ET is still power hungry

[Death_Reincarnated]

Nice read/explanation on the ping. The one thing I used to observe regarding lag issues was when being in a game poistion where there is no action to heaps of it - like facing a corner and then facing an open area with lots of action. This would make me lag for a while. Not sure if this was related to the increase in data receiving/sending or due to hardware/fps lag. This used to happen on an older machine.

 

Anyways, you believe you have lag due to hardware here are some cvars to help you out. It will not make the game look as pretty but it should improve your game performance in higher numbers.

 

- cg_showblood 0 - turn of blood splatter

- cg_bloodflash 0 - similar to above

- com_maxfps 76 - instead of the usual 125. I honestly don't see much difference between 76 and 125 but at least the game should be smoother in more situations.

- r_picmip 2 - the higher this number the crappier the textures will look, but it will improve performance. I think 1 is default.

- cg_wolfparticles 0 - turn of explosions and such. This chews up lots of hardware as it shows airstrikes and things blowing up.

- cg_atmosphericeffects 0 - turn of rain/snow.

- r_dynamiclights 0 - turn of the lighting around dynos as such.

- r_shadows 0 - turns of shadows under players.

- r_fastsky 1 - how often do you look up at the sky and notice that one of the clouds reminds you of something? Prolly never lol.

- r_drawsun 0 - like the sky, you dont look at the sun.

- cg_marktime 0 - turn off bullet marks.

- cg_brasstime 0 - turn of empty shells coming of the gun during shooting.

- cg_coronas 0 - turn of glow around lights.

- cg_coronafardist 0 - distance you see the glow.

- cg_muzzleflash 0 - removes muzzle flash during firing.

- r_lodbias 2 - higher number means better fps but lower geometric detail.

 

I think that is all lol. Play around with these and see. I would suggest backing up your config first.

 

GL

 

D_R

[SunLight]

Nice read/explanation on the ping. The one thing I used to observe regarding lag issues was when being in a game poistion where there is no action to heaps of it - like facing a corner and then facing an open area with lots of action. This would make me lag for a while. Not sure if this was related to the increase in data receiving/sending or due to hardware/fps lag. This used to happen on an older machine.

I think it's because when players and other entities (to check what's an 'entity', play a demo and type: /r_drawworld 0 to see only entities and not the map) are in your area (the so called PVS or potentially visible set) the game has to deal with them, even if they are behind a wall.

 

Then your graphics card draws the walls in front of them, and you don't see those entities, but your cpu has still to do the same work as if they were visible.

If you are in a zone where no other entitiy is in your PVS, then your cpu has to work less.

 

In open maps (or poorly designed ones) quite a lot of stuff is constantly in your PVS, in maps with closed areas and corridors it lags less even in servers full of players.

 

That's how I would explain it, maybe it's not a perfect explanation, but I don't think it's network related, unless you have a really slow internet.

 

- The data you send is higher if you move your mouse a lot. cl_packetdup obviously increases your upload, and also playing with high fps (there is a command every frame).

- The data you receive contains your own 'old' position in the server, and other stuff (for the client prediction) and the data about all the entities in your zone (PVS) and events (for example bullets, footsteps, and many others) plus other commands and things like chat, etc. (and something else I forgot )

 

So it also depends on how many players are in your PVS, how many bullets they shoot, how many medipacks they put on the ground (that's why etpro lags a bit less, when you kill medics, you don't have suddenly a lot of entitites on the ground, just their weapon and the body)

 

---

 

As regards the config values, they are the same as in my config, with a couple of exceptions (picmip for example). But with my graphics card I get the same fps regardless of that setting, so I keep it at 0.

A way to test this, is to record a (possibly short) demo when you have low fps, then replay it with /timedemo 1

It will play as fast as possible, and you can open the console after it's over and see the average fps. So you can use it as a benchmark. I get the same fps with picmip 0 and picmpip 3. So I prefer to have more detail.

