Only for OpenGL based systems: Linux and macOS.

It is on git, so it is updated and has history of all changes (commits). I update the config almost daily according to suggestions and my own improvements. The config is constantly being improved so I don't think there will be a final version. However, I'm planning on polishing much more soon after the pyro update comes out.

Benefits of having it is you get updated graphical and memory libraries for TF2, which improves compatibility and performance. But, you might get VAC banned.

ohmg dude i hate you i had headphones and accidentally had volume on high :((

though it's mostly my fault for being dumb and having volume up

net_maxcleartime 0.5
net_maxpacketdrop 100
net_splitrate 3
rate 167580
net_splitpacket_maxrate 167581

Could you try these CVars? They'll help with network choke. Also it would be helpful if you shared your net settings, or are they Comanglia's default?

Maybe r_pixelfog 0 is the solution. However, I think the less dense fog is only possible with pixel shaders 2.0b, now that I think about it.

Hitreg update? cl_SetupAllBones has been there since release and forces every bone to be set up. Generally, the engine has a good idea of what bones to set up and which ones to not set up, so I don't think the CVar is necessary.

-noipx doesn't exist anymore, nor does -nocdaudio. Were you able to notice a difference in sound or networking with primarysound, snoforceformat and reuse?

Foggy? Could you expand on that please or maybe share a screenshot?

Well bones affect animations in a way, but I don't think cl_SetupAllBones sets up any useful bones. There hasn't been any proof from anyone that cl_SetupAllBones actually helps hitreg.

0 is better because it doesn't force uselessly setting up bones.

In local servers, you'll have to set host_thread_mode 0. Not a dumb question. Remember to set it back to host_thread_mode 1 when you're going to join a server.

They've been there since release. I just looked through the source code to find them.

I don't know why Chris claimed this because I can verify even without source code that nojoy does affect the game's behavior. TF2 outputs "Did not detect any valid joysticks" to stdout when nojoy is not present. Not sure about the mouse acceleration commands since I haven't used Windows in years. But from my memory, I remember them affecting mouse acceleration.

I would definitely recommend cl_SetupAllBones 0. Also any reason why you didn't use -nops2b?

After looking at the code again, I think setting any of the violence_ cvars to 0 enables low violence mode/silly gibs, but I'm pretty sure the code I'm looking at is outdated so I'll have to test in game. The best way to disable gibs is through the prop break max pieces commands, I believe. Not sure about blood, how was this done traditionally?

Found a few more launch options, though I haven't fully investigated them yet

-mat_softwaretl : forces software vertex processing, might be useful for bad GPUs
-limitvsconst : limits hardware vertex shader constants to 256, don't think this one will be too useful
-primarysound : uses a single sound buffer, not sure what the real impact of this is
-snoforceformat : skips force setting of sound buffer format, not sure what the real impact of this is
-reuse : allow network sockets to be reused when busy (sets SO_REUSEADDR), not sure about this one but i tested it earlier and i didn't have any bad results with it, maybe someone with a better feel for net settings can test this
-usetcp : allows usage of TCP, probably useless because I'm not sure where TCP would be used in TF2, I'll have to investigate further

Some more kinda useless ones:
-shadersondemand : initialize shaders on use, though it still seems to allocate memory for the shaders, so don't see any use case for this...
-random_invariant : use the same seed for local RNG every launch, don't see a use besides debugging
-small : allow resolutions smaller than 640x480
-filmgrain : may enable a film grain effect, crashed on my PC when I entered a game

So benchmarks soon then?

He's trying his best, and admits when he is unsure about things.

You know which launch option causes this? Most likely -nouserclip

Also are you still using dx8?

Eh, unlikely. Most likely placebo.

I've found some launch options that you peeps might want to test:

-nosrgb : disables SRGB (limits variety of colors) results in flatter/smoother colors, and brighter lighting
-nouserclip : disables user clip planes and enables fast clipping (I experienced a fps boost with this)
-nops2b : uses pixel shaders 2.0 instead of 2.0b, reduces fps impact of some shaders
-softparticlesdefaultoff : disable depth blending for particles and sprites

Please don't use these settings with my config. I already set these to optimal values in my main config and maxframes.cfg.

Main offenders: r_lod 2 (sets all models lod 2, even if they can go down to a lower lod, use r_rootlod 2 instead and set r_lod to -1), blink_duration 0 (doesn't do anything except divide by zero and possibly cause some errors/memory leaks), g_ragdoll_fadespeed 0 (this is the rate at which the ragdoll fades out, if it is 0 then it just takes up memory trying to do the impossible fade task per frame, though this might have been fixed in Tough Break), and I thought violence_hgibs 0 enabled low violence mode, which is worse for performance.

It is better to remove any extra configs you have for graphics and just suggest values to me. They might be ruining performance.

I know, but most ragdolls disabling configs get a value wrong and cause a memory leak (possibly fixed in Tough Break). I wanted to make sure what values they were using.

How did you remove gibs and ragdolls?

What is LOD tweak?

And I recommend using the highest dxlevel supported by your card. But it comes down to what the benchmarks say, and personal preference. There are some things in dx8 that hurt performance depending on your specs. It makes up for it by disabling some effects not supported by it, but I think most of the important effects can be disabled in dx9 anyway.

Hm, so I guess it comes down to benchmarking it with some perf tool.