P3Dv4.3 : my observations

[GSalden 4,656]

Hi Michael,

With LC scenery I normally use High and with photo real scenery Ultra as that is rendered lower at each setting .

To see if the total settings that put load on the graphics card is a cause I have the opportunity to test with a 1080 next to my 1080 Ti. The 1080 Ti will drive the 2 monitor NVSurround monitors and the 1080 the left one.

As I always use Shadows quality Low but Shadows distance High it might help.

I have to say that my 1080 Ti almost never goes above 90% load , but who knows.

The only thing I do not know is if I can set the clockspeeds for each card differently ( Eg : 1080 Ti : OC 1926 MHz ) ?

[GSalden 4,656]

Just read that to avoid stutters it might be better to run SLI ( Pete D’s experience ) That might take a while as I have no one near with another 1080 Ti.

Anyone tried SLI and what is your experience with FPS regarding  lots of clouds / high cloudshadows radius ?

[djbully 210]

There some info and test results here: https://forums.flightsimlabs.com/index.php?/topic/18460-p3d-v43-4k-observations/

If you have access? (FSL A320) 

[AnkH 2,233]

Basically it is very easy to "test" if more graphic performance would provide you with better sim performance. Simply check your GPU load. If it is anything above 90%, adding another 1080Ti might improve the situation. If your GPU load ist mostly under 90% (also in cloudy conditions), adding another 1080Ti would help nothing. Vice versa, the "resolution test" should also work in P3Dv4.x: if you get more FPS with 720p compared to your normally used resolution and you get considerably higher FPS (not just 5-10% more), adding more GPU power will help. If the FPS stay roughly the same, adding more GPU will do nothing...

[Guest]

1 hour ago, GSalden said:

Just read that to avoid stutters it might be better to run SLI ( Pete D’s experience ) That might take a while as I have no one near with another

I thought Pete uses ProSim737 and full cockpit with no VC?  Not rendering a VC should help performance regardless of multi-GPU.

Cheers Rob

[GSalden 4,656]

1 hour ago, Rob Ainscough said:

I thought Pete uses ProSim737 and full cockpit with no VC?  Not rendering a VC should help performance regardless of multi-GPU.

Cheers Rob

Pete uses Prosim like me.

His server has 2 graphics cards in SLI. One of them is connected to his 3 projectors in 3 seperate views.

After all tests and reinstall I feel that my issue is the  / 3 drops in the Bucket - 2 drops out the Bucket / issue because of my settings icw 3 screens and the continuesly evolvement of P3Dv4.

As I can fly very well over Orbx TU Netherlands when using Fair Waither - no ASP4 - Cloud DD 50 nm - no shadows , I was thinking that SLI might help ( clouds and shadows ) ...

[Andreas Stangenes 1,628]

My observations in regards to VR is that native VR in p3d v4.3 is very improved. The fact that you can combine it with HDR is also a big benefit. In previous iterations of p3d I found VR mostly a gimmick. Now it is quite solid, and it plays together with several good quality camera addons (like chaseplane) making it my new preferred way of VR'ing in flightsim. 

P3D v4.3 is rock solid for me.

[GSalden 4,656]

3 hours ago, AnkH said:

Basically it is very easy to "test" if more graphic performance would provide you with better sim performance. Simply check your GPU load. If it is anything above 90%, adding another 1080Ti might improve the situation. If your GPU load ist mostly under 90% (also in cloudy conditions), adding another 1080Ti would help nothing. Vice versa, the "resolution test" should also work in P3Dv4.x: if you get more FPS with 720p compared to your normally used resolution and you get considerably higher FPS (not just 5-10% more), adding more GPU power will help. If the FPS stay roughly the same, adding more GPU will do nothing...

My graphics card is always 50-90% , mostly 50-80%.

Pete wrote that his smoothness improved with SLI . Even with 1 graphics card not loaded above 90%. Now the GPU loaded are around 20% most of the time....

[Chris Bell 165]

Gerard in [GRAPHICS] section find
TERRAIN_SHADOWS_RECEIVE=1 
change to
TERRAIN_SHADOWS_RECEIVE=0

[GSalden 4,656]

11 minutes ago, Chris Bell said:

Gerard in [GRAPHICS] section find
TERRAIN_SHADOWS_RECEIVE=1 
change to
TERRAIN_SHADOWS_RECEIVE=0

Then I will have no cloudshadows ...

[Chris Bell 165]

you should still get cloud shadow with no blurries 

beside... you're scenery is too realistic!
you dont need it anyway

[Guest]

51 minutes ago, GSalden said:

Now the GPU loaded are around 20% most of the time....

That's actually not good IMHO ... but perhaps the tools measuring load aren't accurate?  The objective with SLI and processing overall, is to move as much as one can onto the GPU.  So if the single GPU is around 90% under heavy load, then adding a 2nd GPU with SLI "surround" mode (if that is the mode being used) should NOT drop load to 20%, that's doesn't really add up?  Even if using DX12 with EMA mode with SFR (and re-writing all the graphics code) you'd never realize that kind of reduction in GPU usage ... so something isn't accurate there.

SLI should help with AA processing.   However SLI is very driver version dependent (each nVidia driver version release was a roll of the dice on what it may or may not break in P3D when using SLI) I eventually just gave up on SLI ... along with a few other bugs that surfaced because of SLI, bugs that rely on nVidia to fix which may or may not happen any time soon.

Best way to test SLI is crank up GPU intensive tasks, like 8XSSAA, DL, DR, reflections, shadows and shadow quality (my prepar3d.cfg tweaks to shadow quality will also increase GPU loads) ... these should easily put a single GPU at 100% load (any GPU even an OC Titan V).  If SLI suddenly reports 50% load on two or more GPUs, then something is not right, either the measuring software isn't measuring correclty or something else, there is no way one could realize a 2X reduction in load under current DX11 constructs with SLI under P3D ... that would be suggesting ZERO overhead with adding a 2nd GPU and that's just NOT the case, ever.

Cheers, Rob.

[mokeiko 39]

Quote

However SLI is very driver version dependent (each nVidia driver version release was a roll of the dice on what it may or may not break in P3D when using SLI) I eventually just gave up on SLI ... along with a few other bugs that surfaced because of SLI, bugs that rely on nVidia to fix which may or may not happen any time soon.

Totally why I gave up SLI as well. To me, It wasn't worth the premium to add another card only to have updated drivers break it.

G-SYNC is the bees knees now for me.

Fly safe

Edited July 26, 2018 by mokeiko

[westman 257]

Here is my 1080TI SLI and Single, i think it work ok no super test but you see the difference anyway

https://www.dropbox.com/s/i2lmqaw93yfuail/Single.png?dl=0

https://www.dropbox.com/s/ginh6hpu1mr5yor/SLI.png?dl=0

 

[Guest]

8 minutes ago, westman said:

Here is my 1080TI SLI and Single

Any chance you can do a video with the MSI overlay or Precision OCX overlay?  I'm puzzled at why one GPU is maxed at 99% and the other fluctuating (I would have expected both to be fluctuating)?

Cheers, Rob.