Is it possible to speed this up?

Sirius · Post by **Sirius** » Sat Jun 25, 2005 11:43 pm

Running Descent 3 timetest under OpenGL (D3D has slower average although a better minimum), medium detail settings, vsync and bilinear filtering off, 800x600, mip mapping on...

44.86 Descent3 v1.4
2 Min
122 Max

I have the rest of it but it's 80-some seconds so it's quite long. The minimum figure I think is inaccurate; lowest I saw was 12, but that's nothing to be proud of either.

System is a P3 667, 768 MB PC133 SDRAM, Radeon 7500 64MB, Windows XP SP2...

I find it strange that a machine with even these specs - significantly over recommended for Descent 3 - can give these sorts of results. It's also worth noting that this is the single player timetest. In less well-designed multiplayer levels with heavier fire, it can and frequently does drop below what must be 5 fps.

I can play with this, but I'd rather not have to... are there any strange things I can try to improve the performance (besides buying any hardware, that's not really an option)?

Duper · Post by **Duper** » Sun Jun 26, 2005 12:45 am

what do you have your -"framecap xxx" set at? That might help. No guarintees. :\

Sirius · Post by **Sirius** » Sun Jun 26, 2005 1:30 am

0. The maximum framerate was, not surprisingly, not as fast before... but performance is still quite variable.

SuperSheep · Post by **SuperSheep** » Sun Jun 26, 2005 8:44 am

-nomusic -nomultibmp -width 1024 -height 768 -framecap 200 -nomotionblur -nosparkles

This is the command line I used for a long time. When I used to have a 3dfx Voodoo card, setting the resolution lower helped, but when I got a NVidia card, I found setting it lower actually hurt performance.

Make sure you set your framerate in windows to as high as you can for the same resolutions you use for d3.

Don't forget to add -fastdemo to your commandline for timetests.

Disable any services that you may have running, and check your IRQ sharing and see how many devices are sharing the sound and video interrupts. It is possible this could be a soundcard issue. Test that by adding -nosound and running the test again.

WarAdvocat · Post by **WarAdvocat** » Sun Jun 26, 2005 10:10 am

-fastdemo is part of -timetest. No need for the extra command line parameter.

we lost a lot of old threads with the upgrade, but here's one with some good info, mined from the DBB with the search function:
phpbb/viewtopic.php?t=1006

That 2 FPS minimum is horrid. I was playing on a borrowed machine one time and I was getting that frequently in dogfights. It made me die a lot. No fun at all.

Mobius · Post by **Mobius** » Sun Jun 26, 2005 2:34 pm

Don't forget, online, 2 FPS = add 500ms of lag.

Krom · Post by **Krom** » Sun Jun 26, 2005 3:05 pm

D:\Games\Descent3\main.exe -launched -framecap 0 -nointro -nosparkles -nomotionblur -deadzone0 0.06 -deadzone1 0.06 -nomusic -usesmoothing -width 1280 -height 960

My standard multiplayer command line. Additionally, in the video drives I have 4x FSAA and 8x Aniso filter enabled. In detail settings everything but dynamic lighting is disabled (Though only weapon effects is the one I really need turned off). The timetest reading 2 FPS is probably something loading in the background, make sure there are no programs running that would steal CPU cycles from the game.

Nosferatu · Post by **Nosferatu** » Sun Jun 26, 2005 3:53 pm

I agree with krom. I know Win2k had a whole bunch of services that are turned on by default that are not needed and knowing M$ bloatware, its probably only gotten worse in XP.

BTW my argument list:
-nointro -framecap 200 -usesmoothing -launched -pilot NOSFERATU -vsync

The framecap is in there more or less just to ensure its not using the game default of 50 so the vsync will kick in. I hate screen tearing and this eliminates it. I think Im more accurate when there is no screen tearing.

Mobius · Post by **Mobius** » Sun Jun 26, 2005 8:58 pm

I've never really noticed screen tearing before - even at quite low frame rates. Yes, VSYNC was disabled in both game and drivers when I have tried to detect tearing.

Personally, I'm all for just letting the back and front buffers go crazy, and let the GFX card draw whatever frame it wants.

Never did understand the desire to sync the frame buffer swap with the screen fly-back. But I spose that's because I never see tearing....

Sirius · Post by **Sirius** » Sun Jun 26, 2005 9:58 pm

Rather curious... using -nosound parameter,
86.23 Descent3 v1.4
29 Min
168 Max

Minimum still isn't -good-, but it's not half as bad by any means... makes me wonder whether EAX absolutely kills performance.

Edit: Will try a few more things (such as disabling other services) later...

Nosferatu · Post by **Nosferatu** » Sun Jun 26, 2005 10:20 pm

Sirius wrote:Rather curious... using -nosound parameter,
86.23 Descent3 v1.4
29 Min
168 Max

Minimum still isn't -good-, but it's not half as bad by any means... makes me wonder whether EAX absolutely kills performance.

Edit: Will try a few more things (such as disabling other services) later...

