You are not logged in.
Pages: 1
Any tips on improving FPS? It's rather sad that it's 2013 and I'm still getting fps down to 50 in HL1 mods (with my laptop[1]). I've been playing Day of Defeat for a few days and the low fps is pretty irritating. I don't know how many decades it takes to run a mod from the year 2000 perfectly. I'd like the fps to be constantly 100. The fps is slow for all the games I've tested. The issue exists for native and wine "emulated" Steam. I'm currently sticking with native Steam.
I'm using 1920x1080 resolution, but other than that all the couple of graphics options are set to low quality. I tried to use a lower resolution but the impact on fps was pretty miniscule. I tried to use the Nvidia card according to the Wiki's instructions with primusrun, but the fps didn't improve noticeably. VSync is off. I'm using the libraries bundled with Steam. My desktop environment is Lxde (which is apparently one of the lightest desktop environment's out there).
1. My laptop was bought in early 2012:
Intel® Mobile Core i5 2430M / 3MB L3 Cache, 2.40GHz - 3.00GHz
Memory: 8 GB - 2 x 4096 MB - DDR3 SDRAM, 1600 MHz
Graphics: nVIDIA® GeForce® GT555M 2GB
Hard drive: 120 GB SATA Solid State Drive
Last edited by nawitus (2013-08-15 18:49:45)
Offline
do you use your external gpu when plaing? BTW: 30 fps is enough, the rest is waste of power.
Linux odin 3.13.1-pf #1 SMP PREEMPT Wed Mar 5 21:47:28 CET 2014 x86_64 GNU/Linux
Offline
Like I stated, I'm currently using the Nvidia card with primusrun. For me, 100 fps is the target since it looks nicer, and I believe HL1 game logic requires 100 fps to control the recoil better.
Offline
See hints in thread.
Offline
Arg forgot this see here:
https://github.com/amonakov/primus#faq
Linux odin 3.13.1-pf #1 SMP PREEMPT Wed Mar 5 21:47:28 CET 2014 x86_64 GNU/Linux
Offline
Tried 640x480 resolution windowed and with nice -20, and the fps improved somewhat. The fps still drops to 60fps all the time if I'm to believe the fps display in 'net_graph 3'.
Offline
60 FPS is of course the usual rate with vsync enabled.
Offline
I know, but vsync is not enabled in DoD options, and the fps is constantly between 60-100fps. With VSync on the FPS should always be below (or at) 60.
Offline
I know, but vsync is not enabled in DoD options, and the fps is constantly between 60-100fps. With VSync on the FPS should always be below (or at) 60.
this is controled by primus not by your game.
Linux odin 3.13.1-pf #1 SMP PREEMPT Wed Mar 5 21:47:28 CET 2014 x86_64 GNU/Linux
Offline
Tried 640x480 resolution windowed and with nice -20, and the fps improved somewhat. The fps still drops to 60fps all the time if I'm to believe the fps display in 'net_graph 3'.
By the way... nice for games is useless. Why do you want the games to have a higher priority than your kernel or X? Doesn't make sense.
Offline
nawitus wrote:I know, but vsync is not enabled in DoD options, and the fps is constantly between 60-100fps. With VSync on the FPS should always be below (or at) 60.
this is controled by primus not by your game.
I already tried vblank_mode=0 before issuing primusrun, but the fps didn't improve.
nawitus wrote:Tried 640x480 resolution windowed and with nice -20, and the fps improved somewhat. The fps still drops to 60fps all the time if I'm to believe the fps display in 'net_graph 3'.
By the way... nice for games is useless. Why do you want the games to have a higher priority than your kernel or X? Doesn't make sense.
Nice was a tweak recommended on the internet so I tried it. It helped someone.
---
I tried fps_max 200 and turned off vsync and set various settings to performance in nvidia-settings. No effect on FPS.
Offline
1. My laptop was bought in early 2012:
Intel® Mobile Core i5 2430M / 3MB L3 Cache, 2.40GHz - 3.00GHz
Memory: 8 GB - 2 x 4096 MB - DDR3 SDRAM, 1600 MHz
Graphics: nVIDIA® GeForce® GT555M 2GB
Hard drive: 120 GB SATA Solid State Drive
this should rock ... i have a core2 with a GF9800GT, 4G ram and have stable 60fps at 1280x1024 (yes, net_graph shows that), everything full (with vsync of course, above 60fps, you just waste the power) in games like portal, counter-strike: source, etc., so both valve and source engines.
You can narrow down the issue with:
1. try without any sound
2. try a different window manager too (if possible something lightweight), to rule out a faulty setting.
3. i'm out of date with the new i5 CPUs, but acording to ark.intel.com the CPU has an integrated Intel® HD Graphics 3000, if so, are you sure you are running the games on the nvidia card and not on the integrated intel?) - you can make a simple test, use a software render setting in any valve (half-life 1) engine at 640x480 or 800x600, the cpu should run that fluently, and compare that to OpenGL setting inside the game.
Offline
nawitus wrote:1. My laptop was bought in early 2012:
Intel® Mobile Core i5 2430M / 3MB L3 Cache, 2.40GHz - 3.00GHz
Memory: 8 GB - 2 x 4096 MB - DDR3 SDRAM, 1600 MHz
Graphics: nVIDIA® GeForce® GT555M 2GB
Hard drive: 120 GB SATA Solid State Drivethis should rock ... i have a core2 with a GF9800GT, 4G ram and have stable 60fps at 1280x1024 (yes, net_graph shows that), everything full (with vsync of course, above 60fps, you just waste the power) in games like portal, counter-strike: source, etc., so both valve and source engines.
