Pre Steampipe

2639 frames 22.804 seconds 115.73 fps ( 8.64 ms/f) 4.545 fps variability

2639 frames 21.340 seconds 123.66 fps ( 8.09 ms/f) 5.027 fps variability

2639 frames 22.715 seconds 116.18 fps ( 8.61 ms/f) 4.542 fps variability

2639 frames 21.136 seconds 124.86 fps ( 8.01 ms/f) 4.940 fps variability

Tried changing "cl_threaded_client_leaf_system" to 1(def 0), didn't make a difference. Possibly because no new areas of the map are shown during the test.

CPU and overclock: x4 955 @ 3.8
Graphics Card: HD 6850

Dxlevel: 81
Resolution: 800x600
Full-screen or windowed: Windowed
Configs enabled: Chris Highframes
Shadows enabled/disabled: disabled

Additional notes: Second 2560x1440 monitor running from 6450 (no Crossfire).

Edit:

According to a guy on Facepunch in 2010, the threading cvars should be left disabled:

//cl_threaded_bone_setup 1 [b]//Causes crashes on disconnects![/b]
//cl_threaded_client_leaf_system 1 [b]//UNSTABLE![/b]

Further testing required to see if things have changed in the meantime.

Edit edit:

Stabbystabby's thread about threading stability on SPUF http://forums.steampowered.com/forums/showthread.php?t=2971352

not sure what cvar makes bots move quickly or what cvars you inherit from someone else's pov demo; maybe someone can help

also wow dude you more than tripled your avg fps

//cl_threaded_bone_setup 1 [b]//Causes crashes on disconnects![/b]

Hmm haven't experienced this yet, I will definetely say something if I do start experiencing this, but how I think cl_threaded_bone_setup works I would not be surprised if this is true if someone had a processor that didn't support to many threads (most likely less than 3, and you will probably have issues).

also like to say adding -32bit to my launch option increased my average FPS by about 3, but this should only help if you're running a 64bit OS.

other interesting results

Pre-SteamPipe

2639 frames 26.708 seconds 98.81 fps (10.12 ms/f) 4.014 fps variability
2639 frames 26.091 seconds 101.15 fps ( 9.89 ms/f) 4.014 fps variability

the only difference between those 2 results, is me moving my mouse around and pressing keys occasionally. (the one with less fps I was moving my mouse/using my keyboard)

2639 frames 22.900 seconds 115.24 fps ( 8.68 ms/f) 4.659 fps variability

CPU and overclock: i7-3770K @ 4.3 GHz, 4x4GB DDR3-2133 CL11
Graphics Card: AMD Radeon HD 7970 @ 1150/1650

Driver version: Catalyst 13.5 Beta2
dxlevel (default is 90): 90
Resolution: 1920x1080
Full-screen or windowed: Full
FPS configs enabled: None
Shadows enabled/disabled: Enabled

Additional notes: SteamPipe Beta (not that this really matters, I have an SSD anyway).

Because of how unvaried the current benchmark demo is, I recorded another one, which includes a lot more camera movement and some places with no action. It's about 2 minutes long, still shouldn't take too long.

Here's a DL link https://mega.co.nz/#!YoFDiCZC!CTaoM0DYyh_KwuVUf87YEwEeQPBOYJi7pZ2Gc2dxjaw

My time for it with dx9/shadows off/gibs+ragdolls off

8509 frames 89.021 seconds 95.58 fps (10.46 ms/f) 8.416 fps variability

Same settings, launched with Razer Game Booster

8509 frames 90.728 seconds 95.79 fps (10.66 ms/f) 9.643 fps variability

8509 frames 89.021 seconds 95.58 fps (10.46 ms/f) 8.416 fps variability

8509 frames 67.324 seconds 126.39 fps ( 7.91 ms/f) 7.479 fps variability
(same specs as post #68)

2639 frames 17.599 seconds 149.95 fps ( 6.67 ms/f) 7.077 fps variability
C

CPU: i7 2600k@4.2 Ghz
dxlevel (90)
Resolution: 1920x1080
Full-screen or windowed: Fullscreen.
FPS configs enabled: Chris dx9frames
Shadows enabled/disabled: Enabled.

