1000 FPS + on Debian 5 please

[Monk]

*shakes head*

Changing distros isn't going to do anything.. Just because one person gets good results on one distro, doesn't mean another user will, because they use different hardware.

HIGH FPS DOES NOTHING BUT SUCK UP CPU AND CAUSE HEADACHES. THE FPS WARS ARE OVER. IF FPS IS IMPORTANT, HACK THE PRINTF AND MAKE IT RETURN 1000, I GUARANTEE YOU AND YOUR CUSTOMERS CANNOT TELL A DIFFERENCE

[lol554]

Have you ever heard of different libs or kernels? I had terrible results with Ubuntu 9.10 after the OB update, now when i use 10.10 i get excellent results. Still the same hardware.

I doubt it's Santa that are playing around, it's too early for that.

[Monk]

(11-15-2010, 05:05 AM)lol554 Wrote:  Have you ever heard of different libs or kernels? I had terrible results with Ubuntu 9.10 after the OB update, now when i use 10.10 i get excellent results. Still the same hardware.

I doubt it's Santa that are playing around, it's too early for that.

What results?

Seriously people, if -fps- matter that much, VALVe would've added it to their technical documentation, but there is only mention of tickrate, and even that is deprecated.

[lol554]

A lot more stable fps.

[Monk]

I don't think there's any technical documentation out there that shows that 'stable' FPS does anything useful. If you could provide me with non biased information on this so I can look over it, that would be great.

[lol554]

(11-15-2010, 05:32 PM)Monk Wrote:  I don't think there's any technical documentation out there that shows that 'stable' FPS does anything useful. If you could provide me with non biased information on this so I can look over it, that would be great.

Ok, so in fact, a spiky fps-server does run as good as a stable fps-server - in your world? I don't need to provide any information, i feel it in the gameplay if the server fps are spiky.

Why don't you prove the rest of the forum wrong?

[Monk]

Prove you wrong? I asked you for technical proof, and this is your technical answer:

'I don't need to provide any information, i feel it in the gameplay if the server fps are spiky.'

I could take that in a few ways. The first way is, placebo effect. You think it's better because you have the notion of a 'fixed' number in your head. You constantly do rcon stats and the number stays the same. The physical gameplay is different every instance you play (meaning different maps and players all the time), but yet you think it's better even though there is no scientific way to prove it.

The 2nd way is that you completely discount what I said because you know deep down there is a possibility I am right. You are horrified at the idea that FPS doesn't matter (at all)

The final way is that you just think it's that way because you have 50,000 other morons who say it on a forum / web site / server companies who have no clue at all and just repeat what others say.

I did post an email from Alfred @ valve a few months ago on here regarding server FPS. He states it's pretty much useless (over tickrate), so if it's useless over tickrate, then stable fps is useless as well, because there is no documentation on valves site or anywhere (non biased) that says it's significant.

You should try and have an open mind instead of discounting what people say when it doesn't agree with your point of view.

[lol554]

We have different points of views, but you really don't respect that. Im afraid that your one of those persons that only believes in yourself and doesn't listen to others. My servers are running at 67-70 fps, as it has been stated on the forum that higher fps is unnecessary, and just causes the cpu to be sucked up.

When i mean gameplay, i mean how the tickrate is, and how the gameplay is. People want a smooth gameplay and tickrate, and thats the reason of having a stable fps. If you have the fps bouncing from 100 to 30, the server doesn't run smooth and the tickrate will drop several times.

Just call everybody but yourself a moron, you won't get any popularity that way..

[Monk]

As long as the FPS is 1:1 with the tickrate, then nothing else should be needed. If stable FPS / anything really mattered, then VALVe would list it on their site under their technical documentations.

[lol554]

(11-15-2010, 06:25 PM)Monk Wrote:  As long as the FPS is 1:1 with the tickrate, then nothing else should be needed. If stable FPS / anything really mattered, then VALVe would list it on their site under their technical documentations.

Maybe Valve haven't mentioned that in their technical documentation, but most people can figure out the puzzle themselves. For example, of you pay for a 10 Mbps connection, you expect that connection to be stable as possible, and not bounce from 1 to 8 Mbps.

[Monk]

That's an amusing analogy. Thank you for the laugh, it's nice to see someone attempt to justify their reasoning by using the a network speed as a point of reference when they have nothing to do with one another..

[lol554]

(11-15-2010, 06:34 PM)Monk Wrote:  That's an amusing analogy. Thank you for the laugh, it's nice to see someone attempt to justify their reasoning by using the a network speed as a point of reference when they have nothing to do with one another..

What are you laughing about? It's exactly the same situation, but of course, you can't realize that.

[Monk]

So..

If your server has an FPS of 66 <-> 99 and your tickrate is @ 66hz, if the FPS bounces between 66 and 90, that is bad? or 66 to 68? How is that bad?

[BehaartesEtwas]

guys...


valves technical documentation isn't comprehensive, so we unfortunately cannot conclude anything from something not being mentioned there.

but: experience from various people show that since the orangebox update the fps are not so important anymore. my explanation for this is, that the orangebox engine has a dedicated thread for receiving network packets. as the only reason for high fps in the past was the uncertainty in the time when a packet arrived, that was caused by looking for new game packets only at each frame and which therefore could be reduced by rising the fps, this completely solves the need for high fps.

but: the ticks (i.e. the "snapshot" of the world) are calculated during the frame. so if you want a stable updaterate (=tickrate in this context), you need your fps to be stable. the important thing is in the fps-meter graph the "TIME RMS", which gives the variance in microseconds. large variances mean automatically variances in the updaterate. 1ms (i.e. a TIME RMS of 1000) is probably okay, as network ping variations are in the same order of magnitude. but if you want to be sure, you should stay below it, so the quality is only determined by the network connection (which is the optimal case).

so: regardless of the fps you reach, the quality of the server is given (on software side) by the fps variance ("TIME RMS") only, supposed the fps permanently stay above 66 (=the tickrate). lowering the fps to the tickrate will make things easier (as it lowers cpu usage), but I doubt it will solve all of your problems. it looks now that nobody can directly tell you what you are doing wrong.

I recommend you to start over and do it in small steps: the kernel isn't so important anymore (any recent kernel should do it). important is the server runs at realtime scheduling, and that nothing else disturbs it. don't use the idler, don't change nice levels, don't use ld_preload libs, and don't do any other "tricky" optimization - for the moment. start with a fresh installed srcds without any plugins or special settings. use fps_max 68 (or so) and run it with a standard map and with realtime priority under a non-root user. make sure it's the only game server running.

if that works you can add step by step everything else you need.

[Terrorkarotte]

Why use the meter if you can turn the net_graph 4 on and look at "var".
The smaller "var" is the better the server. 10 and above is really bad