New D3D FSAA Viewer

[dr3amz]

im just playing with this more, alternating from 2xaa to 2xaa+tempaa=2

and im starting to wonder just how much of a difference there is... perhaps not as great as i first thought...

for example, the hut on nature (3dmark2k1) the roof looks exactly the same on aliasing using either method... except with 2xaa you have constant aliasing, rather than on/off (which can leave jaggies around)...

 

[iodsnips]

When comparing the temporal AA (which you guys say forces vsync on) to regular AA, shouldn't you force vsync on in both tests?

vsync off normal aa VS vsync on temporal AA will of course show major performance differences when used with double buffering

I can't wait to try this new temporal AA stuff. Gotta put backin my 9800pro first.

You guys were saying that the higher hz make the effect better? Lucky my 22" CRT can do wild high hz levels

800x600 at 200hz, 1280x960 at 133hz, and 1600x1200 at 113hz should prove the be very interesting.

 

[tEd]

I'm trying to get it to work...

The FSAA viewer shows the frame by frame differences in windowed mode. However, when I switch to full-screen, the "temporal" aspect of it stops, and reverts back to plain AA.

I switch back to windowed....and the temporal mode returns. Anyone else have similar problems? (Using latest official Cats...and DVI interface to flat panel)

Add "TemporalAAFrameThreshold" to the reg with no value or 0 it should solve the problem you have

 

[dr3amz]

When comparing the temporal AA (which you guys say forces vsync on) to regular AA, shouldn't you force vsync on in both tests?

vsync off normal aa VS vsync on temporal AA will of course show major performance differences when used with double buffering

I can't wait to try this new temporal AA stuff. Gotta put backin my 9800pro first.

You guys were saying that the higher hz make the effect better? Lucky my 22" CRT can do wild high hz levels

800x600 at 200hz, 1280x960 at 133hz, and 1600x1200 at 113hz should prove the be very interesting.

basically there is nothing between the performance of the two, and whilst i thought tempaa gave a big increase in quality - im not so sure now, here are the benches again.

http://homepage.ntlworld.com/dr3amz/tempaa.htm :?

sorry, looking at the FSAAViewer you can clearly see how it helps, by alternating the pattern 2xfsaa looks more like 4xfsaa.

however, comparing parts of 3dmark the 2xfsaa and 2xtemp just look so similar `in game`

 

[MuFu]

The "TemporalAAFrameThreshold" setting has no effect for me (assuming it is what we think it is). I have it set to 60 and you can pick certain gaming conditions in which the drop from 60=>30=>15FPS etc is clearly accompanied by distinct levels of flicker severity. This is with v-sync on or off in the control panel.

MuFu.

try "TemporalAAFrameEnableCount x" while x=1 then the threshold should work

Hmm... nope, still doesn't seem to work; even after a restart.

 

[Dwarden]

I myself noticed this

using 4x2 and 4x3 in FlashPoint Resistance with maximal details revealed one thing, 4x3 got NO stuttering (or almost undetectable) ... 4x2 gave some, and 4x got many especially on camera scripted movements ...

can anyone look into it?

 

[dr3amz]

I myself noticed this

using 4x2 and 4x3 in FlashPoint Resistance with maximal details revealed one thing, 4x3 got NO stuttering (or almost undetectable) ... 4x2 gave some, and 4x got many especially on camera scripted movements ...

can anyone look into it?

interesting, especially as benchies prove there is no difference in speed between the two modes... :?

ATi, help!

 

[tEd]

The "TemporalAAFrameThreshold" setting has no effect for me (assuming it is what we think it is). I have it set to 60 and you can pick certain gaming conditions in which the drop from 60=>30=>15FPS etc is clearly accompanied by distinct levels of flicker severity. This is with v-sync on or off in the control panel.

MuFu.

try "TemporalAAFrameEnableCount x" while x=1 then the threshold should work

Hmm... nope, still doesn't seem to work; even after a restart.

it works for me in FSAAViewer didn't try any game though

 

[Bambers]

When comparing the temporal AA (which you guys say forces vsync on) to regular AA, shouldn't you force vsync on in both tests?

vsync off normal aa VS vsync on temporal AA will of course show major performance differences when used with double buffering

I can't wait to try this new temporal AA stuff. Gotta put backin my 9800pro first.

