yo yo...

after toying around a fair bit in pov duel and 4on4, I _think_ this is what is best for me..


cl_maxfps 308 (=1.6ms vidlag with show vidlag command, 500 hz mouse = 2ms so its ok? doesnt matter or?)

cl_earlypackets "1" (somehow with this on 1, it feels like i am more "connected" to the floor, and somehow gameplay feels just a tiny bit "slower" )


150 hz crt display in 640 x 480





can you please post your maxfps/earlypackets setting, and screen hz/resolution?

1. mouse 1000hz (but looks capped at around 700hz according to mouserate.exe)
2. cl_earlypackets 1
3. cl_physfps 77
4. cl_independentphysics 1
5. crt 154hz
6. cl_maxfps 616

I think anyone would benefit from 1-4.
Jjust make sure #5 x Integer = #6

Fail. (1) (2)

As for the kungpun, yes, it's (earlypackets 1) definetely better on any and all situations. As for the vidlag, I'm not so sure it works ok without vsync, or if there was such a limitation in the first place. However, the real frametime you have at 308fps is actually 3,247ms but since you are using 150Hz display, it becomes 6,67ms. (yeah, that's right, even the mighty CRT has "video lag" in one sense)

im not sure what your on about renzo?

clearly you want cl_maxfps to be a multiple of screen refreshrate? Or maybe i should read your links...

I know there is no need to sync physfps and cl_maxfps any longer.


But cl_maxfps and refreshrate feels better when synced. Might be placebo...

There's absolutely no such "rule" of monitor Hz * integral = maxfps. The "rule" is for people who don't use cl_earlypackets, and it's between cl_physfps and cl_maxfps. And I think it's mentioned quite clearly in the theory of smooth quake. (right, ok, I ignored 60/75Hz TFTs which tend to have fps/Hz related smoothness problems)

ok, that rule might or might not be important. But those settings aint "fails". It's just settings.

We both pointed out that the dependence between physfps and cl_maxfps is gone with earlypackets. But i dont see why earlypackets should change anything between cl_maxfps and monitor refreshrate.

Don't get me wrong, I understand totally how difficult it is to get 60Hz/75Hz TFT to look smooth in any situation. But for CRT (something kungpun has) running at 150Hz 154fps is more than enough to make it better than any TFT there is. (kungpun: it's just an example, use your 308fps instead)

As for the failing part: You will get jitter without cl_earlypackets, if you use other than physfps * int = maxfps formula. So he has 150Hz monitor (not 154Hz) and he would set 300fps or 600fps according to your advice, it would be worse than running 154/231/308/385/etc fps.

There was a specific ezQuake build by Qqshka that had internal timings monitoring. At cl_physfps 77 and cl_maxfps 500 there was +-2ms jitter in the internal timings, while with 154fps, 231fps, etc fps there was 0,004ms jitter. Guess which one looks worse?


//EDIT: more accurate explanation

//EDIT2: also, moving to a bit different section

Not if he also set cl_earlypackets 1

if you got a monitor that supports max 150hz

i would try

cl_earlypackets 1
cl_physfps 77
cl_independentphys 1
cl_maxfps 600/300/150
crt 150hz

or

cl_earlypackets 0
cl_physfps 77
cl_independentphys 1
cl_maxfps 616/308/154
crt 150hz

I dont see how that goes against what you said here renzo.

if you try to sync fps with monitor refresh you without using vsync, won't you get a slow moving tear because they arent quite synced? I know ive seen it at lower refreshrates/fps... cl_earlypackets is not perfect yet, I get severe movement prediction errors when i traverse a ramp with cl_earlypackets 1, 308 fps, and a latency to the server that varies from approximately 8-13ms.

1. mouse 500 Hz (unlike old Logitech this A4 X7 is able to deliver stable 500 Hz data)
2. 150 Hz crt
3. cl_maxfps 150
4. vid_vsync 1
5. cl_vsync_lag_fix 1
6. cl_vsync_lag_tweak 0.5
7. earlypackets 1, indepphys 1, physfps 0

I clam this to be the smoothest quake (tm)

150hz,308fps,500hz mouse,cl_earlypackets 0.... FYI: cl_rollalpha 99999999999999 (disables smoothing of rolling), cl_rollangle 4, cl_rollspeed 200, v_viewheight 0

It has severe input lag, but it will look the smoothest indeed.


It can't get better than that. The problem with your case is that very high variation of pingtime. What you see will be delayed from 9,74ms to 16,23ms with your configured fps if the ping values you gave are really 8,0ms and 13,0ms (3,25ms/6,49ms/9,74ms/12,99ms/16,23ms @ 308fps).