 

As regards com_maxfps, well, if the problem is that I can't reach a certain number of fps, the only difference would be to have the same 'feeling' all the time instead of having it fluctuating in different areas of the map. Maybe that's what you meant.

[Vanaraud]

I think it's because when players and other entities (to check what's an 'entity', play a demo and type: /r_drawworld 0 to see only entities and not the map) are in your area (the so called PVS or potentially visible set) the game has to deal with them, even if they are behind a wall.

 

Then your graphics card draws the walls in front of them, and you don't see those entities, but your cpu has still to do the same work as if they were visible.

If you are in a zone where no other entitiy is in your PVS, then your cpu has to work less.

 

 

Thats what I´ve thought all the time, when Im in spec in at 1st spawn of attacking team and the game is played lets say in 3rd spawn then I get less FPS drops.

Also when I played in jay3 the other day I limited my FPS to 333 but it went haywire jumping around 250 and 285 so capped it to 250 with 77 ping and it was very smooth and no lag with 9 vs 9 players. Also I could do the same maxFPS 250 in jay2 with 140 ping(not that the ping value matters anyways as the calculations must be done every time, only the numbers differ. In my opinion its no different if you put 70 or 140 in interpolation formulas;)) and 10 vs 10 players but when more players join and server reaches 17vs17 or even more I get lagging, part of it could be ISP also but my experiment showed that with loading higher clock CPU profile in 2min  and playing 20 vs 20 without no FPS lag drops, it shows that server isn´t to be blamed of the lag. More the CPU or maybe ISP(though I have same ISP and tried 100KB\s, 2mbit\s and 5mbit\s speeds and as ET uses around 100kbit\s at max theres no difference imho. maybe 64kb dial up modem would be too less for ET:P) I know in gaming it more improtant the quality of connection than speed, if you drop a packet but have 150mbit\s connection then you loose one shot;) But I wonder if with todays technology the amount of information that server sends gets so big that my ISP couldn´t handle it going from 10vs10 to 20vs20. If it were solely ISP problem I should have lag all the time?

Actually could check it all up experimenting with CPU speed and also is there cmds to see how much traffic I receive on ET?

[SunLight]

Thats what I´ve thought all the time, when Im in spec in at 1st spawn of attacking team and the game is played lets say in 3rd spawn then I get less FPS drops.

Also when I played in jay3 the other day I limited my FPS to 333 but it went haywire jumping around 250 and 285 so capped it to 250 with 77 ping and it was very smooth and no lag

[cut]

Actually could check it all up experimenting with CPU speed and also is there cmds to see how much traffic I receive on ET?

About fps, why do you cap so high? imho you shouldn't go higher than the value when you don't feel it goes any smoother.

I don't know what's your monitor refresh rate, but I doubt that anything over 166fps would feel smoother.

 

The more fps you have, the more commands you have to send to the server. If it's too high you could even get a connection interrupted.

 

---

 

As regards the traffic, well I guess there are several utilities that show your DL / UL speed on your desktop, I don't know which one I could suggest.

If you aren't using other stuff while playing, that would tell you the bandwidth used by ET. Maybe your firewall could tell you that info as well.

 

On ET itself, there are some commands (they spam your console with lots of text so you have to set them back to 0, sometimes blindly, after you read)

there is cl_showsend for upload (it could be 0 or 1) and cl_shownet (it accepts values from 0 to 4)

 

cl_showsend: The number in parenthesis is the number of commands you send in that packet. It is obviously higher if you use cl_packetdup, because you send duplicates. Ideally it should be 1 (without cl_packetdup, 2 with cl_packetdup 1 etc.) imho, because if it's too high, it means you don't send commands at once, but you group them together and you send them in groups.

 

cl_shownet 1: should show you the size of the server message you received.

cl_shownet 2,3,4 is more detailed with the info in that server message, it tells what's in it (messages, snapshots, and some data on the entities in the snapshot)

cl_shownet 3 shows your ping as well.