8509 frames 89.021 seconds 95.58 fps (10.46 ms/f) 8.416 fps variability

8509 frames 87.177 seconds 97.61 fps (10.25 ms/f) 5.715 fps variability
same settings that were in post #36 difference of about 4 fps

8509 frames 89.021 seconds 95.58 fps (10.46 ms/f) 8.416 fps variability

While I appreciate the thought, keeping the benchmark demo the same from the start will make everything much easier. Additionally, my benchmark demo simulates a worst case scenario the entire duration of the benchmark, which is much more helpful for the people trying to make sure their avg fps exceeds a certain minimum (say 120 FPS) so they can reliably predict if their FPS might drop in a critical 6v6 moment like a midfight.

Attempting to run the benchmark after Steampipe gives me a bunch of

Doesn't seem to be any increase in FPS.

2639 frames 23.640 seconds 111.63 fps ( 8.96 ms/f) 4.906 fps variability

i5 2500k
gtx 560
maxframes config
1680x1050 at 120hz
...looks like i need to get more frames

2639 frames 25.157 seconds 104.90 fps ( 9.53 ms/f) 6.542 fps variability

CPU: i7-3770k @ 3.5GHz (not OC'd, so lazy :( )
Graphics Card: GeForce GTX 660

Driver version: 306.97
dxlevel: 90
Resolution: 1920x1080
Full-screen or windowed: Windowed (native res)
FPS configs enabled: default, this means ragdolls w/ no fade.
Shadows enabled/disabled: enabled

Additional notes: Two monitors. Aero enabled.

I'll update the driver and post results.

After updating the gfx driver for the GTX 660 to version 314.22 my results:
2639 frames 24.433 seconds 108.01 fps ( 9.26 ms/f) 4.576 fps variability

W/ update and disabling ragdolls and gibs.

quick comparison:
2639 frames 25.157 seconds 104.90 fps ( 9.53 ms/f) 6.542 fps variability <-older
2639 frames 24.433 seconds 108.01 fps ( 9.26 ms/f) 4.576 fps variability <-newer
2639 frames 21.537 seconds 122.53 fps ( 8.16 ms/f) 8.037 fps variability <-w/o ragdolls+gibs

8509 frames 89.021 seconds 95.58 fps (10.46 ms/f) 8.416 fps variability

True, but the demo is both shorter, so it's a little more stochastic, and doesn't involve any large scale movement, which makes it not have very much rendering load. Though it is late.

2639 frames 19.947 seconds 172.30 fps ( 7.56 ms/f) 6.574 fps variability

i5 2500k @ 5.2
2x GTX 680 SC
1920x1080
Chris DX9 frames (shadows disabled)

So I guess TF2 really does like high clock speeds T_T

edit: I'm questioning the consistency of this. I just ran it a second time with the same settings and got 164.4 fps.

^This will likely be true for about 99% of people who play TF2, but people with low rpm HDs and/or bottle-necked by their RAM may see a FPS increase.

Taking an average of 3 results would be better, but it's more time consuming. 5.2 is a pretty sweet overclock

Have you overclocked your 2500K? Maybe mess around with your Nvidia driver settings a bit.

2639 frames 19.947 seconds 172.30 fps ( 7.56 ms/f) 6.574 fps variability

I've tested this demo in a lot of different ways, things as simple as moving your mouse or pressing a key effects your FPS. I lost about 2% FPS just by doing those things in that test.

I've also played pubs for about and hour and a half and my average FPS went from 101.8 to 89.6 in that demo.

Point, alot of things effect your fps. Also restart TF2 if you've been playing for awhile and are about to do a scrim/match.