Hmmmm. This could possibly be another one for dxdiag.
Try doing a start->run->dxdiag
Let it do its checking then go to the audio tab and notch down the audio acceleration one notch.

Krom · Post by **Krom** » Sun Jun 26, 2005 10:44 pm

EAX does kill performance on slower CPUs, it eats up a lot of CPU time in D3, use directsound and NO 3D SOUND. That should get your framerates up a lot closer to constantly playable levels.

SuperSheep · Post by **SuperSheep** » Mon Jun 27, 2005 7:22 am

When I built my home system, I bought a lot of top notch components(at the time

) but everytime I enabled EAX, it would cut my frametimes in half. I had an Audigy card, then an Audigy 2 card and both exhibited this. That was with a P4 2.5GHz, and a GeForce TI4600.

Software or DirectSound is the best choice until you can get your overall framerates up with a faster CPU.

Sirius · Post by **Sirius** » Mon Jun 27, 2005 7:32 am

Although at this time of day multiplayer is unlikely to say the least, I have playtested the current setup in the first three levels of Descent 3. It seems now I tend to get an average of 60-90 FPS, dropping only about as low as 30 in dogfights and outside, which is much more acceptable (I know a few people here would hate it, but it isn't really a problem for me).

So, this seems to have had quite an impact. Thanks for the input guys.

WarAdvocat · Post by **WarAdvocat** » Mon Jun 27, 2005 11:11 am

That sounds more like it, for your system.

By the way, thanks for the inspiration! I re-tweaked my system, and got my timetests down to 11 seconds with about 200/450 min/max. Of course, that's on an AMD 64 3000+ / Geforce FX 6600 GT system.

Krom · Post by **Krom** » Mon Jun 27, 2005 2:05 pm

Don't let me scare you off War, but I could probably get a 5 second timetest out of your system.

Done it before on mine, 500+ FPS averages and all with my XP 1700+ and FX 5900 ultra. =)

Timetest wrote:530.73 Descent3 v1.4
420 Min
707 Max
688 Second 1
707 Second 2
439 Second 3
478 Second 4
420 Second 5

Deadmeat · Post by **Deadmeat** » Mon Jun 27, 2005 4:04 pm

XP3200+ with a Radeon Soft Mod 9700Pro.

433.50 Descent3 v1.4 345/325=core/mem
322 Min
632 Max
500 Second 1
632 Second 2
518 Second 3
322 Second 4
390 Second 5
350 Second 6
387 Second 7

Yee Haw, I beat the Bunster.

Grendel · Post by **Grendel** » Tue Jun 28, 2005 10:37 pm

Well, it's all relative. I have a 6800U and my timetest reads:

308.78 Descent3 v1.4
245 Min
345 Max
292 Second 1
339 Second 2
345 Second 3
334 Second 4
330 Second 5
298 Second 6
297 Second 7
310 Second 8
313 Second 9
295 Second 10
314 Second 11

Pretty bad ? Not really, this is at 1280x1024 w/ all D3 eye candy, max detail, deep render, 4xAA, 8xAF and Quality setting. I don't need more FPS on a TFT w/ 60Hz refresh rate

Without qualifiers a FPS value doesn't really say much (unless it's in the <20 realm -- that's bad in any case).

Edit: timetest cmd line:
-nosparkles -nomotionblur -width 1280 -height 1024 -bumped -z32bit -framecap 0 -nocompress -nosound -nonetwork -timetest secret2.dem

BUBBALOU · Post by **BUBBALOU** » Wed Jun 29, 2005 5:53 am

Seriously people any more FPS than your refresh rate is just an ego booster in D3. You can have 2000FPS in D3 if you fly like a brick does it matter?? Yet the guy running 60FPS is easting you up with the vauss.

I have my both of my D3 systems locked at 100 FPS they never faulter!!

XP Intel 2.53Ghz/512PC2700/GF4TI4200
98SE AMD K7600/256PC133/3dfxV3500

BTW ~ Grendel why are you running your timetest is D3D on a 6800U? You should be running the timetest in OGL and need to remove those D3D only switches from your command line!!

[poke]

Grendel wrote:-nosparkles -nomotionblur -width 1280 -height 1024 -bumped -z32bit -framecap 0 -nocompress -nosound -nonetwork -timetest secret2.dem

WarAdvocat · Post by **WarAdvocat** » Wed Jun 29, 2005 6:06 am

Yeah, Krom, I could probably add some frames too. I have all the bells and whistles set to 'on'. I'm content with middling fast.

SuperSheep · Post by **SuperSheep** » Wed Jun 29, 2005 7:18 am

BUBBALOU wrote:Seriously people any more FPS than your refresh rate is just an ego booster in D3. You can have 2000FPS in D3 if you fly like a brick does it matter?? Yet the guy running 60FPS is easting you up with the vauss.

Actually, FPS does matter just not for the reasons you might think. 100 fps = 10ms lag. That lag is for everything, controller input, sound effects, packet handling, etc. Getting a 2000fps for a fraction of a second doesn't mean much, but if you could hold 500fps because of it, well, that does matter.
I wouldn't let my system run uncapped, because I like to deal with a stable lag. Hence, the framecap 200, but that only is possible because my timetest has a minimum higher than 200.

