%limit_fps bug/suggestion

[LabRat]

I am not sure how to categorise this one, or where would be the best place to put it - here or bug reports..

 

I am running a nice fast system (AMD 3300+, ATI x700 256MB), and have set VSYNC to 'on' in the ATI OpenGL control panel.

 

While running the game with %limit_fps 0 the CPU usage is as near to 100% as the game can manage without crashing Windows, and giving a constant 50FPS (I am running on TV out until I replace my exploded monitor).

 

If I %limit_fps 50 my CPU usage drops to around 12%-20% - it appears that frames are still being drawn even though the screen doesn't render them.

 

Upgrading the SDL libs to 1.2.11 introduces a new attribute:

SDL_GL_SWAP_CONTROL

which allows the program to set/unset VSYNC.

 

I can't see how to get the monitor's current refresh rate using SDL anywhere on google, but if a way could be found then setting %limit_fps to match that should reduce processor usage immensely when activated.

 

In the meantime, maybe forcing %limit_fps to a max of 100 or 85 or 72 or 60 (72 sounds a fair number to me) will help reduce the CPU usage.

 

If %limit_fps is set to 0, then just let it continue as-is for benchmarking purposes.

 

Anyway, if you can make sense of that lot, feel free to comment.

[0ctane]

I am a bit surprised that your CPU load is near 100%. At limit_fps of zero I get a full GPU load, but not CPU. Regardless, when I actually use a fps limit, typically 50fps, the reported fps always seems to be quite lower, 35-40fps. I have noticed this effect for both by Athlon 2200 / Geforce 440MX and the more powerful Macs I use. I highly recommend a FPS limit, otherwise you are going to stressout that videocard. EL will try to use every ounce of GPU power possible at limit_fps=0.

 

I thought the NTSC standard was 30fps for broadcast TV. I am not sure about PAL. Do (non-HD) TVs go up to 50fps? I am not sure.

[LabRat]

Picture 1: Task Manager with %limit_fps 50

 

Picture 2: Task Manager with %limit_fps 0

 

The PAL refresh rate standard is 25Hz interlaced, but TV out graphics cards just do 50Hz and output half the picture each pass.

 

I normally run on %limit_fps 12 as a rule - I don't watch the game closely enough for the slight jerkiness to bother me, but I was just checking my benchmarks as I have just overclocked my CPU from 2GHz (3300+) to 2.2Ghz (3600+)

 

Edit: The bottom picture shows where I loaded MSPaint if you are wondering about the extra threads etc..

Edited November 15, 2006 by LabRat

[emajekral]

I've noticed both behaviours as well. The GPU one in Linux, the other in Windows. Dunno how to solve though, so unless I'm doing speed tests I set limit_fps to about 5 fps over my goal and enable vsync too.

 

That refresh rate looked right to me labrat

 

I hear that American movies are generally pitched too high in PAL format because they just speed up the film from 24fps to 25fps. Is that true? (Offtopic, I know)

Edited October 6, 2006 by emajekral

[LabRat]

Nah, the movie is still played in realtime, my TV and video just playback at NTSC speed so it makes no difference at all.</offtopic>

[majestyk]

I've noticed both behaviours as well. The GPU one in Linux, the other in Windows. Dunno how to solve though, so unless I'm doing speed tests I set limit_fps to about 5 fps over my goal and enable vsync too.

 

That refresh rate looked right to me labrat

 

I hear that American movies are generally pitched too high in PAL format because they just speed up the film from 24fps to 25fps. Is that true? (Offtopic, I know)

 

I had a similar effect while being at University. For our cinema project, we had very old russian 30mm projectors. Due to the different voltage applied they ran always a bit faster than in their original country; we calculated the speed to be around 26.2 fps, which averages to about 5-7 minutes shorter movie runtimes over the length of a complete hollywood movie. You didn't really notice the difference while watching, but you always got home early. Nice way to save time. :-)

[Entropy]

I had a similar effect while being at University. For our cinema project, we had very old russian 30mm projectors. Due to the different voltage applied they ran always a bit faster than in their original country; we calculated the speed to be around 26.2 fps, which averages to about 5-7 minutes shorter movie runtimes over the length of a complete hollywood movie. You didn't really notice the difference while watching, but you always got home early. Nice way to save time. :-)

 

The difference is more apparent in the audio.