 

edit:

I tried it now and cl_shownet obviously shows less data if I free-spec an area where there is nobody. Because the server sends you only the playerstate, commands, and maybe a couple of entities, while instead in areas with lots of action there's a lot of entities.

Edited May 20, 2013 by SunLight

[Death_Reincarnated]

@SunLight Smooth fps is better than having it set high and fluctuating. How low does your fps go when player numbers increase high enough? Any fps value you set that is above the Hz value of the monitor is pointless. There are monitors that go 100+ Hz but they cost fair bit - 50/60Hz is he general.

 

Out of curiosity:

- Have you tried rolling back to older GPU driver?

- What about your com_hunkmegs? It should be at least 256MB. Not sure how jay/nq mods work with this, but the pub mod would reset this value to default so you would have to put an extension in the shortcut to the game.

- Tried lowering your resolution?

 

P.S. Just found this on the net - http://www.defused.nu/index.php?/topic/408-low-fps-fix-nvidia/ I think you said you have amd-based? But, there might be a setting similar to that found in the link. Might help as my friend had this problem and this fixed it for him.

Edited May 20, 2013 by Death_Reincarnated

[SunLight]

@SunLight Smooth fps is better than having it set high and fluctuating. How low does your fps go when player numbers increase high enough? Any fps value you set that is above the Hz value of the monitor is pointless. There are monitors that go 100+ Hz but they cost fair bit - 50/60Hz is he general.

 

Out of curiosity:

- Have you tried rolling back to older GPU driver?

- What about your com_hunkmegs? It should be at least 256MB. Not sure how jay/nq mods work with this, but the pub mod would reset this value to default so you would have to put an extension in the shortcut to the game.

- Tried lowering your resolution?

 

P.S. Just found this on the net - http://www.defused.nu/index.php?/topic/408-low-fps-fix-nvidia/ I think you said you have amd-based? But, there might be a setting similar to that found in the link. Might help as my friend had this problem and this fixed it for him.

 

I didn't try to go back to previous drivers, but since I had low fps from the start then I don't think it would help. On Windows I haven't updated my drivers for a long time, so it's much older than the version I have on Linux, same fps.

 

And on Linux drivers aren't such an easy business as on Windows. For example I can't upgrade xorg now because Amd doesn't support latest xorg version, and I have to use a repo with the old xorg. Then I don't know if older drivers work with kernel 3.9.3. On Arch Linux there is a new kernel every other day...

 

---

 

The resolution doesn't affect fps, like picmip. I tried benchmarking with /timedemo 1. After all the bottleneck is the cpu, not the graphics card (which says 20% usage even while benchmarking an ET demo at full speed) I could even enable antialiasing and stuff if I wanted.

 

There isn't a setting like that one you showed me in my control panel. There is a thing called 'catalyst a.i.'

On windows it doens't affect performance at all. On Linux, my fps improve noticeably (i.e. I get more fps than on Win) but it uses 2 threads with 100% cpu, and I get mouse lag, because the cpu is too busy to read my mouse movements properly. Let alone the rise in cpu temperature. Maybe there is a way to give lower priority and cpu% to that thread only but I didn't investigate much.

 

---

 

As com_hunkmegs I use 196. I could rise it to 256, but in game I can check how much memory is actually allocated (via /meminfo) and it doesn't make sense to set it too high if it's not being used. I will check how much memory the game allocates when I play. Imho it's lower than 196, but you never know... I set it to 256 now.

 

---

 

For the monitor thing, you're right, it doesn't make sense to have anything above your monitor Hz, but there is a possible reason to use fps above your monitor's Hz:

since nobody plays with vsync on (vsync lags all mouse movements, for some reason) there is the problem of the combined delays from mouse poll, game fps, and monitor fps.

 

- I move my mouse, then I have a little delay according to my mouse polling rate (1ms at 1000Hz)

- The game reads the mouse move every frame, so if I move my mouse when a frame has just 'started' I have to wait until the next frame for my movement to be read.

- when the game displays it, then you have to wait for the monitor to display it, according to your monitor Hz.

