Sparks off by default
Sparks off by default
Sparks absolutely kill my fps. I go from a steady 25-35 fps to 8 fps and below (in a local game, with one AI). That is really unacceptable, IMO. Maybe this is only a problem for Macs?
G4 800 mHz ATI Radeon 7500 Mobile (32MB VRAM).
G4 800 mHz ATI Radeon 7500 Mobile (32MB VRAM).
-
- Dr Z Level
- Posts: 2246
- Joined: Sun Mar 20, 2005 4:03 pm
- Location: IM: luke@dashjr.org
- Lucifer
- Project Developer
- Posts: 8640
- Joined: Sun Aug 15, 2004 3:32 pm
- Location: Republic of Texas
- Contact:
They don't hurt my fps at all. I get a steady 140fps with or without sparks. I think sparks give that really nice out of the box smell and should be kept on, and I also think it should be easier to turn off some of the details, maybe with some keybindings? Or a special detail menu that can be turned on on its own in-game to make it easier for new players to turn it off?
When I get the particle system up and running (for Bacchus), sparks should get a significant boost in performance, so I'd just as soon leave them on right now and revisit the issue after Artemis has been released.
When I get the particle system up and running (for Bacchus), sparks should get a significant boost in performance, so I'd just as soon leave them on right now and revisit the issue after Artemis has been released.
Perhaps it's an ATI thing rather than a MAC thing? Provided Jonathan was using the old 0.2.7.1 codebase where there is known wasted CPU time due to the object-object-interaction which should be gone in current CVS.
Has anyone rough performance figures for PCs with ATI or Macs with NVidia?
No matter how it turns out to be system dependant, we should disable sparks on those system where they are known to be problematic for now and reenable them when Lucifer's particle system sparks are ready.
I don't think adding hotkeys for detail settings makes things easier for the novice. Likewise for the extra menu hotkey. But maybe that's just me.
Nemo: I find the general FPS you get strangely low. With just one AI and at sufficiently low resolution so that fillrate is not the limiting factor, your CPU should be capable of at least 200 fps in the beginning of the match (Less after you let Jonathan's LRBot run for a while). Is this still in debug mode?
Has anyone rough performance figures for PCs with ATI or Macs with NVidia?
No matter how it turns out to be system dependant, we should disable sparks on those system where they are known to be problematic for now and reenable them when Lucifer's particle system sparks are ready.
I don't think adding hotkeys for detail settings makes things easier for the novice. Likewise for the extra menu hotkey. But maybe that's just me.
Nemo: I find the general FPS you get strangely low. With just one AI and at sufficiently low resolution so that fillrate is not the limiting factor, your CPU should be capable of at least 200 fps in the beginning of the match (Less after you let Jonathan's LRBot run for a while). Is this still in debug mode?
- Tank Program
- Forum & Project Admin, PhD
- Posts: 6711
- Joined: Thu Dec 18, 2003 7:03 pm
Tank: Yes, that's why I asked for low resolution results. Besides, according to
http://www6.tomshardware.com/mobile/20020408/index.html,
the 7500M plays in the same league as the Gf4Go which is what I have, and it gives me quite reliable 50fps (my refresh rate, I guess) in 800x600. With a P4M 1.8Ghz and about 20% CPU load. Without sparks. They don't cost me much on average, but cause an occasional frame drop. So I'd still expect the 7500M to perform a tad better, unless the one nemo got is one of those shared memory things, but they get a fancy sounding name attachment AFAIR.
Iceman: Yes, but you're still using a version that has the spark performance bug (0.2.7.1 and earlier) and a snailspeed CPU where that bug really hurts, right? And while I've got your attention: you were reporting strange freezers in 0.2.7.1; I think it would help us if you got one of the latest snapshots and make a debug recording of one or two of them. Then we could see better whether there's a correlation between the freezing and anything we do.
http://www6.tomshardware.com/mobile/20020408/index.html,
the 7500M plays in the same league as the Gf4Go which is what I have, and it gives me quite reliable 50fps (my refresh rate, I guess) in 800x600. With a P4M 1.8Ghz and about 20% CPU load. Without sparks. They don't cost me much on average, but cause an occasional frame drop. So I'd still expect the 7500M to perform a tad better, unless the one nemo got is one of those shared memory things, but they get a fancy sounding name attachment AFAIR.
