 said:

Even if it is still only 3fps gain, why do you think you get that gain? , with SS=1.0 your 1070 should not be loaded at all.

 

I know the Oculus runtime does some fancy stuff when the CPU misses a frame submit. I am guessing that in these edge cases sometimes having the GPU able to crank out the frame really fast can help. So yes, both 1070 and 2080 are not loaded very much, but 2080 gives you a frame faster and therefore an improvement in those edge cases.

[VBF-12_Stick-95]

50 minutes ago, chiliwili69 said:

We can not (and must not) extrapolate VRMArk test results to IL-2VR.

Why?

because IL-2VR is currently limited by CPU (if you have 1080 or above)

 

We shall see...

?

[chiliwili69]

On 10/13/2018 at 2:08 PM, Wulfen said:

IL2 likely needs to go down the Vulcan route along with other optimizations.

 

Yes, we discovered that about a year ago. IL-2 VR is bottlenecked by CPU.

So, we will not benefit too much from new CPUs or new GPUs.

But the VR users in IL-2 is still a small minority.

I dream with the day that IL-2 is able to load all the cores of a Ryzen or a 9900K. But I am afraid this will not happen soon. So we try to get the most of the current hard/soft/settings

On 10/13/2018 at 10:07 PM, BlackMambaMowTin said:

I'm pulling my hair out trying to figure out why that bomber attack runs so poorly. If I just play the game on a mission, I get much better performance. 

 

Hey, I also really intrigued about your case.

Have you tried to run a mission in Kuban with the sun around

I really would like to help you here. Please, do the following:

 

- Tell me you Mobo brand and model

- Tell me your RAM speed

- Try to run the track with LOW settings and  all of the other settings at minimum

- Be sure you put ASW off

- What HMD device you use?

- Install MSI afterburner and enable some trends for GPU load, CPU frequency, CPU temp, GPU temp, etc while you run the track.

 

 

You can also try to run fraps in one of your missions and record that mission as a track. And then compare the fps of the live vs. recorded track.

On 10/13/2018 at 11:10 PM, dburne said:

That bomber track is in no way representative on what one would see playing the game - at least that is my perception.

I have flown many bomber intercept missions with multiple bombers to attack , in both PWCG and the Career Mode, and get way better performance actually flying them than what is given playing that track file.

 

Well in fact, the re-playing the recorded track should give you a bit more fps that the same flight in live game.

This was commented by a Sharpe in another thread. He said that FM calcs are not performed in a recorded track.

 

Looking from time to time the in-game fps counter is not an objetive way to measure performance.

Could you use an application like fpsVR and record a video of your mission? So we can learn how you achieve a better performance in similar missions. (Kuban, Sunset, Spit, clouds, etc like in the track).

 

 

[chiliwili69]

On 10/13/2018 at 11:10 PM, dburne said:

I would advise anyone not to get hung up on those numbers with that bench.

I will take what I get playing the game, versus benching that track file any day of the week and twice on Sundays.

 

Just remind that the benchmark is just a measuring tool with two objetives in mind:

- Study how every Hardware/software/settings is affecting the performance of our most loved IL-2VR

- Compare the performance with your peers and see if they are aligned, if not, something wrong could be in your PC.

 

Apart from that, the IMPORTANT THING is your subjective feeling with the game, not what you obtain in the benchmark.

If a person is happy running at 30fps in VR, that´s perfectly fine.

40 minutes ago, VBF-12_Stick-95 said:

We shall see...

 

We already have seen it with the Alonzo test with the 2080 versus 1070.

[dburne]

1 hour ago, chiliwili69 said:

 

Well in fact, the re-playing the recorded track should give you a bit more fps that the same flight in live game.

This was commented by a Sharpe in another thread. He said that FM calcs are not performed in a recorded track.

 

Looking from time to time the in-game fps counter is not an objetive way to measure performance.

