It is a few hidden ConVars that are being set for dxlevel 90 and 95, using the dxsupport_override.cfg file. These changes aren't relevant for dx8. What exactly do you prefer with dx8?

The max performance preset is good for toasters. You'll also probably want to use the iGPU preset.

I seem to experience strange hitches with "net_queued_packet_thread 0"--what are the supposed benefits?

net_queued_packet_thread 0 shouldn't be causing hitches unless your CPU or CPU scheduler is being dumb. Without the queued packet thread, split packets are sent without a delay.

what's the default value for net_queued_packet_thread? i'm experiencing a similar problem, dunno if it's that setting or another, but i've only begun having that problem since installing this config.

The default value is 1.

Values:

0 - Send packets and split packets immediately
1 - Send packets and the first split packet in a sequence immediately, but queue the rest of the split packets on a separate thread and send them after a delay in between each split packet
581304 - Queue all packets on a separate thread and send them after a delay in between each packet

I tested 0 on Windows and didn't encounter these issues, but if other people experience it, then I'll set it to the default of 1. By the way, what CPU are you using?

Also, stabby, are you using the -NoQueuedPacketThread launch option with net_queued_packet_thread 1?

EDIT: After looking at the documentation about how the Windows network system works, and then crying, I've decided that net_queued_packet_thread 1 is unfortunately necessary for Windows, except in some rare cases.

=

i didn't have -NoQueuedPacketThread enabled in my launch options and haven't noticed the hitching/stuttering since enabling it, although the problem occurred quite randomly to begin with. could ya link relevant documentation about the windows network stuff + net_queued_packet_thread for other tech savvy chaps to look over? thanks for putting so much work into this config, it's good to know someone is actively testing stuff out and explaining things for the rest of us plebs.

No iteration of network settings have had a negative effect for me thus far, however, I do disable smoothing as I only get prediction errors during specific collisions. It might be placebo but sometimes I feel like shots register sooner. I do notice warping but only from people who are blatantly playing on default settings (20 tick).

I would be interested to see odd interp settings scrutinised based on the premise of other people's update rates however.

I really like dx8 because of the shorter rocket trails. i cant see shit after firing a rocket in dx9. i wish so badly that dx8 gave better performance. also, i kind of feel like dxlevel 100 is giving me shitty performance compared to 91?

So do you mean you don't get stuttering and hitching with net_queued_packet_thread 0 and -NoQueuedPacketThread?

https://docs.microsoft.com/en-us/cpp/mfc/windows-sockets-in-mfc,
https://social.technet.microsoft.com/Forums/windows/en-US/ead342a2-3fbd-48c7-80f8-85b34054e64b/windows-7-suddenly-limits-outgoing-udp-packets?forum=w7itpronetworking, and comments in the Source Engine's net_ws.cpp.

thanks sooo much!!! the kind words really mean a lot! <3

Strange, do you have the -NoQueuedPacketThread launch option and net_queued_packet_thread 0 ConVar set?

cl_smooth does not affect other players, it only affects your own eye location. Interp and update rate won't affect other users, cmd rate will, but only slightly. Ping is the thing that matters the most.

Hm yeah, the particles are a lot simpler in dx8. Not sure if those particle files are sv_pured or not, so maybe you could move them over to dx9.

Well 91 doesn't exist, 91 will just go to 90. The available settings are 90, 95, 98 and 100. 98 is only allowed on Xbox 360 and I don't think 100 does anything but I haven't investigated it enough.

Hey, idk if this has been asked before but is there any version of this config that has gibs and ragdolls disabled? I was trying to edit it to have them disabled earlier and it wasn't working at all.

The default config (comfig.cfg) should disable ragdolls. You can disable gibs with the comp, compquality or maxperformance presets.

What version of the config are you using? 4.3.0 should fix this.

I'm still using comanglia's + your launch options + you fixes for comanglia and I get no issues with net_queued_packet_thread 1 and -NoQueuedPacketThread although I believe these nullify each other, no?

No, net_queued_packet_thread 1 will queue and delay packets regardless of -NoQueuedPacketThread, but will do this unthreaded if you have -NoQueuedPacketThread.

You should either use net_queued_packet_thread 0 + -NoQueuedPacketThread (which I recommend if you don't encounter problems with it), or net_queued_packet_thread 1 without the launch option.

Mainly the particle effects like the rocket trails, pyro flames, sniper scope and other annoyances.

Calling testers:

add "ConVar.cl_clean_textures_on_death" "1" to section 3 of dxsupport_override.cfg.

Added somewhere around Gun Mettle, I think. When you get a freezecam/die, it removes unused textures.

^ didnt notice any issues with that

however after 4.3.0/4.3.1 (i was on 4.3.0a until just now) there's a lot more visual clutter like blood bursts on death and grass on the maps. is that intended? im still on the comp.cfg preset

edit: adding -dxlevel 95 to my launch options then removing it again fixed it? fucking shit game

I'll try it out and get back to you.

On a side note, I take it that the game checks for mat_dxlevel and then loads appropiate dx overrides, right? Running -dxlevel all the time makes the game start considerably slower for me.

Yeah you only need mat_dxlevel 95, not the launch option. I already have it in the comfig.cfg. And thanks.

Nice timing :)
Anyway, I tried net_queued_packet_thread 0 with the -NoQueuedPacketThread switch and all is good. I'm on windows 10 btw.

Dunno if it's placebo but when I set net_queued_packet_thread to 1 I had some weird hit reg. Now that I have net_queued_packet_thread 0 it seems to be fine although I havent tryed experimenting with the said launch option.

Alright, sounds good. Seems like we will have to find out if stabby was using -NoQueuedPacketThread.

Well you should also have the launch option if it is at 0. It saves memory and CPU which is being idly wasted.

Also, no, it isn't placebo. When packets are bigger than net_maxroutable, which is maxed out at 1260, Source divides them up into smaller chunks called splitpackets. These splitpackets are delivered in a sequence and with net_queued_packet_thread 1, each splitpacket after the first splitpacket is sent on another thread out after a delay that depends on your rate setting and the splitpacket size. If you have it set to 0, then the splitpackets are sent at the end of the frame immediately.

So like. How exactly do i install this? i have only used comanglias config before this and all you needed to do was put autoexec in cfg folder but this has multiple cfgs and folders so dofak i m ment to do...

I think the problem is usually anti aliasing, try to mess around with that.
This is what it says in comfig.cfg:

//mat_antialias 0 // Do not do antialiasing (commented due to HUD and overlay glitches, set it once and then comment it out)

She says how to do it in the OP:

Please don't downfrag people who just want help!

(also sachiko, she*)

Sorry! I didn't know. I just corrected it.

Calling all testers with CPUs that have more than 4 real cores: add -threads core count - 1 to launch options

ryzen 5 1600 @3.8GHz
technical stuff

benchmarks

The default setting is your cores - 1, at a max of 3, just FYI.

im aware of that, i just wanted to know the fps variance