 

That's why I notice an improvement over 60fps even if my monitor is 60Hz.

 

 

with high fps the delay can be very low sometimes
 
mouse (500Hz): * * * * * * * * * * * * * * * * * * * * *
game (125fps):   +       +       +       +       +       +   
monitor(60Hz):        |               |               | 
 
with 60 fps the delay is fixed (well, assuming that the game goes as steady as the monitor, which can't be)
and in the worst case it can go really bad, here is an 'average' case
mouse (500Hz): * * * * * * * * * * * * * * * * * * * * *
game (60fps):   +               +               +          
monitor(60Hz):          |               |               |

[Antichrist]

I've had issues running ET smoothly for as long as I can remember. I could NEVER maintain steady FPS. With 2 monitors, I couldn't even maintain the server minimum. I would constantly change between 15-30FPS.

 

The only way I was able to change my problems, was a massive upgrade (Used a 512mb card before). Installing a 1G video card seemed to help more than minimizing graphic settings. I can now run 2 monitors, with almost full graphics, and maintain 76FPS steady.

[SunLight]

- Have you tried rolling back to older GPU driver?

Lol it was a while since I last benchmarked on Windows vs Linux, you're right! When I installed Linux I got the same fps as Win, and that for a long time, now I tested again and lo and behold, on Win i get 20 fps more -.-

Reading your post gave me this idea of comparing the old version at windows, without timedemo benchmark I wouldn't have realized...

 

I quote an old post fom another thread, this radar demo (old etmain demo) is often used by ET players as benchmark:

 

radar.zip

• downloaded with extension dm_83 renamed to 84
• new profile with default setting
• used etmain folder

 

if someone with a fast pc (or slow, still interesting) could try this, so we can compare our fps:

 

- put the demo in etmain/demos

- load et main screen (etmain, not other mods!, just open et.exe without loading any mod)

open the console:

/timedemo 1

/r_drawfoliage 0

/demo radar

 

(don't touch anything while the demo runs)

 

after it's over, open the console again and read.

 

here are my values:

 

Linux (latest drivers)

2896 frames, 50.3 seconds: 57.6 fps

 

with catalyst a.i. on, (but the whole machine lags, so it's unplayable)

2896 frames, 34.3 seconds: 84.4 fps

 

Windows:

2896 frames, 35.3 seconds: 82.0 fps

 

 

darn! fu microsoft and amd!

[Vanaraud]

Heres my results on Win7, OC default with Power Boost to 3.8Ghz and memory @1600:

2896 frames, 16.9 seconds: 171.0 fps

and with CPU OC-d to 4.2Ghz and RAM@1600mhz:

2896 frames, 14.4 seconds: 201.1 fps

 Thats about the realtime advantage I gain in Oasis map also. I tried yesterday with players amount 13vs 13 near the allies CP door where you can jump over to other wall, where I got FPS drops.

With default settings(see above) I got FPS drops from 125 to 80, with OC-d from 125 to 100FPS.

So I really don´t get it how is it possible to get with any number of players and 3-4y old PC "stable" 125FPS.

Also thats why I´ve capped it to 76PFS, it combines fine with rest of settings(maxpackets), is one of the "magic FPS" and even though my monitor doesn´t go above 60Hz higher FPS is still better than lower;) .

[SunLight]

Heres my results on Win7, OC default with Power Boost to 3.8Ghz and memory @1600:

2896 frames, 16.9 seconds: 171.0 fps

and with CPU OC-d to 4.2Ghz and RAM@1600mhz:

2896 frames, 14.4 seconds: 201.1 fps

nice values, after all your Ghz is (more than) double than mine, as expected you go 2x as fast and more...

 

now I was trying some other things, I tried to run the same test with r_skipbackend 1 (careful because when you set it to 1, it won't draw anything and all the other commands must be entered blindly, on Windows)

 

on Linux I get 314.9 fps on Windows 282.1 fps, which means that the 'cpu-only' part goes better on Linux, only the current catalyst drivers are not as optimimzed as Windows ones it seems... -.-

 

I tried to compile ET on Linux with SMP 1 (a feature they didn't put in the game for Windows at all, and even on Linux it doesn't work)

I get this message: 'Using dual processor acceleration' but then after '----- finished R_Init -----' the game hangs.

