Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Motigg]

Alright since theres been quite a few posts lately trying to figure out which SDL mode the best, I figured I'd post my findings and open anyone else to do the same just so that there is at least some kind of concensus as to whats most likely to be better should someone ask.

Inspiration for this post came from someone asking if .31.06 seemed slower to which someone responded try the different modes, but .31.06 does seem a little slower then the previous two releases while still being faster then .31.03.

Relevant information:

Video Card: Nvidia GeForce 6150 SE (Crap I know)
Processor: AMD Athlon 64 x2 dual-core processor 3800+ (2 GHz dual-core)
OS: Windows 7 x64

I'm also using the Genesis mod by Deon with the Ironhands graphic pack.
It is a 3x3 embark with a volcano in dead center, underground has not been breached.
The Fortress is 4 years old with 122 dwarfs (only 2 idle) and 100 animals (20 of them pets... was up to 150 before I started culling) on top of having 7500 stones lying around.  I kinda let things get a little outta hand when starting out.
FPS is capped at 200 while GFPS is capped at 10 for more speed.
Temp/Weather are both on.

These numbers were found with just FF with this window open, Task manager, the df folder and the init file.

Standard: 35ish FPS, would range from 33-37
2D: 38 FPS, 37-39 range
2DSW: 38 FPS, 37-40 range,
2DASYNC:39 FPS, 37-40 range
Partial:2: About 40 FPS on load screen flashes alot so I didn't test very long.  Partial:1 and Partial:6 did not work for me.
ACCUM_BUFFER: Lots of flucuation from 33-40, tended to stay lower 30's more often then not
FRAME_BUFFER: 36 FPS, 35-38 range
VBO: 37 FPS, 35-40 range

In conclusion, it appeared that 2DASYNC is the best mode of average although for some reason on fresh embarks VBO beat all other modes when I tried it orginally.  Another interesting thing worth noting was during all tests, DF did not use more then 52% of the available processing power which was about the same as .31.03 while .31.04-5 used around 60-65%.  My guess is that for whatever reason none of the modes are utilizing the second processor even though I'm pretty sure Baughn said that it should for graphics.

Edited to reflect recent findings.

[Untelligent]

Interesting data. Now we need to figure out if Partial is better than the rest, for those it works for.

[Motigg]

Well I went back and tried it again, got Partial:2 to work ran at about 40 FPS on load but it flickers as well as not being able to resize at all else the screen goes black, Partial:1 just flickers and Partial:6 was just a black screen. I'd try more but it just doesnt seem useful for me at least.

[Motigg]

Shameless bump

[xtank5]

Have you tried setting the priority higher?  If you make it a hi-priority process, it will use more of the power available to it. 
You should also test it with it running on just one core.  This can be achieved through the process manager within the task manager.  Simply right-click, set affinity, and change it so DF is the only one on a single core.
Also, how are you testing the FPS?  Do you load the save, let it run for a while and then save?  Or do you let it run and then end the process so as not to alter the save file?

By having as much information as possible, we can get the most varied results to compare data to while using the same test-bed. 

[Motigg]

Yea, sorry forgot about some of that stuff.  I was really tired when writing the post up that I forgot to keep the save and such.

I run DF on high priority at the time I was testing.  I was opening the save and letting it run till one of my dwarfs started a construction from a mood. I'd then end the process and restart from the beginning.  I made sure to not touch it the entire time to make sure that everything happened as close to the same as possible.  I had also stopped my mass butchering of animals just to make sure that did not affect FPS during the testing.  Each test ran for about a minute and I averaged the FPS.  I made sure that I had no extra processes running (I nuked a couple processes that aren't really needed but took up some processing power).  I could try running it on one core if I could find a time close to when i was getting these numbers, I hadn't even thought of that.

Ideally if someone has a large fort I could run the tests again with default ascii as well as other graphic sets.  I don't really mind devoting time to it if it helps inform people.  I backup every season as well so I'm sure I could go through on find out around what time it was rather easily to upload the save at a relatively close time.

[Ninjabob]

On my computer Partial Print helps... I don't have any statistics as of now, but I could come back later today and drop some off. In the last version of DF:

My main computer usually averaged on 100fps when embarking, 40-50fps when 80-100 dwarfs, lower framerates with more dwarves and complexity. Partial print seemed to not change it at all. In fact, most settings didn't matter much.

On my laptop, I think I used to embark with aprox 50fps without partial print, think it was close to a hundred fps with PP on, but it fell quickly with more dwarves. With 30-40 dwarves it was usually down to about 40fps, 50-70 it was generally in its twenties, and so on.

Oh and according to the wiki, Partial Print is turned off in v31.04

[Motigg]

Per request I went back and tried it using just one core.  I only did it with DASYNC since that was the one that I had originally reported with the highest FPS.  I found a save relatively similar to the one I had been using (no traders like before 121 dwarves though and some miasma still around) that had given me about the same FPS as before with both cores set to run DF.  Having only one core running DF did not seem to affect DF's FPS although I did experience a spike when there were announcements for like 1 frame (coulda been the fact that like 3 went off at one time).

I attempted Partial:3 but that also doesn't work for me.

@Ninjabob:  Yea thats right, in 31.04 Baughn had said somewhere that it was in yet or something to that a affect I believe he added it in in 31.05 but I really just used 31.05 to test some military stuff not really having a serious fort.  It wasn't till 31.06 that I noticed there seemed to be a slight drop or so it appeared.