I've closed the survery after reaching 332 submissions. You can view the statistics here.

I ran the same demo, so that's weird...

I can't add this info now since it would skew the results. I'll probably make another survey a week or two after this including more relevant info and see how results change.

I didn't bother to add it because 200+ is generally all you'd need to know. I plan on doing this again sometime, and including stats for processors, GPUs, PSUs, etc, so maybe I'll add that next time.

And thanks!

I decided to make a survey to figure out people's general FPS in TF2, and so far have gotten around 240 responses thanks to Reddit. However, I'd like some more, and that's where you guys come in. Just click here and fill out the questions. I'll release the data and give some averages once September 25th comes, or when I feel like I have enough responses.

Please note that I made a mistake and forgot to add a "None" option to to question where I ask if you're using an FPS config, but I can't add it now because it would skew the results. Just select "Other" instead.

Added some more convars and a new experimental convar section. Convars in this section may have a supposed FPS gain or none at all, and are undergoing testing. They arent' detrimental to FPS however, by the looks of it. If you're experiencing crashes, then just delete everything in that section and comment it, or, if you want to help out, comment convars until you figure out which one is causing crashes and report it.

Thanks to Comanglia for helping me find some convars. I'm still looking through the entire convar list for TF2, though, to see if I can find anything else helpful.

I gave them a try on my crappy PC and got these results with the benchmark;

(all experimental benchmarks and r_pixelfog on)
2860 frames 73.944 seconds 38.68 fps (25.85 ms/f) 2.656 fps variability

(filesystem_native set to 1, host_thread_mode set to 0)
2639 frames 68.168 seconds 38.71 fps (25.83 ms/f) 2.682 fps variability

By the looks of it, the FPS variance definitely went down from yours either way, but I don't think they did anything that would affect the demo. Perhaps it'd be more effective in real play, I don't know, but they don't seem to harm anything. I am, however, leaving r_pixelfog 0 commented since it seems to lower FPS in some situations.

Added some new convars;

mod_load_anims_async 1             // Loads model animations async
mod_load_mesh_async 1              // Load model mesh async
mod_load_vcollide_async 1          // Load model vcollide async

Note that with the above benchmarks and the one below, all of my standard graphics settings are set to their lowerest in windowed noborder at 1366x768. Gibs/ragdolls are on.

2639 frames 66.843 seconds 39.48 fps (25.33 ms/f) 2.654 fps variability

Looks like a very small FPS boost if anything, not a big deal but if it's not detrimental and doesn't cause crashes I'll go with it.

Didn't notice this, thanks. I'll add it.

That list is outdated by two years, new convars are sure to have turned up or been missed. In fact, I found some experimental ones that I'm about to benchmark with my config;

ai_expression_optimization 1       // Disables NPC expressions when you can't see them, not sure if it works on regular players in TF2, but that's why it's experimental.
//r_pixelfog 1                     // Supposedly increases FPS via fog, but it doesn't seem like it. Feel free to play with it to test. I'll run benchmarks later.
fast_fogvolume 1                   // Undocumented but I didn't experience any FPS loss with it enabled, will run benchmarks later.
filesystem_native 0                // I think there's an FPS gain with this, anyway.
host_thread_mode 1                 // runs host in threaded mode if multicore

Looks like some interesting stuff, I'll add those to my config. Past that, I'll probably print all cl_ and mat_ commands to a text file from console and look for potential improvements.

I just wish convar documentation was better, because this can only be painful to do. Hopefully they improve it in Source 2.

That's... interesting. Why would it decrease FPS that much just by a value change? What does it do visually other than allow Halloween hats? If that's the cause, does using the NoHats mod counteract it?

That's what I'm trying to do with mine, and that's why Comanglia released his, to help out my project.

I appreciate that.

I'll give this a try and add it as an option if it works.

EDIT: Looks like r_pixelfog is set to 1 by default. I'll play around with it some more, though.

EDIT2: r_pixelfog 1 means that pixel fog is off by default. Setting it to 0 shows the fog, but it's not like normal map fog that is used as an optimization; this fog decreases FPS by 20~ in idle situations and isn't beneficial in active situations.

I'd rather not argue over this, I'd rather we help each other to make a great FPS config; that's what I'm trying to do with mine, and why I put it up on github.

It is a project intended to be superior in every way, and that's progressive. It is indeed documented more, if you look (Chris didn't describe what most of the convars do, I did), and it's user-friendly/intuitive thanks to the in-game setup.

People have reported FPS gain on Reddit and here, and I myself have gained performance, hence why I released it. As stated a page back, this is a chance to make a community project that can help us all out, instead of arguing about it. What is deemed unneeded, I remove. I never stated that I'm an expert. This project depends on information already released or the information people are willing to disclose for it.

I'll finish off by saying that it matters that this exists; Chris' configs were abandoned and public improvements wouldn't be made any longer. There was no one willing to step up to the plate to update them publicly until now.

On another note, I've updated the config again. The two convars that were causing crashes are now commented and mentioned in the FPS section in the wizard (for people that still want to use them), and cl_software_cursor is commented since it was preventing some people from moving their cursor up, theoretically.