too bad!

is that when you play 30 minutes on the same map or just having the game launched? do you only have to do it once or every 30 minutes?

https://youtu.be/GSfnxqlj2T4?t=284

thanks shounic

I believe it resets on map change. It progressively gets worse as time goes on, but for me it starts to become noticeable at around 30 minutes. If pregame lasts 20-30 mins I think it's worth doing before readying up, but it's usually not that big of an issue in competitive.

You will need to do it every 30 mins or so if the map doesn't change regularly though, like on jump servers. Which is why it's useful for those 3+ hour T6 pleb times.

I mean I saw the entire video. Beyond the jankiness of ignoring assists if another banner is ranked higher,
something that's not brought up are the cases where no debuff of healing are involved.

I swear I've seen it happen several times - That I do 200 damage to someone right before they die and a teammate gets the kill and NOBODY gets an assist.

I've seen this bug a few times as well. Can't think of any on camera, though.

https://github.com/mastercomfig/team-comtress-2/releases/tag/0.0.14

Playtesting on connect tf2.mastercomfig.com

https://github.com/mastercomfig/team-comtress-2/releases/tag/0.0.15

This one is a banger and I highly recommend testing it!

Got rid of a huge bottleneck for particle simulation!

https://github.com/mastercomfig/team-comtress-2/releases/tag/0.0.16

We're playing on connect tf2.mastercomfig.com!

https://www.youtube.com/watch?v=EnBDMZ-9-Dw

If this is being looked at by Valve is there any possibility of bringing mat_picmips below -1? You used to be able to read the signs on the walls with -5 and the textures had less compression artifacts. Seems to me this should be the default for ultra textures considering almost every graphics card from the last decade can run this setting with little to no performance impact.
Example:https://wiki.teamfortress.com/w/images/thumb/3/3b/Tf2compare1pp2.jpg/400px-Tf2compare1pp2.jpg

Mat_picmip < -1 can cause issues with some hardware and assets. The better way to handle this is to just generate those assets with proper mips in the first place, but then you're talking about asset changes as opposed to code changes, so that's kind of outside the scope of this project.

If this gains a foothold with Valve, and we're able to contribute directly to the live game, then absolutely, it's something we'll look into.

Those are some juicy threading changes. I'd love to see some benchmarks or frame graphs with before/after.

Just found out about this, have not seen it in the log but has the reference pose bug been fixed? I hope it isnt because its quite goofy... is there a gamebreaking part to that that i dont know about? Anyways, huge thanks, yall are legends!

Not at all detailed, and only a single run of each (not a bunch of samples), but on a computer with NO background tasks, running at 1440p max ingame settings, 5930k, GTX 1080:

TC2 0.0.16 dustbowlbench: 16250 frames 86.039 seconds 188.87 fps ( 5.29 ms/f) 26.496 fps variability
TF2 8/27/20 dustbowlbench: 16250 frames 94.149 seconds 172.60 fps ( 5.79 ms/f) 14.829 fps variability

I'll be grabbing frametimes comparisons later. Those are more helpful for this sort of thing, but I just didn't have the time before.

This demo is spectating a full lobby of bots (can't remember if 24 or 32) fighting on Dustbowl's first map, until they cap the final point.

https://i.imgur.com/gAfLahF.png

we did it boys. we got the frames.

i was gonna argue that i get those frames in the current build of tf2 but its actually only at 745 with a 1070 i7 6700

the frames arent all that high, this is actually just a screenshot of the first linux build attempt, stuff is broken AF because of a couple known reasons, hence most models being fucked

is that not just because ur being healed by a med- cant give both of u an assist (i dont know if this is a thing, its just a guess).

https://github.com/mastercomfig/team-comtress-2/releases/tag/0.0.17

Love the MvM fixes, gotta try them out with some bots. I'll edit this with any bugs I find.

Okay so if you go to the upgrade station the class you're playing as will constantly say the "Get to the upgrade station" non-stop until you leave the upgrade station. Also for whatever reason certain upgrades that shouldn't be avaliable showed up:

https://i.imgur.com/VejmvuD.png

This is because Team Comtress 2 uses an in-dev version of Jungle Inferno, when things were being reworked a bunch. For some reason, the jetpack and gas passer upgrades appear on nearly every weapon in the game, because they weren't fully implemented. Not much we can do here.

Okay, so uh. For those of you wondering about performance. I keep saying "wow here's how many frames we get in itemtest" or staring at walls or whatever. But here's the real impressive numbers.

First off, some timedemos. You can download this demo file here, if you'd like to test it yourself.

All measurements were done on an i7 5930k, GTX 1080, at 2560x1440, max in-game settings (except motion blur off), no configs.

Dustbowlbench is a demo recorded in TC2 0.0.9 of 24 bots fighting on Dustbowl Stage 1 for a little over 4 minutes. Lots of explosions, tracers, particles, etc. Lots of panning over large fights.

TF2 8/27/20 dustbowlbench: 16250 frames 94.149 seconds 172.60 fps ( 5.79 ms/f) 14.829 fps variability
TC2 0.0.16  dustbowlbench: 16250 frames 86.039 seconds 188.87 fps ( 5.29 ms/f) 26.496 fps variability
TC2 0.0.17  dustbowlbench: 16250 frames 78.406 seconds 207.25 fps ( 4.83 ms/f) 18.302 fps variability

So those are some nice numbers. But they don't tell the whole story.

Timedemo is an inherently laggy function; it basically tries to render a demo as fast as possible, without caring for delta time. This normally means it's very stuttery, as it basically prioritizes rendering game ticks over frames.

With the scheduling and wait changes implemented in 0.0.17, timedemo is smooth as butter. I can't even convey it. I'll record some playthrough later or something so you can really see it. It's absurd.

But that doesn't tell the whole story either. Here's some raw FRAPS data, showing frametimes, as well as 1% and 0.1% frametimes. Remember, framerate does not matter. The only thing that matters are frametimes; you can have a really stuttery experience with half of the frames taking twice as long as the others, and it will still measure with a high framerate even though it feels horrible. Consistent frame pacing is very, very important.

Vanilla TF2:

https://cdn.discordapp.com/attachments/755408387617062942/756883044773789716/unknown.png

TC2 0.0.16:

https://cdn.discordapp.com/attachments/755408387617062942/756886223380807851/unknown.png

TC2 0.0.17:

https://cdn.discordapp.com/attachments/755408387617062942/756887871310528592/unknown.png

Yes, that is literally double the framerate in fights using TC2, at max graphics settings, but also the 1% and 0.1% frametimes are dramatically improved. You can really feel it.

We're going to be running a playtest tonight. Would love to have a full server. Check out the Discord server for more info.

I don't understand this really, can someone give an example of what exactly this will change?

Edit: deleted graph. Might be bad data, don't want to misinform.

https://github.com/mastercomfig/team-comtress-2/releases/tag/0.1.1

Come join us! connect tf2.mastercomfig.com

Testing on Suijin!

https://cdn.discordapp.com/attachments/389091069091905546/759920979710967849/Team-Comtress-2-main-menu.mp4

edit: im getting 160FPS ON 32 PLAYER SUIJIN

ur nuts

Installing this somehow turned on mouse accel in base tf2. The option for mouse accel is not checked in the menu. Mouse accel is not stuck on in tf2classic or team comtress, however.

Lol. Fixed it. I was so focused on mouse accel that I hadn't noticed that somehow raw input had been turned off. Still don't understand how that happened but hey, easy fix!

Thank you mastercoms you are unironically my hero!

Bruh how are you real

Lmao and then he goes "?" Like he isnt mocking this post
Smh

?