Iceman: Yes, but you're still using a version that has the spark performance bug (0.2.7.1 and earlier) and a snailspeed CPU where that bug really hurts, right? And while I've got your attention: you were reporting strange freezers in 0.2.7.1; I think it would help us if you got one of the latest snapshots and make a debug recording of one or two of them. Then we could see better whether there's a correlation between the freezing and anything we do.
- Jonathan
- A Brave Victim
- Posts: 3391
- Joined: Thu Feb 03, 2005 12:50 am
- Location: Not really lurking anymore
It was CVS around 0.2.7.2.pre1, with broken winzone. I stopped updating because I started putting more time in doing non-Armagetron things and less in fixing stuff that broke (like CVS merging after trying to change little things throughout the whole game. Also at that time some big changes were made).z-man wrote:Perhaps it's an ATI thing rather than a MAC thing? Provided Jonathan was using the old 0.2.7.1 codebase where there is known wasted CPU time due to the object-object-interaction which should be gone in current CVS.
I have access to a 2.6 GHz P4 with desktop Radeon 9600 running Windows. Without sparks performance is fairly similar (PC running 0.2.7.1, Mac running my version without tweaks like no glFinish) but with sparks the PC is so much faster. But since removing all sparks code (including Render() of course!) doesn't make it faster I wouldn't suspect the GPU.Has anyone rough performance figures for PCs with ATI or Macs with NVidia?
Actually it was called rlbot.(Less after you let Jonathan's LRBot run for a while)
ˌɑrməˈɡɛˌtrɑn
I don't think its an ATI thing, because the same happens on a Powerbook G4 1GHz with a nvidia geforce4mx. With sparks on, the game is really unplayable for me, and I think it may be turning away a few new Mac users because they are on by default. Now would be a good time for a settings change on the mac side, because everyone will get default settings with 0.2.8.0 (user.cfg location has changed).z-man wrote:Perhaps it's an ATI thing rather than a Mac thing? Provided Jonathan was using the old 0.2.7.1 codebase where there is known wasted CPU time due to the object-object-interaction which should be gone in current CVS.
Has anyone rough performance figures for PCs with ATI or Macs with NVidia?
No matter how it turns out to be system dependant, we should disable sparks on those system where they are known to be problematic for now and reenable them when Lucifer's particle system sparks are ready.
I don't think adding hotkeys for detail settings makes things easier for the novice. Likewise for the extra menu hotkey. But maybe that's just me.
With out sparks: 110 fps in the beginning of the round, but after 10 seconds fps has dropped down to 30. With sparks: 110 fps in beginning of the round, dropped to 30 after 2 seconds, and I am not in debug mode. Oh, and the card is not using shared memory.z-man wrote:Nemo: I find the general FPS you get strangely low. With just one AI and at sufficiently low resolution so that fillrate is not the limiting factor, your CPU should be capable of at least 200 fps in the beginning of the match (Less after you let Jonathan's LRBot run for a while). Is this still in debug mode?
I am not saying my card is the best, but the difference with sparks and without sparks is huge on my computer.Tank Program wrote:This is a 7500 you know, not exactly known for stunning 3D performance... IMO, it's just a case of the old hardware blues.
So, maybe the poll should be changed to "Should sparks be turned off on Macs by default?".
The evidence gathered so far strongly suggests this. The relevant code is the crash_sparks definition in gCycle.cpp.nemostultae wrote:So, maybe the poll should be changed to "Should sparks be turned off on Macs by default?".
Jonathan: is your Mac running OSX or some flavor of Linux? (Read: should the decision be based on the OS type or the architecture?)