I don't think a lot of people realize how expensive net_graph 3+ is to render. I have a 4670k @ 4.2GHz and I can easily lose 40-50% of my fps from net_graph 3. Also, you're starting at a wall in your "before" screenshot, vs half of upward in your "after" screenshot. If you want to compare before/after fps, you should be using a timedemo or another deterministic benchmark.

If you need to test things in local servers, then go ahead and set this command to 0. Unless you have a single-core cpu, your fps is going to benefit greatly from having this at -1.

Mat_alphacoverage 1?

Don't really want to inject another demo file into the existing TF2 Benchmarks thread if people don't think this demo is an accurate representation of gameplay. I recorded this POV demo on a valve pub the day after tough break dropped on koth_highpass. It clocks in at a fair bit longer than the previous benchmark demo, but at 4 minutes of playback time at normal speed, hopefully it should more accurately represent the current state of TF2's performance.

http://puu.sh/m3k2S/1d0650ec0d.dem

Same idea should be followed as from the original thread which can be found here. As always, I'd recommend running the demo 2 times and only posting the second result so I/O doesn't interfere with your fps.

If anyone sees a problem with this demo or something that could be improved before everyone starts basing performance numbers off of it, let me know and I'll pull this demo link and try to replace it with a better one.

Disclaimer about my horrible tracking while playing heavy: I recorded this demo while playing with -default on the command line which also prevents you from changing your sensitivity away from the default of 3. For me, that means I was playing at a lovely 0.85 in/360 instead of my normal 8.5 in/360.

4843 frames 33.129 seconds 146.19 fps ( 6.84 ms/f) 17.798 fps variability

CPU and overclock: 4670k @ 4.2 GHz
Graphics Card: GeForce GTX 770

Driver version: 358.50
dxlevel (default is 90): 98
Resolution: 2560x1440
Full-screen or windowed: Full
FPS configs enabled: custom
Shadows enabled/disabled: blobby

0 is not a valid value for mat_antialias. The minimum value for it is 1, and if you set it to 0 it'll just be clamped by the game to 1. The same is true for mat_forceaniso.

640x480 is the minimum resolution source supports, because VGUI only supports scaling up and not down (all coordinates for vgui are based off of 640x480)

Where is the option for this/what is the cvar for this? I can't seem to find it.

yes lets remove all the classes 6s players happen to find annoying

Probably to increase the draw of medic in pubs. I know sometimes I've considered playing med in pubs, then realizing there's only 20 seconds left and deciding against it because I wouldn't have uber by the time the round started.

you post this but not the tf2 patch notes smh

you do realize this isn't csgo.tv

I wouldn't get too excited about a multi-day update, seems it was just left over from gun mettle.

http://puu.sh/lZpCR/0a2a154b70.png

1) cl_crosshair_file crosshair3

2) I can just feel it (with the ambassador especially). Even if I'm wrong and it's just a placebo, it'd still be nice to put it where I *think* it should be.

Is it possible to change the positioning of the builtin crosshair (the one that's changed by cl_crosshair_scale & crosshair 1/0)?

Currently mine is about 1 pixel to the left and bottom, makes sniping a bit more difficult. I'm on 2560x1440 if it matters.

Host_framerate overrides host_timescale. If you set host_framerate you don't need to touch host_timescale.

And Jerm, its important to let VirtualDub set the frame rate to match the audio duration because the actual framerate will end up being something like 62-63 fps.

So you're telling me that loading and managing the additional masking textures for phong, additional masking textures for specular, 2 additional lightmap channels, and almost double the texture memory usage (since pretty much everything uses a normal map) will have 0 performance impact?

Even if it's negligible, there's still an impact. I didn't comment on the significance of it.

mat_phong, mat_specular, and mat_bumpmap all affect fps. I can't comment on the exact impact they'll have, but they all require additional cpu logic in addition to just using more texture memory.