Posters Name: tmmicklabs
Posters Email: tmmicklabs@mwgl.org
Subject: Dungeon Siege Tweaking Guide
News:
Ah yes, Dungeon Siege. You either love it or you hate it. I'm hooked harder than a bad smack habit. As some of you may have noticed, Dungeon Siege brings most PC's to their knees. I've been working vigorously for the past month trying to get Dungeon Siege to run smoothly on my PC. It's been a real challenge getting this game to play smoothly, but after a month of blood, sweat, and tears I've finally figured it out. Find out how I got a 75 % increase in performance in Dungeon Siege.
Full Story:
Ok kids, here we go.
Step 1 - You need to have Dungeon Siege installed.
Step 2 - Click here to download the official Dungeon Siege Benchmark.
Step 3 - Install and run the Benchmark.
Step 4 - Go to the folder that Dungeon Siege is installed in and open "fps.log" to view the results.
Here are the Specs of the Test PC.
Athlon XP 1600+
Epox 8KHA+ Motherboard with KT266a core logic chipet with Via 4in 1 driver version 4.38
256mb PC-2100 DDR-Ram
ATI Radeon 8500 64mb video card using 6071 reference drivers.
Sound Blaster Live
Windows XP Pro
Here are the game settings I ran the benchmark at.
1024/768 32bit color
Object Detail at Max Setting
Trilinear Filtering enabled
Shadows set to "all complex".
Here are the results.
FPS Summary
-----------
Total Sample Time : 192.80 s (6454 frames)
Min Frame Delta : 5.68 ms (175.94 fps)
Max Frame Delta : 160.58 ms (6.23 fps)
Average Frame Delta: 29.87 ms (33.47 fps)
This was unacceptable in my opinion. I set out to find a way to get the performance at a playable level.
The first thing I did was set the shadows setting from "all complex" to "none".
FPS Summary
-----------
Total Sample Time : 192.74 s (8146 frames)
Min Frame Delta : 4.33 ms (230.76 fps)
Max Frame Delta : 166.67 ms (6.00 fps)
Average Frame Delta: 23.66 ms (42.26 fps)
WOW !!! A 27 % increase by disabling shadows. Trust me, you won't miss the shadows.
The next thing I did was set the filtering method from "trilinear" to "bilinear".
This showed an increase of 2 frames per second. I did not notice any sacrifice in image quality. This may offer an even larger increase in performance on older video cards. Give it a try.
FPS Summary
-----------
Total Sample Time : 192.58 s (8460 frames)
Min Frame Delta : 4.30 ms (233.40 fps)
Max Frame Delta : 166.67 ms (6.00 fps)
Average Frame Delta: 22.76 ms (44.93 fps)
The next step was to change the resolution from 1024/768 32-bit to 1024/768 16-bit.
This only an increase of less than one frame per second. Definetly not worth them sacrafice in image quality on newer video cards. May help on older Video cards though.
FPS Summary
-----------
Total Sample Time : 192.70 s (8578 frames)
Min Frame Delta : 4.33 ms (231.76 fps)
Max Frame Delta : 134.64 ms (7.43 fps)
Average Frame Delta: 22.46 ms (45.52 fps)
As you may have noticed from above, EAX is disabled. For this test I enabled "EAX" under the audio option. This gave me a decrease of 11 frames per second. This is a huge decrease. Disabled EAX again to make the game playable.
FPS Summary
-----------
Total Sample Time : 192.98 s (6592 frames)
Min Frame Delta : 4.52 ms (221.41 fps)
Max Frame Delta : 159.19 ms (6.28 fps)
Average Frame Delta: 29.27 ms (34.16 fps)
The final step I did was with the Dungeon Siege Video Configuration Tool. You can access it from start\programs\dungeonsiege. My Rendering Device was set to "Radeon 8500\8500LE Hardare TnL" I set the Rendering Device to "Radeon 8500\8500LE Hardare " expecting a decrease in performance. Boy was I wrong. Oh my Heck !!! A 38 % increase in performance. This is huge.
I can now play this game without cursing and gnashing my teeth. Looks like the Harware TnL function in Dungeon Siege does not like to play nice with my Ati Radeon 8500. I don't have a Nvidia Geforce card at this time to furtner test the TnL theory with.
FPS Summary
-----------
Total Sample Time : 192.63 s (11182 frames)
Min Frame Delta : 3.98 ms (251.10 fps)
Max Frame Delta : 166.67 ms (6.00 fps)
Average Frame Delta: 17.23 ms (58.05 fps)
So there you have it. A 75 % increase in performance with a few easy clicks of the old mouse. Now it's time to move on to part 2 of the test.
For the next test I took out the ATI Radeon 8500 and put a 3dfx Voodoo 5 5500 in its place.
Here are the Specs of the Test PC again.
Athlon XP 1600+
Epox 8KHA+ Motherboard with KT266a core logic chipet with Via 4in 1 driver version 4.38
256mb PC-2100 DDR-Ram
3dfx Voodoo 5 5500 64mb running the lastest drivers from 3dfx underground...ver 1.09 beta 2
Sound Blaster Live
Windows XP Pro
Here are the game settings I ran the benchmark at.
1024/768 32bit color
Object Detail at Max Setting
Trilinear Filtering enabled
Shadows set to "all complex".
Here are the results.
FPS Summary
-----------
Total Sample Time : 193.80 s (3350 frames)
Min Frame Delta : 5.94 ms (168.47 fps)
Max Frame Delta : 166.67 ms (6.00 fps)
Average Frame Delta: 57.85 ms (17.29 fps)
Next Step, disable EAX. Oh my Heck. Only a 2 frame per second increase. This makes no sense. Disabling EAX on the Radeon 8500 gave me an 11 frame per second increase. Something fishy is going on here.
