Performance difference between Nvidia's SSAA and P3D's own?

[Sethos 4,343]

Now P3D v4 has its own native SSAA implementation, which is great. However, I heard someone mention the one applied with Nvidia Inspector / Profile Inspector was more performance friendly for them, with the same visual result. Has any of you guys done any comparisons? :)

[Lenny777 299]

Great question. I'd like to know because I see a huge performance hit when I use 8x as compared to 2X using P3Dv4 choice instead of my Nvidia card.

[ComSimPilot 623]

To me the P3D 4xSSAA while removing all "edges" it's smoothens out the aircraft display textures. Also it has a framerate hit in specific conditions. I find the 8xMSAA to have the best visual/performance ratio. I haven't tried using the nvidia profile instead. 

[blackbird7 3]

This is a good question.  My intuition tells me that you would see a bigger performance hit with NVIDIA's SSAA, but that contrasts with the report you mentioned.  

I have nothing to back this up except the eye test, but it seems that Prepar3D's SSAA is selective.  As Daedalus mentioned, 4xSSAA smoothens out aircraft textures, but my airport textures and such are then worse than 8xMSAA.  It's as if the 4XSSAA option sets the aircraft textures to 4xSSAA and everything else at 4xMSAA.  

This could be really good if that means that I can bump up to 8xSSAA without applying sparse grid supersampling to the clouds, but I don't know how cloud antialiasing is performed in P3D v4.  I need to do more testing.

[Lenny777 299]

I did the override using the Nvidia card but not the app choice yet though the latter has been the method I've always used. So I'll try the reverse and report back.

 

But using the Nvidia settings and monitoring GPU usage my main assessment was visual and GPU usage with FPS. This test was done with the new TFDi (beta) 717 which was a hefty FPS hog in P3Dv3 so no small feat. At Aerosofts new ENGM airport, 30% traffic, minimal cloud cover using AS16(v4 beta) and AF 4X. Resolution 2560x1440.

8xSSAA - 100% GPU external (of course more textures etc) and 80-90% GPU internal. FPS 24-25 internal, external 53-60.

8xMSAA - 70% GPU usage external and 40-45% GPU internal. FPS 24-26 internal, external 64-75 FPS.

I did the same test for all the SSAA and MSAA  settings and of course the SS GPU usage at 4x and 8x is virtually the same at 100% GPU.. The in-cockpit though is around 65% at 4SS.

The best compromise for performance and looks if you want to use SS is 2xSSAA - 75-80% GPU load external and 50% internal.

There is then a significant frame drop of about 20% external if you use 8xSSAA over 8xMSAA but only a marginal difference with the internal cockpit. However, the GPU load difference is huge.

As to looks - got to tell you not a great difference between 8xSSAA and 8xMSAA. I've been using 2xSSAA and there is a difference quality wise in favour of 8xMSAA over that and given the performance difference this will be my new setting. 2xSSAA has a 75%-80% GPU hit external and 50% internal both over 20% more than 8xMSAA.

i7-7700k 4.2 GHz, DEEPCOOL GAMMAXX 400 CPU Cooler 4 Heatpipes 120mm PWM Fan with Blue LED, ASRock 270 PRO4 MOBO, ZOTAC ampExtreme 1080 8GB, 16 GB DDR4 Ripjaws, Thermaltake Core V71 Case, RM850xCorsair psu, ADATA M.2 2280 240GB SATA III TLC Internal Solid State Drive (SSD) WIndows 10 Pro, 2 TB Samsung and 3 TB WD HDs .

 

 

[Lenny777 299]

So I did the Nvdia app controlled settings. Same tests all settings noted above The bottom line is I saw a very slight advantage using the Nvidia override settings but it has to be within a margin of error owing to whatever variables. If anything I'd favour but only slightly the override.

 

I did one further test which doesn't involve the issue of override. Just using app choice on Nvidia and that was an extra long shot from the top of that airport noted above. Difference between 8xSSAA and 8xMSAA.

8xSSAA - 26-33 fps, 100% GPU

8xMSAA - 38-40 fps, 70% GPU.

