Some people experience, some do not. My point was that it really isn't the end of the world if you don't remove it.

Benchmarks for Aspect Ratios

4:3
1280 x 1024

2639 frames 11.506 seconds 229.35 fps ( 4.36 ms/f) 11.492 fps variability

640 x 480

2639 frames 11.525 seconds 228.98 fps ( 4.37 ms/f) 11.855 fps variability

16:10
1680 x 1050

2639 frames 11.643 seconds 226.65 fps ( 4.41 ms/f) 11.369 fps variability

16:9
1920 x 1080

2639 frames 11.843 seconds 222.84 fps ( 4.49 ms/f) 11.428 fps variability

1280 x 720

2639 frames 11.825 seconds 223.18 fps ( 4.48 ms/f) 12.324 fps variability

Considering I've left -dxlevel in on 3 different systems for over a year now and every issue I've had could be tied to something other than -dxlevel it's fairly safe to assume that it's perfectly fine to leave it in so long as you have an autoexec or something that is exec'd upon every load. Keep in mind chris and several others said "cl_threaded_bone_setup 1" was unstable on ALL systems not just low end ones.

mat_specular 0 doesn't necessarily cause a bug with mat_trilinear 0 you just get less FPS for reasons I don't know.

http://teamfortress.tv/thread/7598/tf2-benchmarks#28

You can see -some- of the benchmarks I have done on a older system of mine. I've redone them on my New PC but haven't shared it cause it just confirmed what was already said.

Colorcorrection I left in to mess around with to see what kind of FPS impact it has with various settings. They're all disabled so it shouldn't matter. r_queued_colorcorrection 1 is also set but useless just in case someone decides they want to use colorcorrection as well.

mat_phong 1 is mostly preference for DX9 users I've found non-conclusive results of lower fps on 2 different systems. Which in itself doesn't matter cause DX 8 > DX 9 for fps anyway.

Rate 60000; this is what I use and I did add the comments of the rough rates that would be maximum needed. The 62,000 figure I provided was from a 32 person custom map server (No Heroes Payload)

This is exactly why I posted this thread. Take any of the commands out of mine you want. Listen to my advice, use the mods I suggested or don't.

I put a lot of time into finding the best FPS settings for 2 very different systems but it will not apply to everyones systems. Just trying to share what I've learned.

It doesn't really affect anything, I assume this is all loaded on the GPU and and since TF2 is a CPU limited game it has no effect on framerate. I have them both turned on cause turned off is a little bit distracting and make everything look less distinct.

mat_picmip -1 can usually be set (high quality textures) with extremely minimal fps impact to no impact at all for same reasons as mat_filterlightmaps and mat_filtertextures

This is where I put mine
C:\Program Files (x86)\Steam\SteamApps\common\Team Fortress 2\tf\custom\my_custom_stuff\scripts
and save as a .txt file in your scripts folder

