What are your settings James? Let me set those precisely.
 
I have just been running the MJCQ400, no changes with that either.

Why not look at this from another angle, in DX10 and P3D DX11, only the transparency setting makes a change, the "Antialiasing - Setting" does not make a difference - because it's for D3D9.

Why not look at this from another angle, in DX10 and P3D DX11, only the transparency setting makes a change, the "Antialiasing - Setting" does not make a difference - because it's for D3D9.

 

Well, yes, we already know that the base AA setting in inspector has no effect in DX10 or DX11

 

Your original post states that transparency sparse grid super sampling (SGSS) AA does nothing / does not work in DX9. 

Yes we are discussing that.

 

Since the NI settings add up and go into registers on the GPU, it's possible some combination of NI and Driver in the past exhibited unusual behaviour and the sgss setting made a difference somehow in D3D9. If it did I would suspect the same improvement would be had by choosing simply a stronger AA method from the D3D9 selection. I'm happy to load up that driver and NI combination and test it.

Steve, same settings here as in your most recent shot there.

It's unfair to Brian dragging this discussion down. Irrespective of the sgss setting - he wants to set up his card so he can be sure it's working properly simple as possible. But see an improved AA to compare the old card.

 

Brian, make a fresh profile and try the basic setup - and no tweaks:

Antialiasing - Behaviour Flags = None

Antialiasing - Mode = Override any application setting

Antialiasing - Setting = 2x2 [2x2 Supersampling (D3D only)]

 

I don't see any worse performance with 2x2SS, and it seems marginally better than 8xS which is 1x2SSAA + 4xMSAA. Also 2x2 is the D3D only setting and should work if the card's good.

 

(Notes I've been keeping in a text file and got mixed up a bit and beware I copy pasted some wrong settings around earlier, got late.) I'm still getting the no change with sgss result and don't doubt you guys are seeing something.

You can make a default profile by using the restore button in CP, and apply, that deletes the fsx profile. Then make any setting in CP to create the new profile, so set Antialiasing - Mode = Enhance the application setting to make the new profile with that already set, go into NI and the fsx profile can be selected, set 2x, and behaviour flags as above. Then try 4x, should be only minutes of work.

 

If you can do that first, we can see if it's working right. Be careful to follow those instructions carefully. If that does not work with the default cfg then the card is not working properly.

 

Hi Steve,

I did all the above. Did restore in the CP which did set the FSX profile (in CP) back to default.  I then changed AA mode to" enhance the application setting" I then went into NI, selected, MSFSX, changed the behavior flags as written above. Then I set 2x, then 4x. I still have the same performance.

When I opened FSX ( the game) all my original setting remained  there. Nothing was changed. Is  the CP restore supposed to change setting to default within FSX itself? The restore did make changes to the CP and NI, but not FSX itself.  I did not remove the tweaks from the FSX.CFG. Should I delete and let FSX create a new FSX.cfg, and along with the settings on the CP and NI you recommended, that should do the trick?

Thanks so much for all your help.

Hi Brain, I will be back shortly and help you out with that. Guys, I think I found what makes the sgss differences. Back later...

Brian, I'm thinking that you have been getting poor results whatever AA etc. It could be something else, maybe to do with the way threading works differently on NVidia cards to your AMD Radion. Can you tell me your CPU and is it hyperthreading enabled? I'll get back to you on the AA, but let's check out the CPU loading first. Are you using FSX-SE or FSX?

http://www.bit-tech.net/hardware/graphics/2006/04/07/transparency_adaptive_aa_explained/1

 

Steve, this is all explained here. These types of textures have always been around in dx9 titles including fsx

I have no idea why you think this is only a dx10 and up thing.

Well, I did say i'm putting it very simply. You're thinking the pictures that overlay the model are the same, they are. With DX10 and DX11, they make shaders with the textures and then they are applied to the geometry.

Yes we are discussing that.

 

Since the NI settings add up and go into registers on the GPU, it's possible some combination of NI and Driver in the past exhibited unusual behaviour and the sgss setting made a difference somehow in D3D9. If it did I would suspect the same improvement would be had by choosing simply a stronger AA method from the D3D9 selection. I'm happy to load up that driver and NI combination and test it.

 

You want very visible proof that SGSS gets rid of shimmering in DX9?  Go get the Quality Wings Avro collection and look at the large annunciator panel next to the engine gauges.  Without SGSS they are a shimmery mess.  With SGSS at 2x they are better, and at 4x there is almost no shimmer.  Now, I will say I have not run DX9 since the 314.22 days so I have no idea if later versions have broken this, but based on other posts I've seen to you recently I'm going to guess that it still works on current driver versions.

