Hello!,

If you want to disable vsync to prevent lag but don't want to force your computer to render a zillion frames per second just to avoid the inevitable screen tearing then you can use the binds below to synchronize the rendering with the monitor's refresh period.

bind MWHEELUP "cl_maxfps 90;wait;cl_maxfps 72.0015;echo resynched"
bind MWHEELDOWN "cl_maxfps 60;wait;cl_maxfps 72.0015;echo resynched"

So with the above binds added if you see tearing across your monitor then swipe the mouse wheel to move the tearing out of view. My monitor's refresh rate is 72hz and so for maximum smoothness my cl_maxfps is also set to 72 to match. You can adjust the initial maxfps in the above binds so that the tearing moves up or down the monitor in bigger steps.

Why goddammit?:

Because I use a fanless mini pc (intel HD5000) that overheats if I force it to render at 308 fps constantly. So if you're on a laptop then it'll probably appreciate the above binds too!.

Please try 308fps again but with sys_yieldcpu 1, hope that helps solve the 100% cpu load.

Thanks for the suggestion, Dimman!

I do use sys_yieldcpu 1 but I find it never seems to make much difference to the cpu usage, even when running at just 72 fps. In the end I found the best solution was to run my machine in 'Power Saver' mode (Win 7) which limits my cpu to 798 mhz. Now I can barely feel any warmth coming from my machine at all. The downside to running it at such a low clock speed is that sometimes QW will sometimes need a little extra cpu which can cause QW to overrun a frame and knock the monitor refresh out of sync, and so then I have to readjust the tearing position again.

Ideally we need QW's renderer to use VBOs and then the cpu usage would probably only be about 10% at the most! But I suppose then the GPU would end up having to do more work and so then maybe my machine would still end up getting sweaty.

Ok :/ Problem is that we lack developers, especially ones with good knowledge of OpenGL. If you know your way around it, we could need your help?

most players use as high as possible FPS to promote smoothness and less tearing. definitely don't use vsync or low FPS when you have the option for high fps. if you laptop is overheating then get a laptop cooler or clean, clean the fan or get some new thermal paste.

I find 1155fps is a sweet spot. some people say there is no advantage over 616fps but they are wrong and this can be easily tested in 2 ways. using the 'vidlag' hud element or alternatively just look at an item on the map and move your mouse left and right either side of it. you will see that 616fps has massive tearing and 1155fps has much lower.

Nah, I've only dabbled in OpenGL briefly. There's also the problem of having to figure out the bsp format so then you can convert the map into a single mesh/vbo on loading.

Hi DB!,,

The more fps you have the more tearing you'll get. So if your fps matches your monitor's refresh rate you will get a single tear line. With a frame rate twice your minitor's refresh rate you'll get two tears. Three times your monitor's refresh rate will result in three tears, and so on...

But in your case you're using a frame rate that's multiples of 77 and not multiples of your monitor's refresh rate, so the tearing is never at a fixed position on your monitor. And also with a 144hz monitor the tearing isn't quite as noticeable as it is on 70hz monitor because you get less time to notice the tear. And with a very high frame rate on a 144hz monitor the tearing starts to turn into a skewing effect and so the image on your monitor will lean to the left or right. So the tearing is still there but it's very uniform and evenly distributed down your monitor and therefore not as jarring.

you are wrong about more fps means more tearing. it is the opposite.

correct, i use a multiple of 77fps, as does most people because it is smoother with the way that independent fps works in QW.

client/server send a packet 77 times a second which used to correlate with the frames displayed on your screen. now we have separate cl_physfps (frames sent to server) and cl_maxfps (frames displayed on screen). if you use above 77 fps, it has to 'drop' the frames (not send them to the server) that don't match the 77fps. so if you use 308fps, you send 1 frame and then drop 3 frames. that is a whole number. If you use, lets say, 2x 144hz so you are using 288fps then it has send 1 frame and then drop 2.74025974025974 frames. it doesn't match, you cannot drop a partial frame. the framerate being out of sync with the physfps (clent/server) causes it to be less smooth.

there are many topics on the forum about this and most will agree with me about using a multiple of 77. I don't know anyone who uses a multiple of their refresh rate anymore, everyone uses a multiple of 77 for this reason and the higher the fps, the less tearing. test it for yourself with the method i listed above.

now combine your super high framerates with cl_earlypackets 1 and you get the best combination for smoothness. with this setting, high frame rates actually let you see the things happen on your screen earlier instead of waiting for 77 or whatever lower fps you are running.

it isn't a competition to see who can get the most fps, but higher is better. it is smoother and u can see things on your screen earlier than you did previously with other settings. many people have challenged me on this claim and i have been able to convince them all otherwise by getting them to do a simple test or two.

