FRAPS 2.0 & Video Recording

[Pete]

The difference is dramatic! Great work, Mike--this is probably the best way of showing the difference in AA quality.

BTW, I'd recommend ffdshow for Divx.

 

[Tim Murray]

nevermind--it no longer exists in any downloadable form

 

[Ollo]

No way! What are the detail levels of UT2k3? (I just re-installed it yesterday on Bubbles so I could benchmark it and the details looked a lot better...and I still game at 1024x768 and it just didn't look as good as mine. )

Forget about the texture quality, while I break out the big guns

http://www.nvnews.net/reviews/sapphire_radeon_9800_pro/video/nv35_quake3_4xaa_noan_divx.avi

http://www.nvnews.net/reviews/sapphire_radeon_9800_pro/video/r9800_quake3_4xaa_noan_divx.avi

But you already knew since you've seen both methods of AA in action.

Hm, the Radeon clip appears to have no AA? Lots of jaggies on the top of the left wall. Am I missing something here?

 

[Bambers]

And another thing, hold on, I'll be right back

Had to run a quick test. It sure is nice having two systems side-by-side now

Anyway, one of the first things I do when I get a new graphics card is fire up Quake 3 and set up a 6 player bot match in DM1 with FRAPS running in the background.

High quality settings with maximum texture detail and geometry. High quality sound. Easy to setup, I use the default frag limit of 20, and I always win

edit:

jsut tried with bilinear and min fps was one at 52 (first time i died, q3 always seems to pause slightly when that happens, rest was all 70+)
Resoultion set to 2048x1536 on an Athlon XP 2700+

GeForce FX 5900 Ultra - Avg: 152 - Min: 93 (Detonator 44.03)

Radeon 9800 Pro - Avg: 93 - Min: 33 (Catalyst 3.5)

Strange. May require some investigation...

ah but was that on nightmare? (1.14 nightmare that should be, not the wussy 1.24 stuff)

Tried that on my 9800pro 128mb and after getting worrying 20fps on starting before realising I'd left 4xAA and 16xQ af on i did dm1 with 6 bots (not 6 player but nm) for about 5 mins with fraps counting, avg 100fps. min 38. one of the 3 drops below 60 which i can atribute to me and a rocket luancher v all 6 bots in a corridor and the ensuing fog of red I couldn't actually see through.

The FXs results do seem quite high. Was that the same demo for each or a seperate run?

 

[MikeC]

Hm, the Radeon clip appears to have no AA? Lots of jaggies on the top of the left wall. Am I missing something here?

Hi Ollo. The Radeon 9800 Pro video was captured with 4X AA enabled at a resolution of 1024x768. Try to get the resolution as close to 1024x768 as possible in the program you're using to play back the video.

Watching the video in full screen or changing the desktop resolution are a couple of solutions. Otherwise, the video may automatically be resized to fit the resolution of your viewer, which will result in distorted playback.

 

[MikeC]

ah but was that on nightmare? (1.14 nightmare that should be, not the wussy 1.24 stuff)

Tried that on my 9800pro 128mb and after getting worrying 20fps on starting before realising I'd left 4xAA and 16xQ af on i did dm1 with 6 bots (not 6 player but nm) for about 5 mins with fraps counting, avg 100fps. min 38. one of the 3 drops below 60 which i can atribute to me and a rocket luancher v all 6 bots in a corridor and the ensuing fog of red I couldn't actually see through.

The FXs results do seem quite high. Was that the same demo for each or a seperate run?

Hehe. I setup a standard deathmatch. The days of multi-player GLQuake are long gone and I've grown old and weary now. Playing in nightmare mode could result in a call to 911

The FX results were from the gameplay section in our preview: http://www.nvnews.net/previews/geforce_fx_5900_ultra/page_2.shtml

Each result was based on playing a 6 player botmatch with a fraglimit of 20. I played multiple matches with the same graphics settings and the variation in the minimum and average frame rates between each match was low.

I've invested a lot of time practicing gameplay scenarios that were specifically designed by me for benchmarking purposes. There are times when an anomaly occurs in which case I will discard the result and run the test again. However, if I end up with significant variations in performance after repeated tests, I may redesign the gameplay scenario or scrap testing with that particular game altogether. However, this is an exception and doesn't happen very often.

 

[MikeC]

The difference is dramatic! Great work, Mike--this is probably the best way of showing the difference in AA quality.

Thanks for the compliment Pete! Now I know that it's been a worthwhile project.

 

[Enbar]

A while back I wrote some fragmentProgram tests and noticed my 9700 at the time had some strange results at resolutions greater than 1600x1200. That led me to write some fill tests and sure enough fill rate dropped off at >16x12. Then I turned on 4x multisample and that fall off turned into a cliff.

After that I just avoid resolutions >16x12 which is a little anoying but bearable.

 

[fallguy]

Hello,

When I record a short 15sec clip, and then try to play it, its all garbled. Only game I have tried it on is BF1942+DC. Im on a 9800 Pro, and Cat 3.6's. Tried it on 3.5's as well, same thing.

Is there some setting I am doing wrong?

 

[Pete]

The difference is dramatic! Great work, Mike--this is probably the best way of showing the difference in AA quality.

Thanks for the compliment Pete! Now I know that it's been a worthwhile project.