Silly reviewers saying there is a 20%+ hit with AF used?

[K.I.L.E.R]

2003-03-07 12:12:28 - UT2003
Frames: 9013 - Time: 112672ms - Avg: 79.993 - Min: 44 - Max: 151

2003-03-07 12:17:05 - UT2003
Frames: 11220 - Time: 142905ms - Avg: 78.513 - Min: 46 - Max: 150

2003-03-07 12:27:27 - UT2003
Frames: 11929 - Time: 137228ms - Avg: 86.928 - Min: 44 - Max: 153

2003-03-07 12:57:39 - UT2003
Frames: 5294 - Time: 72204ms - Avg: 73.320 - Min: 46 - Max: 104

1st test was using trilinear quality AF (128 tap aniso)
2nd was using performance AF (64 tap aniso)
3rd test was using only trilinear filtering with no AF
4th test 4x FSAA and 128 tap aniso

The slider for both was set on 16x. No AA was used.
So where is the AF performance hit using trilinear? Wasn't is supposed to be like 20%?
Benchmarking I have used the exact same path and have jumped and shot at certain points for the tests to remain consistant as possible.

This game isn't the only game where 128x AF takes jack all performance hit over 64 tap and even no AF.

May I ask why? (using cat 3.1)

I played at 1024x768 everything set to max.
$0400000005 AGP rate : 1x 2x 4x supported, 4x selected
$0400000006 AGP SBA : supported, enabled
$0400000007 AGP FW : supported, enabled

$0900000004 Core clock : 378.000MHz
$0900000005 Memory clock : 337.500MHz (675.000MHz effective)

 

[demalion]

That looks like a botmatch result....is it?

 

[K.I.L.E.R]

I wish. LMFAO!!!

I had 1 bot while playing, I made sure CPU limitations diminished.
Play with no bots as well and my framerate is nearly exact, besides the 1% error ratio.

I oly have a 1.8GHz P4, my botmatch scores would not be anywhere near that. This benchmark was done in game not using UT2003's in built bench.

 

[kyleb]

well then it is bound to have inconsistences, conceivably very extreme ones.

 

[K.I.L.E.R]

well then it is bound to have inconsistences, conceivably very extreme ones.

Bah! I tested it in so many ways and my framerates are so close. I tested it with 13 bots too and still the performance hit with AF (even on different maps) was very little and in some cases nothing. I also tried other games like BF1942 and DS, Quake 1, Half-Life and Black and White.

BTW: I am going to do more testing. If I do get a large performance hit I will let you guys know.

PPPPPS: I still have not heard from you guys why my results are funny? Fillrate not as important maybe at my current resolution?

 

[Bambers]

If you're running at 1024x768 then that will explain most of it.

Try 1600x1200

 

[K.I.L.E.R]

If you're running at 1024x768 then that will explain most of it.

Try 1600x1200

Monitor doesn't go that high.

 

[KimB]

Please please please do not use the number of "taps" to describe the level of anisotropic filtering. Just use the degree. The degree of anisotropic is actually descriptive of what is going on.

 

[K.I.L.E.R]

Please please please do not use the number of "taps" to describe the level of anisotropic filtering. Just use the degree. The degree of anisotropic is actually descriptive of what is going on.

Degree of 16 I am using. Why? I prefer taps.

 

[Marc]

"Silly reviewers saying there is a 20%+ hit with AF used?"
=>
"Silly K.I.L.L.E.R must use benchmark.exe or benchmark batch file and scene who are not CPU Limited"

 

[K.I.L.E.R]

"Silly reviewers saying there is a 20%+ hit with AF used?"
=>
"Silly K.I.L.L.E.R must use benchmark.exe or benchmark batch file and scene who are not CPU Limited"

This benchmark was done in game not using UT2003's in built bench.

Bah! I tested it in so many ways and my framerates are so close. I tested it with 13 bots too and still the performance hit with AF (even on different maps) was very little and in some cases nothing. I also tried other games like BF1942 and DS, Quake 1, Half-Life and Black and White.

Open your eyes boy.
I tested CPU limited areas as well as non-cpu limited areas.

 

[Marc]

With a 1.84 G P4 in 1024*768 your are VERY cpu limited ... just look at your result with 4x FSAA & aniso.

As for the "in game" benchmark, it's not exactly reproducable so it's better to use the demo batch.

 

[K.I.L.E.R]

With a 1.84 G P4 in 1024*768 your are VERY cpu limited ... just look at your result with 4x FSAA & aniso.

As for the "in game" benchmark, it's not exactly reproducable so it's better to use the demo batch.

I also said I tested other games like Quake 1 which I doubt would be CPU limited, more fillrate limited.

BTW are you sure it's more CPU limited? My scores really don't differ much at all when my CPU is clocked back at 1.5GHz.
Of course underclocking my card then a larger impact occurs with my framerate.

 

[Nagorak]

It's interesting that you have these experiences, but I don't think you're going to convince everyone else that AA and AF has no performance hit. That hasn't been my experience, in any case, although I have been pleasantly surprised by how playable the R9700 is with those features enabled. Honestly, the R9700 is just one ridiculously overpowered card. That's why it's made the FX look so bad.

 

[K.I.L.E.R]

It's interesting that you have these experiences, but I don't think you're going to convince everyone else that AA and AF has no performance hit. That hasn't been my experience, in any case, although I have been pleasantly surprised by how playable the R9700 is with those features enabled. Honestly, the R9700 is just one ridiculously overpowered card. That's why it's made the FX look so bad.

NO NO NO!!! There is a performance hit with AA, a big one. Just AF I am talking about having a small performance hit. I just added FSAA scores for curiosity.