In a modern engine, these high framerates would not be required as the frames being sent aren't tied to the frames being displayed. but in quakeworld it is.

Just to clarify, I would also use multiples of 77 if I were to use a high frame rate. For me 308 fps would be ideal. But multiples of 77 down at the lower range isn't so great. So 77 fps on a 72hz monitor would result in lots of judder.

The only way to get smooth motion with a low frame rate on a monitor with a low refresh rate is to either use vsync and have lag, or to disable vsync and have tearing. So this is why I came up with binds above to manually move the tearing out of view.

What config are you using exactly? The one that came with nQuake?

My config gets 3-4x the amount of fps than the nQuake config. Many people use mine. I would suggest giving it a try if you want it. I also have aliases to switch between eyecandy, noeyecandy and noeyecandy2 (fastest). It might solve your problems.

Intel HD 5000 should be easily capable of ~500fps (with a good config). At least you can run 308fps without it trying to melt the graphics card.

dirtbox config

Hey thanks! I'll give it a go!

I like your solution of disabling vsync but adjusting tear line position for minimal visual impact. I use this to demonstrate what tearing looks like and how it works to people. It's probably the best solution if you can't run high multiples of your refresh rate, where the tradeoff is more tear lines for less individual tear offset.

I wonder if you have played with vid_vsync_lag_fix, where vsync is enabled, but the engine attempts to render each frame JUST before it gets sent for output. The downside is that you miss a frame, the result is lag instead of tearing. You can adjust vid_vsync_lag_tweak if you are missing frames too often. It will not be more responsive than vid_vsync 0, but better than vid_vsync 1 with the benefit of no tearing.

nice cfg dirtbox :-)

Did you try it out raket?

I did, and i really like it! Shafting feels so much better on mouse2 than mouse3 :-) new record, lg 42% on pov :-)

...

nice... i often forget it is an unusual configuration... I've been playing like that since probably 1998 or 1999. I always would have preferred using mouse3 but it is hard to find a mouse with a good wheel for shafting. g400(s) is the only option in recent times. newer mice have such stiff wheels... I cant even use the mouse3 on a g403 for grenade because i use my middle finger for the wheel, not my index finger

Yeah I was using vid_sync_lag_fix for a few weeks before giving up on it because like you say it's not just as responsive as vid_vsync 0. But with a vid_vsync_lag_tweak 0.2 it was definitely better than just vid_vsync 1 alone.

I also tried your config. As expected I didn't really get a performance improvement but what I did find interesting was the fact that you use rollangle. Is that just because you like the effect or do you feel it gives you some other advantage?.

You got no fps increase at all using my config over the nQuake config? Send me your config.

Rollangle allows you too see what direction your opponent is strafing. Gives you a small advantage.
I increase the rollspeed also so that the rollangle on my own screen isn't as obvious (as that is the downside of using a rollangle). With the speed I use, you can get used to it very quickly and not even notice it on your own screen after a small time. I used rollangle in SW mode until 2004... was experimenting with it again before QHLAN. I can take it or leave it really.

Damn, I knew there had to be something sneaky about it!. Top players tend not to care much for aesthetics and usually anything in their config is there for a good reason. I never realized rollangle also appears on your opponents.

Here's my config:

Are you sure you don't get more fps with my config compared to yours? I just did a benchmark (using 4on4_fs_vs_la_e1m2.mvd ) with both your config and mine and here is the results:

Your config: 461 fps
My config (default): 1327 fps
My config (noeyecandy2): 1642 fps

My config gets 866 fps and 1181 fps more than yours respectively.

No wonder you are getting such low FPS and overheating that computer. You have full particle effects enabled and this will kill your fps big time. I can only get 461 fps with your config and I am running a GeForce GTX 980ti and an Intel i7-6700k (both very high-end last year). With my noeyecandy2 setting on an Intel HD 5000 you can expect easily 616 fps solid. Do you know how to do a timedemo?

Do yourself a favour, put your controls in to my config. Your config is garbage. Get rid of it.

EDIT:

I have an easier option for you.
Just by setting these 2 commands, I was able to increase fps with your config from 461 fps to 1239 fps:

This will give you a big improvement, but you could still get 25% more fps with my config and the /noeyecandy2 setting.

