limith Wrote:I find dismissive attitude like this very disheartening.
Try to see where I'm coming from here. I and many others here have spent years tinkering with these things. I've played around with every single AA option in dolphin, various dll injectors, nvcp, CCC, and nvidia inspector. There are already many threads about it. Then one day a new thread pops up with a new user that claims to have done things that I know from my own testing in the past are impossible. So yes in that situation I will act "dismissive" towards that persons post. I have nothing against you personally but reading your post made me raise an eyebrow. Something like this deserves detailed analysis to arrive at the truth.
limith Wrote:Not everyone knows this. I'm just trying to help others.
That's fine. But I can still criticize it. All I ask is that in the future if you're going to make a thread about some image quality enhancement made by a third party application that you should probably provide screenshots of said enhancements and provide usable benchmarking data with details on how it was collected (settings!). Be as specific as you can. Take the time to be thorough.
limith Wrote:Those weren't numbers made up from thin air
I question the effectiveness of benchmarking by "eyeballing" the fps counter. FPS can vary wildy from one spot to another in the same game. Both dolphin and fraps have proper benchmarking features built in. And since an 85% drop in framerate from a 300% increase in internal resolution is not possible on your hardware I would naturally be skeptical enough to want proof. Since I cannot reproduce it myself.
limith Wrote:I explicitly said to use DirectX9 or DirectX11, there's no difference in numbers for me overall.
You should specify which one was used (d3d9 or d3d11) in each of your results even if it made no difference to you. Otherwise we can't properly reproduce your results. That's why I asked.
limith Wrote:I also explicitly stated what each number's AA method was. 8x = 8x plain. 4xSSAA = 4x SSAA. Did you even read it? No where does it list MSAA.
What on earth is "plain AA"? There is no such thing. You must be referring to MSAA which is what dolphins d3d11 backend uses. You didn't list the backend so I wanted to clarify.
limith Wrote:Again, I explicitly said to use DirectX9 or DirectX11, I know it doesn't work with OpenGL, it causes a black screen for me, I even said in the post above to disable override if you want to use OpenGL again.
What specifically doesn't work with openGL? Some driver AA enhancements do work with openGL. Your list of steps says to set any driver method from either company's control panel. It does not specify a specific type. You seem to be referring specifically to AMD CCCs edge detect SSAA. But you didn't specify that anywhere.
limith Wrote:I'm not comparing MSAA. It's not even an option under DX9 or DX11 dropbox for me in normal Dolphin anyway, only an option for the OpenGL backend.
d3d11 uses MSAA. So you are comparing it whether you realize it or not.
limith Wrote:To be honest I can't notice any difference between the 9xSSAA Dolphin screenshot and the No AA at all Dolphin Screenshot (with no override). Curious how you tell the difference imho. What would be a good location/scene for comparison? Water?
You still haven't told us what IR you're using. Lower IR will make it easier to spot the difference. Make sure it's the exact same spot and angle too. Oh and upload original size PNG images. Don't use jpg compression or image hosting sites that don't provide direct image links.
limith Wrote:I think after looking at my GPU usage numbers that it's probably due to Dolphin not being optimized for my AMD card.
That doesn't make any sense. What exactly would be unoptimized specifically for your card and how would it cause this?
SRO07 Wrote:What the hell are you talking about Op? You can't have SSAA without performance hit, it's literally impossible.
Yes you can. With a low IR and a fast GPU you can.
limith Wrote:So explain to me why when I set override to true, OpenGL backend will fail to render if my graphics card is doing nothing to Dolphin, or why my graphics card is actually doing anything (35% usage).
Just because it's doing something doesn't mean it's working as intended.
limith Wrote:And yes I can without a performance hit, if the implementation of Anti-Aliasing in Dolphin poorly uses the graphics card.
That doesn't make any sense. The implementation is done using the API and even if it were poorly done (it's not, I can assure you) that would only explain the high performance hit of dolphins AA. It would not explain the lower performance hit of the driver AA. Which is what he's referring to.
limith Wrote:If I set AA to None in Dolphin, my graphics card says activity is ~0 (very low) at the main menu (the grass field). If I set AA to 4xSSAA in Dolphin, my graphics card jumps up to 99% usage at the main menu and I get lag, hence the drop in fps. If I set my graphics card to override 24xSSAA game settings and I set the Dolphin AA to None, my graphics card is at 35% usage at the main menu*. In other words, the override is more efficient and my graphics card is doing something. Hence I get "no performance hit" because I get 24xSSAA without the graphics card bottleneck. *Jumps to 59% once the background starts to change, numbers are when the game just starts. No AA in Dolphin/Graphics card uses ~30% gpu potential once the background starts to change.
What software are you using to monitor this? Clearly it's not producing correct readings. From 0% load without SSAA to 99% load with 4xSSAA obviously can't be right. That's just common sense.
If you're doing 24xSSAA on an already high IR without a significant performance hit that should be an immediate indicator to you that something is not right. No amount of optimization can allow you to render a scene like this at nearly 100 megapixels internal resolution without a significant performance penalty. Again, common sense should make you skeptical.
degasus Wrote:Driver overloaded AA may either change the settings for aa framebuffers (so msaa to ssaa or csaa switch) or enables AA for the default framebuffer.
AA in the default framebuffer isn't useful as we don't render into it (and in fact, it will break the ogl backend). Also AA config changes for the other backends can't increase the number of used samples, so every 4xMSAA setting in dolphin would results in a 4x..AA.. result.
As you don't see an slowdown on forced SSAA, I think you've only enabled SSAA for the default framebuffer. So you can be lucky, the rasterfont will look nicer
This explains the low performance hit of the driver AA. I knew it couldn't be working correctly. Thanks degasus.
limith Wrote:Although one good thing is that it made me turn on manual fan control since I just realized the auto was letting the GPU go to dangerous levels for the sake of maintaining a quiet fan. >.>
I have a feeling this is why you saw an 85% drop in performance with 4xSSAA in dolphin. Your GPU might be overheating and throttling. Check your GPU temperatures.