Krom · Post by **Krom** » Wed Jun 29, 2005 9:27 am

WarAdvocat wrote:Yeah, Krom, I could probably add some frames too. I have all the bells and whistles set to 'on'. I'm content with middling fast.

With everything on, I still push over 400 average. But most of the eye candy special effects get in my way when I'm playing so I prefer them off.

WarAdvocat · Post by **WarAdvocat** » Wed Jun 29, 2005 10:06 am

Deadmeat wrote:Yee Haw, I beat the Bunster.

That's a first. You'll get yours real soon now

Krom: Are you suggesting that there's something elementary that I'm missing? I was assuming that you were dealing with a water-cooled, ludicrously overclocked system. If there's something short of that I could do to speed things up, I'd be interested

Hmm. I'm running @ 800 x 600 currently due to text issues. This is soon to change. Could the low res be a piece of the puzzle? What else am I missing, oh wizard oh wise one?

Krom · Post by **Krom** » Wed Jun 29, 2005 12:45 pm

From what I have seen, 1024x768 is faster then 800x600, even on my old GF3. Second, 530 FPS was without sound enabled. My CPU is water cooled and overclocked to 2.4 GHz, but thats about what it would take for a Athlon XP to match a Athlon 64 at 2 GHz. I also stripped off extra tasks that I didn't need for D3.

But essentually all I do is boot up, exit all the tasks I don't need, disable sound, and run the time test with as little open as possible.

Right now at well over a week since the last reboot, with a half dozen applications open including a CPU/memory hog like bit torrent, I only hit into the 200 FPS average at 1280x960 with 4x FSAA and 8x Aniso enabled with -nosound. Course there is a big nasty texture load of some kind going on for the first few seconds that bogs my FPS down as low as 20.

If I rebooted clean, cut all the background tasks and reset my screen resolution/AA settings to something more balanced with the CPU performance I could double my FPS easily.

I could probably tweak 500 out of your system in a few minutes, just like I can get it out of my system, but that's not how I use my computer most of the time. Which is why I could use a dual core.

SuperSheep · Post by **SuperSheep** » Wed Jun 29, 2005 2:14 pm

WarAdvocat wrote:Hmm. I'm running @ 800 x 600 currently due to text issues. This is soon to change. Could the low res be a piece of the puzzle? What else am I missing, oh wizard oh wise one?

...

SuperSheep wrote:This is the command line I used for a long time. When I used to have a 3dfx Voodoo card, setting the resolution lower helped, but when I got a NVidia card, I found setting it lower actually hurt performance.

BUBBALOU · Post by **BUBBALOU** » Wed Jun 29, 2005 6:47 pm

Yeah I know about that too SS.(wuzzup!).. brings back old memories of the old Q2 FPS/Network rate/Ping bug. to the point your FPS would make you run in place on a P C/S

As Far as EAX/CMSS 3D Issues - Disabled except for DVD's, Movies, and Music!

Just a little quick one

Hit CTRL-SHFT-ESC if you are running your XP system lean you should only have 36-42 processes running (w/AV) and having a couple of appz chat and browser windows open

Krom · Post by **Krom** » Wed Jun 29, 2005 8:06 pm

I have 32 processes at normal use, extras I don't need: trillian, taskman, firefox, boinc (x2), daemon tools, winamp, media player classic, net limtier. My mouse driver also runs two processes, and ultramon uses two processes, nview keeps an instance of rundll32 going, disk keeper has a scheduler that runs one process always, my nvidia display drivers are running a process and the audio drivers another.

17 extra processes past what windows needs.

FunkyStickman · Post by **FunkyStickman** » Thu Jun 30, 2005 12:54 pm

Just out of curiosity, has anybody tried to run a timetest in Linux? I ran the test, and it worked fine, but I have no idea where the framerate info is output to... didn't find anything on it either here at DBB or in the READMEs.

I wouldn't have anything to write home about, I'm running a Duron 1.3Ghz and a GeForce 4 MX, 512MB of SDRAM. Still, it would be nice to see what kind of framerate I can get, and compare that to running it in Windows.

Grendel · Post by **Grendel** » Thu Jun 30, 2005 1:02 pm

D3 for win writes the file fps.txt to the D3 root directory.

FunkyStickman · Post by **FunkyStickman** » Thu Jun 30, 2005 1:08 pm

Gotcha. I finally found it, after a little looking. It creates fps.txt in the ~/.loki/descent3 directory, which would make sense if I had thought about it some more.

Just for giggles, I made:
88.78 Descent3 v1.4
41 Min
117 Max
running at 1280x960, in windowed mode, AA and AF on. Not bad for old hardware at high res. It's at least playable!

Krom · Post by **Krom** » Thu Jun 30, 2005 1:40 pm

Actually it would be more interesting if you dropped the res a bit to like 1024x768 and ran the timetest without AA to see how fast D3 can run on linux when you remove some of the video limitations. A Duron 1.3 should be able to reach into the 200-250 FPS range in windows.