Could you use an application like fpsVR and record a video of your mission? So we can learn how you achieve a better performance in similar missions. (Kuban, Sunset, Spit, clouds, etc like in the track).

 

 

 

No, I get way better performance in the game rather than playing back that track.

Sharpe does not even fly in VR.

The track does not even utilize the GPU near as much as flying a mission. All one has to do is monitor their GPU to see this.

 

Not really interested, I just play the game and have fun.

 

How I measure my performance?

If the game is smooth and stutter free for me and very fluid throughout my missions  it is more than acceptable. What else would I want.

Currently flying a PWCG campaign with lots of activity, over Kuban in the Spit at various times of day, and having a blast.

 

 

 

1 hour ago, chiliwili69 said:

 

We already have seen it with the Alonzo test with the 2080 versus 1070.

 

So that sets the standard for everyone?

Perhaps not all systems are created equal...

Edited October 15, 2018 by dburne

[chiliwili69]

30 minutes ago, dburne said:

No, I get way better performance in the game rather than playing back that track.

 

I remember that Carl Sagan said: "Extraordinary claims require extraordinary evidence"

 

 

But, forget this and have fun!

 

 

[BlackMambaMowTin]

So, I solved my stuttering problem by changing the processor affinity for all my Oculus software like OculusDash.exe and OSVRService_64.exe to my last 4 cores.

 

My performance is still bottlenecked by my CPU. So my 2080 ti has not improved performance much but it seems I can increase settings a bit without losing much performance. I'll have to test more to see. 

[dburne]

8 hours ago, chiliwili69 said:

 

But, forget this and have fun!

 

 

 

[Alonzo]

For anyone looking to upgrade for VR, you might consider the RTX 2070 now. It has about the same performance as the 1080, which is about the sweet spot for getting performance out of IL2 (since we are usually CPU bound). Here's a review:

 

 

[Guest deleted@134347]

On 10/15/2018 at 12:56 AM, BlackMambaMowTin said:

So, I solved my stuttering problem by changing the processor affinity for all my Oculus software like OculusDash.exe and OSVRService_64.exe to my last 4 cores.

 

My performance is still bottlenecked by my CPU. So my 2080 ti has not improved performance much but it seems I can increase settings a bit without losing much performance. I'll have to test more to see. 

 

did you also set the affinity on IL2.exe to a core that was not assigned to Oculus?

[BlackMambaMowTin]

7 hours ago, moosya said:

 

did you also set the affinity on IL2.exe to a core that was not assigned to Oculus?

 

No, I assumed it was using more of the first first cores. 

 

I'm planning to download Processor Lasso to manage this. 

 

Here's my theory and I could be totally wrong. Oculus Home is probably smart enough to run games on different CPU cores than those it uses for background process. But if I launch SteamVR, SteamVR will be treated as a game but any game SteamVR launches can end up running on any core. Or worse, SteamVR launches it on cores that SteamVR is not using. 

[Guest deleted@134347]

12 hours ago, BlackMambaMowTin said:

 

No, I assumed it was using more of the first first cores. 

 

I'm planning to download Processor Lasso to manage this. 

 

Here's my theory and I could be totally wrong. Oculus Home is probably smart enough to run games on different CPU cores than those it uses for background process. But if I launch SteamVR, SteamVR will be treated as a game but any game SteamVR launches can end up running on any core. Or worse, SteamVR launches it on cores that SteamVR is not using. 

 

 

thanks for the hint on 'process lasso', i'll give it a try on my setup with vivepro... 

 

As far as the game affinity itself steamVR does launch il2.exe as a separate process, hence I presume you can change its affinity after the start.... hopefully...

