I've been meaning to post some information concerning the performance hits when enabling adaptive AA on the X1900 series, but due to recent hardware changes a complete analysis is not possible. Fisrt off, I do not have my X1900 any longer so doing further benches is rather difficult. Of the games that I began to bench, BF2 and FEAR were first. I was meaning to bench others as well since I have a decent amount of the more recent major DX9 titles. The testing was monitored by using FRAPS and the recordings were sent to a FRAPS log. With FEAR I used the in-game benchmark.
Relevant System Hardware Specifications:
[*]X1900XTX @740/850(1700)
[*]AMD Athlon 64 4400+ Toledo
[*]DFI Lanparty UT SLI-DR Expert
[*]2x1GB G. Skill Extreme Edition DDR500 (PC4000)
[*]OCZ Powerstream 520W
Relevant System Software Specifications
[*]Windows XP 32-bit w/SP2
[*]Cat. 6.6 + Chuck Patch
[*]ATI Tray Tools
[*]ATI Tool
[*]BF2 and FEAR of course and patched to the most recent versions.
Please forgive me for not immediatedly providing specific frames as I do not have my normal computer up and running at this point in time. (Believe me, it isn't fun)
My driver settings were also the highest quality settings I could achieve minus a 6xAA setting and Vsync + Triple Buffering. Instead it was 4xAA set from the games and the Vsynce remained disabled for accurate frames.
I did three AAA (Adaptive Anit-Aliaising) settings: No AAA, Peformance AAA, and Quality AAA. (Note: In Tray Tools there is a setting that displays both "Quality" and "Quality with Trilinear". Just for curiousity's sake, I included the "Quality with Trilinear" setting in my initial testing. The difference between "Quality with Trilinear" and the normal "Quality" setting was within 1~2 frames and therefore I considered it a negligible difference.
BF2
In game settings were set to "4", a level that has to be hacked through the config files and results in an "Ultra-High" setting while in the BF2 video options. It was the absolute highest quality settings I could achieve through BF2 itself. Each simulation was done on Daqing Oilfields in Single Player Mode. Each time I traveled the same path of attack. Starting as U.S. and going towards the middle neutral bunker. I wanted to get a good mixture of chain-link fences and vast vegetation since AAA works intensely with those types of alpha textures. The chain-link fences present in the bunker provided a small amount of alpha textures and a decently heated battle as the U.S. and China sides clashed for the neutral flag. I only ran and moved in positions that were easy to recreate for each test and displayed close to the same amount of objects. After the bunker I would diagnoly make my way towards the edge of the solid fence and turn right so as to get a screen full of the higher ground covered with trees. (Note: this is where the peformance hit was felt the most throught any of the testings.)
Once again I apologize for not being able to provide exact frames at the moment and I will update my own findings once I can access my hard drive again, but I remember them within a 5~10 frame set as well as the performance hits they gave.
[*]No AAA
70~100 frames constantly.
[*]Performance AAA
65~100 constantly. I remember it was very close to a 5% decrease.
[*]Quality AAA
48~80 or 90ish. I vividly remember that this was a 25% performance decrease.
FEAR
I used the in-game benchmark for all tests. The benchmark was ran three times and averaged out to portray the average results for each specific setting. In-game settings were all maxed through the game's video options panel (no config file tweaking). Soft Shadows were off.
(All I can really remember was the average section, the middle one that displays the widest margin for the majority of the frames.)
[*]No AAA
~35-39 frames.
[*]Performance AAA
Almost identical to "No AAA". The difference was negligible and therefore I do not remember exact frames for it.
[*]Quality AAA
~45-49 frames. (Yes, I checked again and again. "Quality" AAA gives around a 10 frame increase to the X1900's in FEAR.)
It came to my conclusion that it really depends on the game whether this particular feature will hinder its performance or not.
This isn't meant to be a set-in-stone test as I did not complete many things. If other X1900 owners want to test my results with what they can get, that would be great since I am curious as to what other owners are getting. I wish I could continue some benchmarking, but I am on hold for a little bit as far as gaming goes. All of my findings are estimates that I have yet to specify, but I did want to make notice that BF2 performance can be substantially decreased by this feature while FEAR can be slightly increased. As far as what other games, maybe other X1900 owners could pitch in some of their findings for the moment.