You guys were saying that the higher hz make the effect better? Lucky my 22" CRT can do wild high hz levels

800x600 at 200hz, 1280x960 at 133hz, and 1600x1200 at 113hz should prove the be very interesting.

you need to use games that can throw out those fps though.


I find ut2004 looks noticably better at 4x2t opposed to the 4xAA I normally run it at. Doubt I'll be able to use it online though as I often drop down to 20 with lots of people about.

 

[MuFu]

The "TemporalAAFrameThreshold" setting has no effect for me (assuming it is what we think it is). I have it set to 60 and you can pick certain gaming conditions in which the drop from 60=>30=>15FPS etc is clearly accompanied by distinct levels of flicker severity. This is with v-sync on or off in the control panel.

MuFu.

try "TemporalAAFrameEnableCount x" while x=1 then the threshold should work

Hmm... nope, still doesn't seem to work; even after a restart.

it works for me in FSAAViewer didn't try any game though

So it shows a non-temporal mode when you set the Threshold value above the v-sync frequency?

 

[dr3amz]

well i can see the difference in halflife more than in farcry i guess, and in halflife i keep a solid 100fps (even with 2xtempaa) it looks more like 4-6x aa with no performance hit - which is nice)

 

[tEd]

The "TemporalAAFrameThreshold" setting has no effect for me (assuming it is what we think it is). I have it set to 60 and you can pick certain gaming conditions in which the drop from 60=>30=>15FPS etc is clearly accompanied by distinct levels of flicker severity. This is with v-sync on or off in the control panel.

MuFu.

try "TemporalAAFrameEnableCount x" while x=1 then the threshold should work

Hmm... nope, still doesn't seem to work; even after a restart.

it works for me in FSAAViewer didn't try any game though

So it shows a non-temporal mode when you set the Threshold value above the v-sync frequency?

yes

I tried it in ut2004 and it works there too.

 

[dr3amz]

so setting to 0 seems the best bet, always on

 

[tEd]

The "TemporalAAFrameThreshold" setting has no effect for me (assuming it is what we think it is). I have it set to 60 and you can pick certain gaming conditions in which the drop from 60=>30=>15FPS etc is clearly accompanied by distinct levels of flicker severity. This is with v-sync on or off in the control panel.

MuFu.

try "TemporalAAFrameEnableCount x" while x=1 then the threshold should work

Hmm... nope, still doesn't seem to work; even after a restart.

it works for me in FSAAViewer didn't try any game though

So it shows a non-temporal mode when you set the Threshold value above the v-sync frequency?

yes

I tried it in ut2004 and it works there too.

so setting to 0 seems the best bet, always on

yes but it is probably a good thing to set it to 60+

 

[Dwarden]

i set threshold to monitor refresh (100/110/120/150 must made util to detect resolution ...) , it looks great and works normal ...

 

[Dave Baumann]

Please, quit with the nested quoting.

 

[Moose]

I seem to get a bit more flickering at 3t than at 2t so I turned it back.

Are the patterns hard coded in the chip? It would be nice to be able to experiment with changing the patterns around to get the best AA with the least ammount of flickering to the eye.

I'll bet ATI has really been playing with this and it will look even better on the R420.

heh, there is no way the Nv40 will be able to touch the R420's AA quality and speed this round. I was already thinking that it would be slightly faster without using this. Now its like to be faster and look much better as well!!!

 

[dr3amz]

i doubt you'd notice the flickering too much if at all in game...

 

[Moose]

i doubt you'd notice the flickering too much if at all in game...

You don't notice it much at all. I loaded up UT2004 and even with the worst flickering at (3T) I had to stop moving and stare at an off angle edge to see it. At 4XAA/2T there is almost no flickering at all.

 

[digitalwanderer]

I changed temporal to 2 and I think I like it better.

BTW-LOL at Wavey! You should seen Quinn go off at ANYONE over on EB for nesting quotes, I hear it really chews up database space.