Total Sample Time : 193.21 s (3716 frames)
Min Frame Delta : 5.64 ms (177.36 fps)
Max Frame Delta : 166.67 ms (6.00 fps)
Average Frame Delta: 51.99 ms (19.23 fps)
Step 3, set shadows from "all complex" to "none". Cool, a 7 frame per second increase. Now we are getting somewhere.
FPS Summary
-----------
Total Sample Time : 192.79 s (5040 frames)
Min Frame Delta : 6.00 ms (166.55 fps)
Max Frame Delta : 166.67 ms (6.00 fps)
Average Frame Delta: 38.25 ms (26.14 fps)
Step 4, change filtering method from "trillinear" to "billinear". Yeah baby, another 3 frames per second. Getting us Ohhh so much closer to the magical mark of smoothness.
Total Sample Time : 192.96 s (5709 frames)
Min Frame Delta : 5.58 ms (179.06 fps)
Max Frame Delta : 166.67 ms (6.00 fps)
Average Frame Delta: 33.80 ms (29.59 fps)
Step 5, change the rendering bit depth from 32-bit to 16-bit. Something magical is happening. OH MY HECK !!! An increase of 10 frames per second. This is huge. Dungeon Siege is smooth like butta now. The image quality is still good. If you look close you can notice that the fire from the dragon does not look as good as it did with 32-bit rendering though. I can deal with it though.
Total Sample Time : 192.77 s (7583 frames)
Min Frame Delta : 6.00 ms (166.76 fps)
Max Frame Delta : 161.05 ms (6.21 fps)
Average Frame Delta: 25.42 ms (39.34 fps)
So there you have it boys and girls. A 129 % increase in performance. This is huge. This game is now playable on the good ol' Voodoo 5. Now let's move on to part 3 of this test.
Here come the Nvidia Geforce 2 results you've all been waiting for.
Here are the Specs of the Test PC again:
Athlon XP 1600+
Epox 8KHA+ Motherboard with KT266a core logic chipset with Via 4in 1 driver version 4.38
256mb PC-2100 DDR-Ram
Nvidia Geforce 2 GTS 32mb running ver. 29.60 Detonator drivers
Sound Blaster Live
Windows XP Pro
Here are the game settings I ran the benchmark at:
1024/768 32bit color
Object Detail at Max Setting
Trilinear Filtering enabled
Shadows set to "all complex"
Hardware TnL = Enabled
EAX = Enabled
Here are the results.
FPS Summary
-----------
Total Sample Time : 193.17 s (5297 frames)
Min Frame Delta : 3.75 ms (266.41 fps)
Max Frame Delta : 166.67 ms (6.00 fps)
Average Frame Delta: 36.47 ms (27.42 fps)
27 frames per second. Not to shabby for a card that is 2 years old. Looks like this card has some potential. The first thing I did was disable Hardware TnL. Doing this helped tremendously on the Ati Radeon 8500.
FPS Summary
-----------
Total Sample Time : 193.07 s (6394 frames)
Min Frame Delta : 3.97 ms (252.06 fps)
Max Frame Delta : 166.67 ms (6.00 fps)
Average Frame Delta: 30.20 ms (33.12 fps)
Wow. 5 frames per second increase. Now we are getting somewhere. Looks like the TnL code is broken in Dungeon Siege due to the fact that TnL is slower on a Radeon card and a Geforce card. Now lets disable EAX and see what happens.
FPS Summary
-----------
Total Sample Time : 192.57 s (7599 frames)
Min Frame Delta : 3.87 ms (258.52 fps)
Max Frame Delta : 166.67 ms (6.00 fps)
Average Frame Delta: 25.34 ms (39.46 fps)
Cool. Another 6 frames per second by disabling EAX. Now lets turn the shadows off and see what happens.
FPS Summary
-----------
Total Sample Time : 192.46 s (11241 frames)
Min Frame Delta : 3.79 ms (263.73 fps)
Max Frame Delta : 166.67 ms (6.00 fps)
Average Frame Delta: 17.12 ms (58.41 fps)
Can we say OH MY HECK ? 19 bloody frames per second by turning off the shadows. This is huge. Now we are dangerously close to the magical 60 frame per second mark =-) Now lets set the Filtering method from Trilinear to Bilinear and see what happens.
FPS Summary
-----------
Total Sample Time : 192.49 s (11440 frames)
Min Frame Delta : 3.84 ms (260.58 fps)
Max Frame Delta : 166.67 ms (6.00 fps)
Average Frame Delta: 16.83 ms (59.43 fps)
1 frame per second. I personally cannot tell the visual difference between Bilinear and Trilinear. Finally, I switched the color depth from 32-bit to 16-bit.
FPS Summary
-----------
Total Sample Time : 192.42 s (12316 frames)
Min Frame Delta : 3.81 ms (262.67 fps)
Max Frame Delta : 166.67 ms (6.00 fps)
Average Frame Delta: 15.62 ms (64.01 fps)
5 more frames per second. The visual quality is not as good on the fire and spell effects though. I personally would put it back to 32-bit rendering. It's up to you, though.
So there you have it. A 137 percent increase on performance on a good ol' Geforce 2 GTS. Hope this article helps all of you out there with frame rate issues. If you need any help or have any input, drop me a line at Tmmick2000@MWGL.ORG. Happy benchmarking.