 

I was curious to check how many frames I could get with multiprocessor, but apparently they didn't finish that feature, so it's in the source code, but it doesn't work. Maybe they fixed it in ET XReal, I should check it, but for now I don't have the time... anyway that radar map is an extreme case... sp_delivery and few players I get 125 fps

[Death_Reincarnated]

Glad I was able to help.

 

So your FPS is ok now?

 

I've read/heard several problems with recent updates to catalyst/nvidia, specifically when 3D/threading was first implemented. There was also one nvidia update that screwed GRID game over because of the sound driver update - it assumed surround sound by default even though you had stereo speakers and the game would crash on screen.

 

I don't know the options in catalyst a.i. but it would be good for you to go through them and turn of any that you don't think are necessary.

 

D_R

[SunLight]

Glad I was able to help.

 

So your FPS is ok now?

Still no, but you put me on the right path

 

Now I'm trying to install some old Catalyst versions on a Fedora 32bit installation I don't use. Because I swear that when I switched to Linux I got the very same fps as Win for some time. I benchmarked with /timedemo 1 so it's no placebo.

 

Then I didn't benchmark anymore, and since on Windows I only play on etpro servers, in those maps with few players it's normal to get more fps, I didn't realize Linux went behind...

Now I wanna see which version of the driver started to get slower, because sometimes I have the impression that the more I update, the slower it becomes ...

 

Then if I find a faster one the problem would be to install it on Arch, because that distro generally is more up to date than, say, Ubuntu, they always put everything last-version and using old stuff is hard. We'll see...

 

Thanks for your help up to now

[Antichrist]

I ran it as well.

 

Windows 7

AMD FX 6100 6-core @ 3.3GHz

16 GB RAM

 

1GB EVGA Video

 

2896 frames, 17.6 seconds, 164.5 FPS

 

Hope it helps...

[SunLight]

2896 frames, 17.6 seconds, 164.5 FPS

 

Hope it helps...

thanks, not bad if compared to my 57 but now I think I fixed it!

I finally found out something, maybe it could help some Linux user with ati who happens to read this, and has the same problem.

 

 

I was sure that on Linux I went faster than on Windows, I didn't dream about it... now I found that messing with xorg.conf and aticonfig I got it back to how it was before (i.e. much faster than Win) and that with the latest Catalyst and Arch Linux 64 bit.

 

In recent versions of Linux there is no xorg.conf, so apparently some values were deafult, and I also found an old ati file I had put (and forgotten) in /etc/X11/xorg.conf.d

 

result:

Intel® Core™2 CPU          6300  @ 1.86GHz (no overclock)

Ati Radeon HD5670

 

Windows Xp:

2896 frames, 35.3 seconds: 82.0 fps

 

Arch Linux before the fix:

2896 frames, 50.3 seconds: 57.6 fps (no catalyst a.i.)

2896 frames, 34.3 seconds: 84.4 fps (catalyst a.i.)

 

Linux Fedora 32bit, uberold catalyst: 64 fps (better than the new driver, but still sucked)

 

After the xorg.conf fix:

Arch Linux 64 bit, kernel 3.9.3, Catalyst 13.4

2896 frames, 34.0 seconds: 85.3 fps (no catalyst a.i.)

2896 frames, 20.9 seconds: 138.6 fps (catalyst a.i. on)

 

138.6 is way faster than Windows, and in fact that's what I remembered. Not bad for 1.86ghz, I wonder how much I would get with a fast cpu, now I only have to solve the cpu usage problem with catalyst a.i. on, but at least I go faster than windoze even without a.i.

now I'm a happy man ^^

 

thanks 2 everyone who helped me and read my post up to now, btw

[Jobba]

rofl