For example, my ADSL2+ connection can do this: (ping variation less than 2ms, well between 9,74ms/12,99ms)

Renzo , always killer feedback :E


does this mean that at 616 fps /earlypackets 1 i have 6.67ms/2 real frametime?


whats the frametime for 1001 fps?


final question, anyone feel better aiming in 616 fps than 308 fps? I think milton uses 308 fps :E

all these tweaks can drive u nuts


keep posting your settings people !

1. mouse 500hz
2. cl_earlypackets 1
3. cl_physfps 77
4. cl_independentphysics 1
5. tft 75hz
6. cl_maxfps 924

ps. for some odd reason, vid_wideaspect 1 seems more smooth (maybe vid_mode -8?).

1. mouse 500hz
2. cl_earlypackets 1
3. cl_physfps 77

maxfps 462 and monitor on 154hz
Changed from 308 fps to 462 with earlypackets. Earlypacks and more fps still feels kinda weird, like i barly miss shots that before was a dh
Also been testing 480 or 500 fps and monitor on 160hz, but always go back to 462&154hz. Maybe since server updates 77fps it still a smal, slight advantage having everything x*77.

king, even if monitor have 6something ms betwhen frames at 150hz it will show the last mouseintput/frame read so inputdelay is lower
(half at 308, 1/3 of 6ms at 462fps etc)

A) No, frametime is 1,62ms
B) frametime is 0,999ms

cl_earlypackets does not affect frametime. Frametime a period of time it takes to finish rendering one frame at certain framerate. Frametime is not the same as screen refresh time, unless the fps = Hz. Therefore, running a CRT at 150Hz will have 6,67ms screen refreshtime regardless of the fps you are getting.

1. mouse 1000hz
2. cl_maxfps 1001
3. 800x600@154hz crt
4. cl_independentphysics 1
5. cl_physfps 77
6. cl_earlypackets 1
7. cl_newlerp 0.1

I have the same as en_karl, except that I think I let server use the max fps with cl_physfps 0, which ofc normally is 77 though.

may I ask why cl_newlerp 0.1? whats the difference in your experience vs newlerp 0 ?



Renzo maybe some feedback on newlerp?

Smoother rockets.

Newlerp is a prediction code for smoother projectiles like blAze said. However on newer servers you don't need this feature since the lerping works actually better after sv_mintic was finetuned for "77fps" limit in the last release of MVDSV+KTX server package (hopefully people got the change somehow, it does NOT default unless you get the newer mvdsv.cfg OR add the change yourself.). The servers in my signature, of course, have it changed like so.

Also, cl_newlerp doesn't work very well with pingtimes close to frametime (or their multiplications). Just try it yourself and you'll see what it does.

Oh I didn't know it's obselete. You learn something new every day.

That cl_earlypackets 1 seems to have worked wonders for my game...

1. mouse 500hz or 1000hz can't remember
2. cl_earlypackets 1
3. cl_physfps 77
4. cl_independentphysics 1
5. crt 150hz
6. cl_maxfps 600

Happy to hear why this might be sucky. There is slow tearing, but on cl_maxfps 308 or 601 or 599 or anything the tearing is worse. I can potentially set monitor to 154hz if it'll make a difference.

I quickly tested yesterday the new settings (not during the real game tho :/):
1) earlypackets 1 (had 0 before)
2) wideaspect 1 (had 0 before)
3) physfps 77 (kept the same)
4) maxfps 385 (has 0 before)
5) nolerp 0 (kept the same)
6) vidmode 7 @ 60Hz TFT

looks way better than before....I shall try this tonite during the games...

If its not smooth on your TFT, try physfps 75 and maxfps 300

it IS definitely smoother than my previous setup (wideaspect off and maxfps 0). I've also noticed that cracking down the maxfps to 385 removes ..... the squeeking of the capacitors on my GF9600GT

LOL. My GTX 285 squeals like crazy at 1000fps or more.

Heha my 8800gt does the same. Have to lower the maxfps when I'm specing with a lower volume .

hot gpu is the main reason i use 462 fps and not higher. If i set cl_maxfps 0 the temp goes up into 90's something rather quickly.

What about cl_physfps_spectator ? I currently have it on defult, but tested 38.5 yesterday and it seemed very smooth.

It can. It's better with cl_earlypackets 0, there are no movement prediction errors. It looks like prediction needs to be revised to account for inconstant 'network framerate'(?). My average ping is 10ms, the latency rarely exceeds the 1/77 seconds.