Are you using a laptop? cause how well your cooling is working would have a significant impact on your FPS if your CPU is getting around 80c and throttling itself down to not damage itself.

~~~Updated April 6th, 2015~~~
In light of seeing Rhapsody's CFG thread I thought I send my file out to the public. I've put a decent amount of time into it.

I took the median output of 3 tests from each

M0re's fps config (I removed shadows and ragdolls to make our basically even)

2639 frames 12.157 seconds 217.08 fps ( 4.61 ms/f) 11.614 fps variability

Chris Max Frames with cl_threaded_bone_setup 1

2639 frames 12.085 seconds 218.36 fps ( 4.58 ms/f) 9.717 fps variability

My CFG

2639 frames 11.478 seconds 229.91 fps ( 4.35 ms/f) 13.286 fps variability

Download WARNING unstable for Single/Dual/Tri Core CPUs

Launch Options I used

-dxlevel 81 -w xxxx -h xxxx -refresh xxx -novid -console -nod3d9ex

go ahead, pick it apart, and recommend settings. I'll try to update it. (and yes I put my name in it but kept most of chris's organization)

Other Files (I used these in all tests)

Harbls No Hat's Mod

No Hat's Mod Particle Effects remover (Very Few Unusual Effects remain)

Install Location for the .vpks

Remove Dust from Bullet Holes

Install Location for removing dust on bullet holes

Screenshots

Hud Elements
All hud elmemnts eat up frame rate, turning off commands like cl_showfps, net_graph 1/2/3/4 , Your Crosshair, any extra Hud Crosshairs should save you some if not a large amount of FPS (if it's net_graph 3/4)

fps_max
From my experience fps_max on systems with good cooling tends to cause fairly bad input lag when you're consistently reaching your max. So if you have a Desktop with a fair amount of cooling you'll definitely want this turned off.

Now if you have a laptop or perhaps a desktop in an enclosed space where cooling is going to be an issue, a fps_max I would recommend otherwise you'll suffer from severe micro-stutters and slowly damage your equipment.

cl_threaded_bone_setup 1
basically this adds another thread to your cpu for the bone animations of models.

Please note that this command isn't stable on Singe/Dual/Tri-Core CPUs

r_queued_decals 1
has been reported as unstable mostly by AMD users.

Aspect Ratios
Resolutions don't matter so much in TF2 as TF2 is cpu limited but since their's descrepancies in the FOV of each aspect ratio they have different framerates.

The Benchmarks http://teamfortress.tv/thread/20172/my-fps-config#18
4:3 Lowest FOV, Highest FPS
16:10 Middleground for both
16:9 Highes FOV, Lowest FPS

Commands that improve FPS that I didn't add

hud_saytext_time 0 (Text Chat disabled) default is 12
voice_enable 0 (Voice Chat disabled) default is 1
hud_deathnotice_time 0 (removes the kill feed) default is 6

Didn't add them cause the FPS gain is small for something that is usually personal preference to have enabled. Their's a few more, but you'd probably only turn them off/on just to do a "absolute maximum benchmark".

Highest FPS to date (personal reminder)

2639 frames 11.083 seconds 238.11 fps ( 4.20 ms/f) 11.971 fps variability
4:3 Aspect Ratio, most recent CFG, hud_saytext_time 0, hud_deathnotice_time 0

I've never had a problem with my G400 and G400s except the mouse feet wore out on my g400 after 2 years of heavy use.

Literally every mouse has 1 flaw or at least 1 batch of the mice are defective, and people immediately believe all mice under that name have that flaw. I'm sure you've heard of Razer Deathadders with faulty scroll wheels and right clicks, but you probably never experienced that.

dxlevel removed
1.2639 frames 12.695 seconds 207.87 fps ( 4.81 ms/f) 9.598 fps variability
2.2639 frames 12.798 seconds 206.21 fps ( 4.85 ms/f) 9.566 fps variability
3.2639 frames 12.983 seconds 203.26 fps ( 4.92 ms/f) 9.107 fps variability

dxlevel kept
1.2639 frames 12.786 seconds 206.39 fps ( 4.85 ms/f) 10.396 fps variability
2.2639 frames 12.775 seconds 206.57 fps ( 4.84 ms/f) 9.395 fps variability
3.2639 frames 12.832 seconds 205.65 fps ( 4.86 ms/f) 9.117 fps variability

Honestly relooking at this it might not actually be statistically relevant. I probs should've tested more. (This is with my custom cfg)

if you had very odd looking motion it sounds like you must've alt+tabbed or changed settings mid-demo

can someone please link some with at least -some- proof that the g400s has mouse smoothing.

The best I could find was people saying it existed at 4000 dpi and they didn't have proof either.

If I recall correctly the BenQ XL2411T was 12 micro seconds faster at pixel gtg than the ASUS I'd assume that's true for the XL2411Z btw 12microseconds = .12milliseconds which probably doesn't matter that much but it's still "better"

I can't seem to find the article to show for that but it was on the BlurBusters site :/

I'll post my benchmarks later today when I get home, and once again "redoing the options each time" doesn't matter when you have a autoexec cfg.

m_rawinput "1"
m_filter "0"
m_customaccel "0"
m_mouseaccel1 "0"
m_mouseaccel2 "0"

-useforcedmparms -noforcemaccel -noforcemspd

-useforcedmparms -noforcemaccel -noforcemspd if I recall these don't actually do anything anymore.

You're already execing the CFG every time you open up TF2 it's detrimental to long term FPS to remove -dxlevel XX

I marked it in bold, if someone is overheating they NEED a fps cap it'll be more fluid and less detrimental to the hardware.

What should I change between these other than cl_threaded_bone_setup 1 r_queued_decals 1 (which probably wouldn't be detrimental on two-core systems anyhow)?[/quote]

I've tested this before if you have less than 4 threads with cl_threaded_bone_setup 1 you'll crash every 10-30 minutes. If you have 4 or more it's perfectly stable.

and the Overheat CFG would basically just add a FPS cap to the other 2. Plus the reasoning behind this is to have an explanation of what each CFG does before people download it and keep them from adding asking "Why do I crash every XX minutes?" "Why am I getting stutters at 150+ fps on my laptop?"

The 8 seconds of black screen at the start is a bit much.

Should try some of these jumps with a winger + FaN