You can narrow down the issue with:
1. try without any sound
2. try a different window manager too (if possible something lightweight), to rule out a faulty setting.
3. i'm out of date with the new i5 CPUs, but acording to ark.intel.com the CPU has an integrated Intel® HD Graphics 3000, if so, are you sure you are running the games on the nvidia card and not on the integrated intel?) - you can make a simple test, use a software render setting in any valve (half-life 1) engine at 640x480 or 800x600, the cpu should run that fluently, and compare that to OpenGL setting inside the game.
I'm using LXDE which uses Openbox, that should be light enough. I'm sure that I'm using the Nvidia card, because the game doesn't start if I don't manually start bumblebee service (which incidentally is another unsolved bug). I might try disabling sound if I have the time..
Offline
nice for games is useless
Wrong! Although, a niceness of -20 is extreme and bad, I agree with you there.
The point is that the game should take priority over e.g. polling for new email every x minutes, and Skype (which takes 2% CPU continuously, as a ridiculous amount of polling) for both CPU & disk contention (ionice).
What I do these days is have X (and a native game) run as SCHED_ISO, which requires the BFS scheduler. And for wine, use wine-rt, so that wine can set relative priorities of in-game threads.
A gamer should be running BFS anyway - CFS is noticeably more jerky, even after tweaking.
Last edited by brebs (2013-08-16 20:23:42)
Offline
ChemBro wrote:nice for games is useless
Wrong! Although, a niceness of -20 is extreme and bad, I agree with you there.
The point is that the game should take priority over e.g. polling for new email every x minutes, and Skype (which takes 2% CPU continuously, as a ridiculous amount of polling) for both CPU & disk contention (ionice).
What I do these days is have X (and a native game) run as SCHED_ISO, which requires the BFS scheduler. And for wine, use wine-rt, so that wine can set relative priorities of in-game threads.
A gamer should be running BFS anyway - CFS is noticeably more jerky, even after tweaking.
I provide packages with BFS and Tuxonice (Linux-pf) for Sandy and Ivybridge CPUs see: https://dl.dropboxusercontent.com/u/172 … index.html
Linux odin 3.13.1-pf #1 SMP PREEMPT Wed Mar 5 21:47:28 CET 2014 x86_64 GNU/Linux
Offline
I'm using LXDE which uses Openbox, that should be light enough. I'm sure that I'm using the Nvidia card, because the game doesn't start if I don't manually start bumblebee service (which incidentally is another unsolved bug). I might try disabling sound if I have the time..
yes, that's light enough. Maybe try starting steam directly from console with optirun, like "optirun bash", then type steam and see the output of console, just to be on the safe side.
Also you could try to disable the Multicore rendering in source engines (settings->video->advance), back in the old days it caused similar frame issues and who knows ... or even after starting a source engine based game, open the console and type cl_forcepreload 1 into it, this will force the preload of all resources like textures into the video ram when loading a level.
Last edited by LyCC (2013-08-16 22:16:30)
Offline
ChemBro wrote:nice for games is useless
Wrong! Although, a niceness of -20 is extreme and bad, I agree with you there.
The point is that the game should take priority over e.g. polling for new email every x minutes, and Skype (which takes 2% CPU continuously, as a ridiculous amount of polling) for both CPU & disk contention (ionice).
You say "Wrong!". I say: I want to see a benchmark test, where you can see a difference outside the error margin. If this is not possible, I say "nice for games is useless".
Offline
benchmark test
It's about smoothness, not throughput. Ya just don't get it, like Phoronix.
Offline
ChemBro wrote:benchmark test
It's about smoothness, not throughput. Ya just don't get it, like Phoronix.
I think we must differ between benchmark and benchmark. To test something which hardware is better or to find some bad side effects is ok.
But this is a bit weird cause you gain no benefit from doing it.
Linux odin 3.13.1-pf #1 SMP PREEMPT Wed Mar 5 21:47:28 CET 2014 x86_64 GNU/Linux
Offline
ChemBro wrote:benchmark test
It's about smoothness, not throughput. Ya just don't get it, like Phoronix.
So you say, there is no method to actually prove your "smoothness"?
And please don't bring benchmark and Phoronix together, they just don't fit.
Offline
LOL, the benchmark I use is my own eyes. Personally, I'm ultra-sensitive to frame-dropping, stutter, input delay.
Some intermittent frame-dropping has to be accepted, but if it happened like clockwork e.g. every 5 seconds, that would be unacceptable. Have you got a latency benchmarking tool that can measure my potential annoyance? ![]()
Offline
I don't have, I asked you. But there aren't any benchmarks that can measure "potential annoyance", so no. There are other benchmarks.
Offline
Tried with vblank_mode=0, with software rendering (with nvidia and intel), and without sound. No help still. Software rendering has somewhat slower FPS.
Offline
Software rendering has somewhat slower FPS.
As it should, because it is software rendering.
Offline
nawitus wrote:Software rendering has somewhat slower FPS.
As it should, because it is software rendering.
Yes, that was a pretty expected result. I tried it because someone recommended trying it out to troubleshoot the issue.
Offline
Pages: 1