Relevant System Hardware Specifications:
[*]X1900XTX @740/850(1700)
[*]AMD Athlon 64 4400+ Toledo
[*]DFI Lanparty UT SLI-DR Expert
[*]2x1GB G. Skill Extreme Edition DDR500 (PC4000)
[*]OCZ Powerstream 520W
Relevant System Software Specifications
[*]Windows XP 32-bit w/SP2
[*]Cat. 6.6 + Chuck Patch
[*]ATI Tray Tools
[*]ATI Tool
[*]BF2 and FEAR of course and patched to the most recent versions.
Please forgive me for not immediatedly providing specific frames as I do not have my normal computer up and running at this point in time. (Believe me, it isn't fun)
My driver settings were also the highest quality settings I could achieve minus a 6xAA setting and Vsync + Triple Buffering. Instead it was 4xAA set from the games and the Vsynce remained disabled for accurate frames.
I did three AAA (Adaptive Anit-Aliaising) settings: No AAA, Peformance AAA, and Quality AAA. (Note: In Tray Tools there is a setting that displays both "Quality" and "Quality with Trilinear". Just for curiousity's sake, I included the "Quality with Trilinear" setting in my initial testing. The difference between "Quality with Trilinear" and the normal "Quality" setting was within 1~2 frames and therefore I considered it a negligible difference.
BF2
In game settings were set to "4", a level that has to be hacked through the config files and results in an "Ultra-High" setting while in the BF2 video options. It was the absolute highest quality settings I could achieve through BF2 itself. Each simulation was done on Daqing Oilfields in Single Player Mode. Each time I traveled the same path of attack. Starting as U.S. and going towards the middle neutral bunker. I wanted to get a good mixture of chain-link fences and vast vegetation since AAA works intensely with those types of alpha textures. The chain-link fences present in the bunker provided a small amount of alpha textures and a decently heated battle as the U.S. and China sides clashed for the neutral flag. I only ran and moved in positions that were easy to recreate for each test and displayed close to the same amount of objects. After the bunker I would diagnoly make my way towards the edge of the solid fence and turn right so as to get a screen full of the higher ground covered with trees. (Note: this is where the peformance hit was felt the most throught any of the testings.)
Once again I apologize for not being able to provide exact frames at the moment and I will update my own findings once I can access my hard drive again, but I remember them within a 5~10 frame set as well as the performance hits they gave.
[*]No AAA
70~100 frames constantly.
[*]Performance AAA
65~100 constantly. I remember it was very close to a 5% decrease.
[*]Quality AAA
48~80 or 90ish. I vividly remember that this was a 25% performance decrease.
FEAR
I used the in-game benchmark for all tests. The benchmark was ran three times and averaged out to portray the average results for each specific setting. In-game settings were all maxed through the game's video options panel (no config file tweaking). Soft Shadows were off.
(All I can really remember was the average section, the middle one that displays the widest margin for the majority of the frames.)
[*]No AAA
~35-39 frames.
[*]Performance AAA
Almost identical to "No AAA". The difference was negligible and therefore I do not remember exact frames for it.
[*]Quality AAA
~45-49 frames. (Yes, I checked again and again. "Quality" AAA gives around a 10 frame increase to the X1900's in FEAR.)
It came to my conclusion that it really depends on the game whether this particular feature will hinder its performance or not.
This isn't meant to be a set-in-stone test as I did not complete many things. If other X1900 owners want to test my results with what they can get, that would be great since I am curious as to what other owners are getting. I wish I could continue some benchmarking, but I am on hold for a little bit as far as gaming goes. All of my findings are estimates that I have yet to specify, but I did want to make notice that BF2 performance can be substantially decreased by this feature while FEAR can be slightly increased. As far as what other games, maybe other X1900 owners could pitch in some of their findings for the moment.
Facebook
Twitter