 

Again, visually, virtually no difference I could see but notice the GPU usage and about 50% gain in fps. I like 8XMSAA with 4xAF and will be my new video setting of choice.

[B777ER 2,486]

18 minutes ago, Lenny777 said:

Again, visually, virtually no difference I could see but notice the GPU usage and about 50% gain in fps. I like 8XMSAA with 4xAF and will be my new video setting of choice.

Why would you use only 4xAF? Your system can cleary handle 16xAF without issue. I'm a 6700k 4.5Ghz OC with a 980Ti with 64GB of ram and run 16AF.

[beachdog2001 21]

Are you guys doing these tests with dynamic lighting enabled?

[Lenny777 299]

1 minute ago, B777ER said:

Why would you use only 4xAF? Your system can cleary handle 16xAF without issue. I'm a 6700k 4.5Ghz OC with a 980Ti with 64GB of ram and run 16AF.

Actually it was completely arbitrary for this test. I agree AF 16 with my current texture use of 2048x2048 is fine with no noticeable fps nor gpu difference from AF x4. Increase textures though to 4K will start to have a difference.

[Lenny777 299]

3 minutes ago, beachdog2001 said:

Are you guys doing these tests with dynamic lighting enabled?

Yes I am.

[Lenny777 299]

31 minutes ago, beachdog2001 said:

Are you guys doing these tests with dynamic lighting enabled?

Actually a better response to this would have been dynamic lighting at night at the above noted airport. And yes there is a big hit and difference with nightlighting, night textures, dynamic lighting using the same variables of that airport etc.

8XMSAA - GPU load internal about 90%. FPS still around 23-24 fps. External 100% GPU. Around 56-60 external fps with panoramic shot 35-37 fps.

8XSSAA - GPU internal 100%, FPS 7-8. External 100% GPU around 19-20  fps. Panoramic shot 14 fps 100% GPU.

 

So over a 100% hit in fps between using 8XMSAA and 8XSSAA. For the other poster I increased the AF to 16x for everything.

[stripealipe 87]

1 hour ago, Lenny777 said:

So I did the Nvdia app controlled settings. Same tests all settings noted above The bottom line is I saw a very slight advantage using the Nvidia override settings but it has to be within a margin of error owing to whatever variables. If anything I'd favour but only slightly the override.

 

I did one further test which doesn't involve the issue of override. Just using app choice on Nvidia and that was an extra long shot from the top of that airport noted above. Difference between 8xSSAA and 8xMSAA.

8xSSAA - 26-33 fps, 100% GPU

8xMSAA - 38-40 fps, 70% GPU.

 

Again, visually, virtually no difference I could see but notice the GPU usage and about 50% gain in fps. I like 8XMSAA with 4xAF and will be my new video setting of choice.

Could you explain exactly what you mean by 'app choice' on Nvidia and how you set up the 8xMSAA to get the FPS increase please? Thanks!

[Lenny777 299]

4 minutes ago, stripealipe said:

Could you explain exactly what you mean by 'app choice' on Nvidia and how you set up the 8xMSAA to get the FPS increase please? Thanks!

On the Nvidia control panel ----> Manage 3D settings - the choice is 'Application-controlled' or: Enhanced, off, or override any application setting. So 'app choice' I'm meaning application-controlled.

In the graphics settings of P3Dv4 is where you have your MSAA etc. controls.

 

[MarkW 1,493]

Question - If you were using Nvidia Inspector for P3Dv3 (still installed) and now you load up P3Dv4, does Inspector still run its profile or do we need to wait for a new Inspector version to recognize this version of P3D?  I have setup 4xSGSS in the sim but have no idea if my previous P3Dv3 Inspector settings are just overriding it.

[stripealipe 87]

Hmmmm, i don't really get it. Doesn't it default to 'application controlled' anyways?

So in your first test, you simply had the Nvidia control panel set to default and you got 24-26fps with 8xMSAA, and then you set the Nvidia control panel to 'app controlled' and then got 38-40fs?!

I think i'm missing something here....!