Edited October 17, 2018 by moosya

[dburne]

EVGA now has RTX 2070 Series Cards up for pre-order:

https://www.evga.com/products/productlist.aspx?type=0&family=GeForce+20+Series+Family&chipset=RTX+2070

Edited October 17, 2018 by dburne

[Guest deleted@134347]

30 minutes ago, dburne said:

EVGA now has RTX 2070 Series Cards up for pre-order:

https://www.evga.com/products/productlist.aspx?type=0&family=GeForce+20+Series+Family&chipset=RTX+2070

 

the easiest way to get them is via nowinstock.net  .. Go to below link, sign up for alerts (e-mail, desktop and/or sms) and just wait for it to pop up. But be ready with a prelogged-in newegg or amazon account   stuff is gone in 4 minutes usually..

 

http://www.nowinstock.net/computers/videocards/nvidia/rtx2080ti/

 

 

although I'm still on the fence about 2080ti. Very hard to justify going from 1080ti to a $1,200 expense for some misunderstood gains.

 

 

[kevman]

2 hours ago, moosya said:

although I'm still on the fence about 2080ti. Very hard to justify going from 1080ti to a $1,200 expense for some misunderstood gains.

 

 

I hear you, I am fully on the fence here - some of the VR gains on assetto corsa and dirt rally look like they are quite worth while.

Looking on the forums i think there will be some improvement in DCS but limited improvements in IL2 (mainly SS) . 

Everything would get a 20-30 fps bump on my 3440x1440 120 Hz G-sync but most games I play are over 60 fps on the 1080ti so would I really notice much difference

 

Part of me thinks just go for it it could be 2.5 years until the 3080 ti 

 

[Guest deleted@134347]

9 minutes ago, kevman said:

 

I hear you, I am fully on the fence here - some of the VR gains on assetto corsa and dirt rally look like they are quite worth while.

Looking on the forums i think there will be some improvement in DCS but limited improvements in IL2 (mainly SS) . 

Everything would get a 20-30 fps bump on my 3440x1440 120 Hz G-sync but most games I play are over 60 fps on the 1080ti so would I really notice much difference

 

Part of me thinks just go for it it could be 2.5 years until the 3080 ti 

 

 

i get the rationale.. but I stopped playing Dirt Rally and AC doesn't really appeal to me.    the DR 2.0 announcement brought some excitement in to my life but apparently Codies turned in to another cross platform console port with no VR support so my hopes were crushed. I don't play DCS due to lack of 'stuff to do' with their WWII birds ... so Il2 is really the only VR sim I enjoy.  I hope next year there'll be a real motivation to get 2080ti.... right now I want it.. but I don't need it...

[kevman]

5 minutes ago, moosya said:

 

i get the rationale.. but I stopped playing Dirt Rally and AC doesn't really appeal to me.    the DR 2.0 announcement brought some excitement in to my life but apparently Codies turned in to another cross platform console port with no VR support so my hopes were crushed. I don't play DCS due to lack of 'stuff to do' with their WWII birds ... so Il2 is really the only VR sim I enjoy.  I hope next year there'll be a real motivation to get 2080ti.... right now I want it.. but I don't need it...

 

Sounds like it is a bit of a tough sell for you if your main game is IL2 - just spoke to my missus she says just buy it its me who is not so sure?

[von_Tom]

14 hours ago, kevman said:

 

just spoke to my missus she says just buy it

 

 

OMG that is surreal.  Hands up if you have to hide the receipts or at least fudge how much IT costs!

 

*Hand up*

 

von Tom

[kevman]

Nvidia FE 2080 ti is now in stock in the UK. 

[dburne]

Hoping my EVGA RTX 2080 Ti FTW3 I have had on pre-order  ships today or tomorrow.

I am not in a big rush but would like to finally have in hand.

[kevman]

3 minutes ago, dburne said:

Hoping my EVGA RTX 2080 Ti FTW3 I have had on pre-order  ships today or tomorrow.

I am not in a big rush but would like to finally have in hand.

Here's hoping. I have evga card on pre-order with the rainforest so have not been charged yet .

Evga have been a bit slow on the uptake this time round conpared to some other aib partners - should be worth the wait though

Edited October 18, 2018 by kevman