For the original demo:

Before cl_threaded_bone_setup 1, cl_threaded_client_leaf_system 1, violence_ablood/hblood 1:

2639 frames 21.222 seconds 124.35 fps ( 8.04 ms/f) 5.175 fps variability

After:

2639 frames 20.044 seconds 131.66 fps ( 7.60 ms/f) 5.503 fps variability

Aero is off, 2 monitors, 800x600 resolution, 100 Hz, dxlevel 95 with dx9frames.

Lucky chip. I had one die so I had to RMA. First one wouldn't top 4.7 without blowing itself up. Only reason I'm debating the haswell upgrade. :P

wanna bump this since watching the demo whenever i play botmaps they just move crazy fast.
i also move ALOT faster for some reason and same goes when i play on pub servers i dont know why

This is a really common issue, i saw b4nny talking about it back when he was streaming. especially if you have a good computer, you'll see in certain maps (like jump maps or tr maps) where if you have the fps_max at 0 the fps can go up to 2000 when you look at walls or go through corridors etc. If there is no human player in your vicinity time will actually go faster.

The solution is to cap your fps. Mine is at fps_max 400.

i have fps cap at 120. no like this never happend before till i tried this benchmark demo.

Example process in a private server, on spawn i notice there are certain areas where i move significantly faster and some areas where i go back to normal speed.
Also i notice the gravity is alot heavier when i do the first sticky jump at the spawn door. when i do the sticky jump i fall down alot faster and quicker that i even get fall dmg from it.

Anyone's m_rawinput changed to 0 after watching this demo or changing to steampipe?

The reason why bots move faster in local servers after doing the test is probably because the way that it tests it is by locking host_framerate to 66, which makes host_timescale vary as it runs faster and slower. "host_framerate 0" should fix the problem.

I'm getting crashes with cl_threaded_bone_setup 1 and cl_threaded_client_leaf_system 1

2639 frames 66.863 seconds 39.47 fps (25.34 ms/f) 4.881 fps variability

2639 frames 152.687 seconds 17.28 fps (57.86 ms/f) 0.877 fps variability

CPU and overclock: M520 @ 2.4 GHz
Graphics Card: Intel HD Graphcs

Driver version: ?
dxlevel (default is 90): 90
Resolution: 1280x800
Full-screen or windowed: Full
FPS configs enabled: Chris dx9frames
Shadows enabled/disabled: disabled

Additional notes: Windows 8, using my shitty laptop, cl_threaded_bone_setup 1

2639 frames 19.867 seconds 132.83 fps ( 7.53 ms/f) 6.139 fps variability

2639 frames 19.838 seconds 133.03 fps ( 7.52 ms/f) 5.569 fps variability

2639 frames 19.861 seconds 132.88 fps ( 7.53 ms/f) 5.279 fps variability

Mean fps of 132.913.

Did this again because I got a new monitor (VG248QE). It seems like resolution has little to no effect on fps, but I'm pretty sure I heard that before anyway.

Not using my secondary monitor at the moment; I need either a new adapter for the secondary monitor or a DisplayPort cable for my monitor.

2639 frames 35.575 seconds 76.18 fps (13.48 ms/f) 2.907 fps variability
//cl_threaded_bone_setup 0

2639 frames 32.501 seconds 81.20 fps (12.32 ms/f) 2.759 fps variability
//cl_threaded_bone_setup 1

CPU and overclock: Athlon II x4 640 ~3.0GHz
Graphics Card: Radeon HD4890 Vapor-X

Driver version: 13.1
dxlevel (default is 90): 81
Resolution: 1920x1080
Full-screen or windowed: Full
FPS configs enabled: Chris highframes
Shadows enabled/disabled: disabled

Additional notes: Windows 7 Pro 64bits, Aero disabled, only Steam and Mumble were running.

Fuck man, I thought it'd be better :( and I don't know if my benchmark is good or bad :/