Did you use my maxframes.cfg addon?

1

You mean #193? I was saying true toaster configs would only boost FPS with very bad PCs, but my maxframes config was for boosting FPS on any PC.

Also, the launch options I suggested work best on dx9.

@mastercoms

Could you please explain what the following commands do ? I didnt understood everything from the comments you left in the config. Especially for that last one, I'm genuinly intrigued.

// mouse yaw sensitivity
//cl_yawspeed 210
// mouse pitch sensitivity
//cl_pitchspeed 225
// mouse yaw factor, mostly for acceleration
// if you want to change yaw sens, use cl_yawspeed
//m_yaw 0.022
//m_customaccel_scale 0.04
//m_customaccel_max 0
//m_customaccel_exponent 1
// moves the player at a faster rate when changing elevation on a slope
// (only in local self predictions)
//cl_idealpitchscale 1.3

cl_yawspeed makes +left and +right faster or slower. Generally used to make you spin around in circles quickly. Mastercom's comment there is wrong, it has nothing to do with the mouse.

cl_pitchspeed is the same as cl_yawspeed except there isn't an up/down alternative to +left/right so it probably doesn't do anything. Maybe for controllers, but controllers already have their own more direct way of doing this.

m_yaw changes your horizontal mouse sensitivity. Some people like to set it to 0.0165 if they're running 4:3 stretched, but it's preference then, and you probably should leave it at 0.022 otherwise.

m_customaccel and stuff should be obvious, it's ingame mouse accel.

cl_idealpitchscale comparison: 0.8 (def) and 1.3. I can't notice a difference within tf2 though.

Thanks mate ! So cl_yawspeed doesnt do anything basically. I still dont understand what that last command does, and your explaination of it is so different from mastercom's that it makes it quite confusing.

Can someone give me a TL;DR?
For those of us who don't want to read through the whole thread/setsul shitstorm, is there actually any reason to switch over to this from comanglias or is it just gonna give 2fps increase at the cost of not being able to make local servers without your tf2 catching aids and dying in a dumpster fire

It gives you similar fps to Comanglias, but it also looks really nice.

A lot of people were also saying that it is smoother/less stuttering.

I don't think there was Setsul shitstorm, just a discussion with a bunch of misunderstandings.

There is a single command that makes you unable to move in local servers, which you can fix by setting host_thread_mode 0 when you are using a local server.

net_compresspackets 0
net_splitrate to 3
net_queued_packet_thread 581304

I tried the above cvar values, and got no negative effects. net_queued_packet_thread had 0 impact FPS when benchmarked, but perhaps that's because it's offline? Not enough time with them yet to be sure, but I must say hit reg "feels" better (again, still totally in the placebo zone, but interesting)

The FPS impact of net_queued_packet_thread is entirely dependent on your CPU.

Did you test "net_splitrate to 3"? Just making sure...

Here are my benchmarks using meb's ancient demo. The more recent benchmark demo causes my TF2 to crash and I'm not sure if there's any other currently used :(.

Regardless, hopefully this is still in some capacity helpful:

http://i.imgur.com/pK8s8Uk.png

What are your specs and OS? Also, did you -autoconfig for when you used a new config? And yes, you should be able to daisy chain exec commands (as in, have exec commands in multiple files). At the bottom of my config, there is a commented exec for using maxframes.

Windows 10, 64 bit
GeForce GTX 1070 with 8gb VRAM
i7-5930K @ 3.5
16gb of RAM

TF2 historically does not seem to like my PC :(

Are you referring to using -autoconfig between each config or using it once before beginning all of these benchmarks? If the former, I can re-do some. I'm not sure of the proper benchmarking procedures :D

I think it is fine to use -autoconfig while benchmarking because I believe that the autoexec runs after autoconfig.

ah just saw your last message, but here are the new benchmarks:

http://i.imgur.com/Dvw0GO8.png

Very strange how maxfps is not having effect on FPS.

How are you installing the config? And what does your exec usage look like?

I had an autoexec with an exec line in it. For each test, it would either be `exec gfx_c.cfg` for comanglia's config or `exec gfx_m.cfg` for yours. `gfx_m.cfg` had the `exec maxframes.cfg` line at the bottom uncommented for the second test. I kept the `maxframes.cfg` filename the same.

I wouldn't write this off as not being an issue on my end, though; I may have accidentally ran maxframes for both tests, or something odd like that. That's the only thing on my end that I can think of, though

Could you share tf\console.log from doing con_logfile log.txt and then cvarlist? (when you have my config+maxframes)

sure, here you go: https://www.dropbox.com/s/lyi55ny9imzafvo/log.txt?dl=0

Seems like my config is not being applied. CVars only in my config and not in Comanglia's like lzma_persistent_buffer 1 are not being set.

Yes, tested them all.

What about a CPU determines the impact? Some feature? I'm on an i7.

Well I meant did you set it to "net_splitrate 3" or "net_splitrate to 3"? It should be "net_splitrate 3".

Depends on how many cores you have, how strong those cores are and how well your OS can manage the network threading using your CPU. I would only recommend it on a 4 core CPU at the very least, so the i7 seems to be fine for that.

Maybe, but 0 gives only a negligble FPS boost and opens the possibility of more significant hit registration issues (the bug it allegedly addresses is very major). Seems better to error on the side of caution, to me.

Yeah, I meant "I set the value for 'net_splitrate' to '3'"

And thanks for the reply, I figured as much. Yeah, it's a good CPU and hyper-threading is enabled and I'm putting it to use (all non-hl2.exe processes that don't experience problems with doing as such are set to use non-physical cores, while hl2.exe is set to the physical ones).

You can edit posts. There is no need to post twice within one minute.

That is not how it works.
You can't run anything on a non-physical core. Both threads run on the same physical core. There is no preference either.

tried the maxframes config and it gave me pretty bad input lag. I did like how it looked though.

Pardon.

I dunno, ProcessLasso calls the cores hyper-threading adds "Non-Physical", not sure whether that's technically accurate. Regardless, I get a performance increase by setting hl2.exe to cores 0;2;4;6 and other processes to 1;3;5;7.

Could you please tell me your specs, and also share tf\log.txt from doing con_logfile log.txt cvarlist.

Most of these programs that magically handle your CPU better (like core unparker, and process lasso) are made by people who aren't informed as much as the OS and firmware devs who actually know what they're doing.

Hi all,

would appreciate it if people with strong GPUs could do a benchmark with r_occlusion 0/1 and r_fastzreject 0/1 individually and together.