Oops!! Yeah I tested my FPS by simply standing in the corner of DM4 at the green armor with my crosshair on the bottom step in the ammo room. But yeah, I'm sure if I did a proper benchmark with timedemo then your config would be waaay faster.

And you're right about my particle count, I didn't realize how high it was. I'll lower it to see if that helps to cool things down a little. Surprisingly setting it to 0 doesn't remove the particles entirely and there's still a reasonable amount to make my trails look okay. So I may even leave it at zero!

Thanks for the tip!

Don't bother lowering it, use ZERO. This then uses the oldschool particles (that were in the original engine) which give you good fps. If you have it at anything above 0, you will still get shit fps.

i get 833 fps (standard)
i get 917 fps (noeyecandy2)

This is with 2560x1440... GTX 970, I5 2500k@3.3ghz. 8gb ram.

i get 1083 (standard)
i get 1163 (noeyencandy2)

... i5 2500k @ 4.6ghz.. (2560x1440)

....
i get 1143 (standard)
i get 1236 (no eyecandy)

... i5 2500k @ 4.6ghz (1920x1080)

1598 (320x199) fodquake-sw :]

http://demos.quakeworld.nu/download.php?id=2

Celeron 466 MHz, GF4 MX440, SDRAM 256 MB 100 MHz, Windows 2000
100 FPS

Sempron (Socket A) 1.9 GHz, Radeon 9100, DDR 1.2 GB 400 MHz, Windows XP
400 FPS

G4 1.67 GHz, Radeon 9700M, DDR 512MB 333 MHz, MorphOS 3.10
265 FPS
193 FPS (SW 320x214)

G5 2 GHz, Radeon 9600, DDR 2 GB 400 MHz, MorphOS 3.10
300 FPS
274 FPS (SW 320x200)

Athlon X4 631 2.6 GHz , Radeon HD 5670, DDR3 4 GB 1600 MHz, Debian 8
674 FPS

i3-4160 3.6 GHz, GF GTX 950, DDR3 16 GB 1600 MHz, Windows 10
2085 FPS

Happy bunnies!

(Edited 2017-04-18, 22:54)

Korni - use the 4on4 demo that comes with nQuake so everyone gets the same/comparative results.

It is called: 4on4_fs_vs_la_e1m2.mvd

dirtbox - Can you upload 4on4_fs_vs_la_e1m2.mvd somewhere? (downloading a ~50MB archive is a meh).

http://www.sensible.place/4on4_fs_vs_la_e1m2.7z

Celeron 466 MHz, GF4 MX440, SDRAM 256 MB 100 MHz, Windows 2000
110 FPS

Sempron 2300+ (Socket A) OC 1.9 GHz, Radeon 9100, DDR 1.2 GB 400 MHz, Windows XP
415 FPS

G4 1.6 GHz, Radeon 9700M, DDR 512 MB 333 MHz, MorphOS 3.10
302 FPS
197 FPS (SW 320x200)
192 FPS (SW 320x214)

G5 2 GHz, Radeon 9600, DDR 2 GB 400 MHz, MorphOS 3.10
315 FPS
276 FPS (SW 320x200)
256 FPS (SW 320x240)

Athlon X4 631 2.6 GHz, Radeon HD 5670, DDR3 4 GB 1600 MHz, Debian 8
699 FPS
808 FPS (3.1 GHz)

i3-4160 3.6 GHz, GF GTX 950, DDR3 16 GB 1600 MHz, Windows 10
2070 FPS (Win 10 RS1)
2313 FPS (Win 10 RS2)

i5-4690K 3.5 GHz (Stock)
2775 FPS (Win 10 RS2)

Configs:
http://korni.ppa.pl/temp/default.cfg
http://korni.ppa.pl/temp/default_sw.cfg

All fodquake
Celeron and Sempron run 0.3
PPCs run the old 0.4 gl and current sw 0.4
The rest run current 0.4
With s_nosound 1

(Edited 2017-06-20, 18:04)

I tried dirtbox cgf and now my QW is alot smoother and the fps is higher and stable
Perhaps I should have tried making my QW smooth like 10-15 years ago

indeed... almost everyone uses a shit fps config...if only ppl tried mine earlier

What would you suggest as shaft-button placement?
I don't want to switch weapons, I want to have it under one button only. Ive been using mouse 3 for shafting but that requires me to slightly change the grip which affects the aim.

And I have jump on Mouse 2, it would be hard to unlearn it now.

I've seen that Logitech G402 and g502 have buttons very close to Mouse 1 - has anyone tried those in QW? they are a bit heavier than most of competitive mice these days (103 for g402 and 113 for g502)