Hm, wasn't able to find it in the code at first, sorry.

What folder did you install it in and what were your launch options when you tried my config?

mat_forcemanagedtextureintohardware should be 0.

If that doesn't work, a workaround that reduces performance is mat_mipmaptextures 0

Ok, thanks for the testing.

Don't use the -NoQueuedPacketThread launch option with anything but net_queued_packet_thread 0.

DX9 can give better FPS, except if you have a very lowend or old GPU.

Not sure what you mean by outdated, they still work

Doesn't exist

Sorry to hear that! Did you properly reset TF2 and install it in the correct folder with the specified launch options? and what preset did you use?

Don't see how it's bad

The lower the value, the most more consistent/stable the framerate.

There's a lot of factors that go into that. Video drivers are more mature and maintained for DX9, DX9 itself has a lot of optimizations and improvements to resource control and GPU commands, and Valve has improved its DX9 renderer a lot more than DX8. It's hard to pin down any one thing that has the most effect, since there are a variety of improvements that make it better.

Thanks for these benchmarks.

hi, thank you so much for doing this. im a big believer in just being happy and nice to people because that's just a good thing to do.

these compliments will go a long way and start a chain of happiness in people's lives so thank you for all that you are doing!!

warning: nerd

i thought i was some leet hax0r so i wanted to call myself webmaster.

however, webmaster is an actual website job title so most sites don't allow it as a username for impersonation reasons.

so, i tried to think of alternative names. i decided to name myself master of all the .coms, because that's essentially all of the internet. i felt it would be fitting to shorten it to mastercoms for my username.

unfortunately, i was not aware of .tv so i'm afraid i'm powerless here

I've played TF2 since it went F2P and I loved every moment of playing it back then (and still do), so much so that I became a PC gamer after using an Xbox for so long and only because of how much I loved this game.

But I owe so much more than just my PC gaming habits to Team Fortress 2. TF2 was a masterpiece of game design and inspired me to learn more about how to create and code my own experiences for people to enjoy as much as I had enjoyed TF2. I don't think I'll be able to achieve something as great and with such a legacy, but I do appreciate the resources Valve has given to not only enjoy the games they make, but learn how they work in-depth and help make your own games and other software.

I know we're all waiting for a great update on the 10th anniversary and we're worried if Valve can deliver. However, I believe our very worry is a testament to how great and capable of a company Valve is. If Valve were a bad games studio, we wouldn't be here discussing it among ourselves, nor would we even be expecting another great free update from them after a decade.

It's amazing how far we have come, with Valve and through our own community alone. Here's to another 10 years. We have both the game and community to take us there, and I'm excited for the journey we have ahead of us.

Blog post comes out. Some stuff about the anniversary. Then, at the end of the post, they say they don't have a date for the Pyro Update yet but they are working on it and it should come out soon. Valve then releases the Pyro Update an hour later.

Then they can say they were early with the update.

That's entirely possible.

You shouldn't have conflicting dxlevels for -dxlevel and mat_dxlevel, and I don't recommend using DX8 for performance reasons on anything except old cards. DX9 is a much more efficient renderer that takes advantage of hardware better and has a bunch of performance optimizations Valve has not ported to DX8. Though, I totally understand if you're using it because you like the look and feel of it.

The dxlevel overrides reduce particles more aggressively. They will be applied on dxlevel 80, 81, 90 and 95.

i really don't know what to say, but this makes me so happy!!!!! thank youuuuuu soooo much <3 love you too

Sorry, I missed this question. I've been thinking about having in-depth explanations about CVars, commands and engine mechanics in TF2 on the mastercomfig wiki. However, I'm going to wait until after the next major update before I consider actually doing it.

Hm, maybe it is some memory allocation issue. Try r_flashlightdepthtexture 1 in addition to the commands listed before. Not sure if it will allow for bleeding shadows, though.

I've also released 4.4.7, quite possibly the most boring release of all time.

I'm anticipating the Pyro Update to come soon and for it to change things in terms of cvars and optimization, as every major update usually does, so I'm trying to refrain from doing bigger changes to my config until then.

The Pyro Update will probably result in a bigger update for my config and thus the release of 4.5.0, or if we get a new engine (volvo pls つ ._. ༽つ giff source 2), 5.0.0.

Thanks for the detailed report. I'll look into this, but it would help if you could share your dxdiag info (search for dxdiag, run it, then click the save all information button at the bottom.

cl_pred_optimize doesn't affect if you predict or not, it controls how much prediction data you reuse if there's no new information from the server, even if the server sends you an update. If you have it at 0, it will repredict everything all over again, even if it has no new information from the server to change the prediction result. 1 will prevent repredictions when you don't get new server information, and 2, in addition to enabling 1, will also prevent repredictions on information that confirms your prediction had no errors. And no, setting this to 1 will not resynchronize with the server, because it only does that if you have errors, regardless of prediction optimizations. So in the cases where 2 would occur, it would not resynchronize if you had 2 off.

Also, I think you misunderstand prediction, and my bad for not fully explaining it earlier. It isn't used for when you lose connection to the server or anything, it is used so your player and weapons react instantly to your keyboard and mouse, and the resultant actions are smooth (like moving forward, your player will move forward smoothly instead of the increments you receive data from the server), because ping and server update rate will delay when you get new updates. And I should emphasize, this only affects your player, no one else.

It isn't about accuracy, it is about repredicting with the same information again, which isn't very useful. So, the optimizations skip repredicting based on unchanged information.

Prediction optimization 1 and 2 are two different cases that don't happen together. 1 happens when the server doesn't give you new information about your player's position/state. 2 happens when it does give you new information about your player's position/state but it doesn't tell your client that its prediction was wrong. What I meant by 2 doing 1 is that when you have pred_optimize 2, it also enables pred_optimize 1.

What Mao-C said is wrong. Synchronization is not done with new data if there were no errors anyway. So, with cl_pred_optimize 1, you would be repredicting with the same data, if you had no errors.

If there are errors, prediction optimization does not apply and does not affect it.

No, prediction doesn't really affect hitreg, it just increases the smoothness and responsiveness of what info you send to the server (like weapon firing and movement).

cl_pred_optimize 1 reuses prediction info from the last time your client predicted if the server didn't update you yet about what's going on.

cl_pred_optimize 2 does 1 and also something similar to 1 if the server gives you an update and the prediction turns out to be true.

Most of the time, these optimizations will happen most often when your frames are higher than your updaterate but can also happen when the server legitimately doesn't have anything to tell you about where you are or that you were wrong on a prediction.

Could you set con_logfile console.log and then share tf/console.log (after you crash)

and I assume you've tried the default config?