First of all if you've been using DX10 Preview, go into the .cfg and delete the DX10 DISPLAY.Device sections. The DX10 sections have the extra '.0' before the final '.0', or '.1' for your second display device.

 

so remove the DX10 DISPLAY.Device section shown here with the .0.0, .0.1, .0.2 etc.

[DISPLAY.Device.NVIDIA GeForce GTX 680.0.0]

Mode=1920x1200x32

Anisotropic=1

AntiAlias=1

 

Now we can set up DX9 without strange results, and see the effect of making extra settings to the AA section in NI.

 

 

Using the settings in NI. Have you wondered why 8xS + ANY value of SGSS looks the same? There's no extra pass of AA by setting the Transparency SGSS settings. I suspected it's because or'd values overblow a register, or create some difference or change to the way the AA strategy is interpreted internally resulting in a straightforward SSAA. That's why the fps is hit because the sim is writing to a much larger area than the screen with SSAA.

 

Choose 3x3 or 4x4 and compare to 4xS+anySGSS, 8xS+anySGSS, and see 3x3SS, and 4x4SS.

 

Can we all have a think about Brian's new GPU now? Speak later Brian, let me know your CPU and HT on or off and we can look at the CPU loading with the new card.

 

Just a quick note on Performance with those AA modes. As I said I'm starting from a saved flight shows up with the exact situation. I'm taking samples from the GPU and compiling a fps graph with MS Access. Setting unlimited fps in the sim and turning off vsync in NI.

 

around 54fps: 8xS [Combined: 1x2 SS + 4x MS] + 2x Sparse Grid Supersampling

 

around 63fps: 4x4 [4x4 Supersampling (D3D only)]

 

Both these look the same on screen, so the drop in fps is unnecessary with the 8xS+2xSG setup.

 

I changed the scene for a much more complex one and result 49fps with 8xS+2xSG and 53fps with 4x4SSAA.

 

"Using the settings in NI. Have you wondered why 8xS + ANY value of SGSS looks the same? There's no extra pass of AA by setting the Transparency SGSS settings. I suspected it's because or'd values overblow a register, or create some difference or change to the way the AA strategy is interpreted internally resulting in a straightforward SSAA. That's why the fps is hit because the sim is writing to a much larger area than the screen with SSAA."

 

No, I haven't, because it doesn't!!!

 

I use 8xS in inspector and without sgssaa set in transparency section I have a shimmery mess with trees, fences lampposts and certain panels.

Again Steve, the difference is night and day and you have now had 3 or 4 people saying the same thing.

 

I'm not sure if you are deliberately being obtuse or if you are trying to not so gracefully back pedal.

I use 8xS in inspector and without sgssaa set in transparency section I have a shimmery mess with trees, fences lampposts and certain panels.

Again Steve, the difference is night and day and you have now had 3 or 4 people saying the same thing.

 

I'm not sure if you are deliberately being obtuse or if you are trying to not so gracefully back pedal.

You just did not read or understand my post Glenn. I too see the night and day difference in the AA - that is not in dispute. Go back and read exactly what I said in my last post.

 

"Using the settings in NI. Have you wondered why 8xS + ANY value of SGSS looks the same? There's no extra pass of AA by setting the Transparency SGSS settings. I suspected it's because or'd values overblow a register, or create some difference or change to the way the AA strategy is interpreted internally resulting in a straightforward SSAA. That's why the fps is hit because the sim is writing to a much larger area than the screen with SSAA.

 

Choose 3x3 or 4x4 and compare to 4xS+anySGSS, 8xS+anySGSS, and see 3x3SS, and 4x4SS.

Just a quick note on Performance with those AA modes. As I said I'm starting from a saved flight shows up with the exact situation. I'm taking samples from the GPU and compiling a fps graph with MS Access. Setting unlimited fps in the sim and turning off vsync in NI.

 

around 54fps: 8xS [Combined: 1x2 SS + 4x MS] + 2x Sparse Grid Supersampling

around 63fps: 4x4 [4x4 Supersampling (D3D only)]

 

Both these look the same on screen, so the drop in fps is unnecessary with the 8xS+2xSG setup.

 

I changed the scene for a much more complex one and result 49fps with 8xS+2xSG and 53fps with 4x4SSAA."

 

I also said a situation within the DX10 DISPLAY.Device section throws the DX9 AA off.

I am reading Steve.

 

You are suggesting that no matter what level of sgssaa I set there is no difference? Right?

 

Well again, there is a clear difference on my setup with 2x and 4x. 4x eliminates all shimmers but at a major perf cost, especially in clouds.

2x is more acceptable in terms of perf but a slightly reduced quality.

Bigger picture here.... You are now agreeing that transparency sgssaa setting works in dx9?

Or are you now saying that selecting sgssaa simply invokes full screen supersampling?