Hey all sorry about the bump. Added a light mode option to the style. You can change it on install, or just by selecting Light/Dark in the style settings. Also if you are savvy enough you can easily customise the colours yourself.

Image:

Worth noting that you should get the style from userstyles.world for the latest version, and not userstyles.org. The one on userstyles.org is old and probably wont get updated.

Hope you like it

Reinstalled Windows 10 for testing, so might as well test different TF2 tweaks

Specs:
5800X PBO (4.85GHz all cores)
RX 580 8GB stock
4x8GB 3200MHz CL14
Windows 10 IoT Enterprise LTSC 2021
benchmark_test

As a reference, from a completely stock TF2 I scored 339FPS. 0 tweaks or custom files.

After testing, this was what I settled on:

4812 frames 8.643 seconds 556.74 fps ( 1.80 ms/f) 68.202 fps variability

This was with DX 8.1, CleanTF2+, 1024x768 fullscreen, mastercoms low with viewmodels and match hud off, lighthud.

Some interesting things I found:

• XMP on or off made quite literally 0 difference. 2133 JEDEC vs 3200 CL14 netted me 0.14FPS. Not sure why
  
• -dxlevel 100 reduced my performance by 6%
  
• Turning off both match HUD and viewmodels gave me the most FPS (other than using a config)

If I wanted to +200MHz to PBO and test CO again, disable chat, maybe overclock my GPU a bit, i'm sure I could get 580FPS+, but I'm not that desperate.

Not bad for Zen 3.

Changelog:

Stylus theme for TF.TV based on the styling of vlr.gg. Dark and Light themes available

https://raw.githubusercontent.com/fakemanoan/TFTV-Stylus-Theme/dark/screenshots/Dark/Home.png

Something I have been working on in my spare time.

Requires the Stylus extension to work, available for most web browsers. Tested on Chrome.

Pictures

Additionally, there is a HQ icon style I made too. Improves the legibility of certain icons, especially when zoomed in. The file size is quite large however.

Credits:
- Me: the CSS, class icons and nerd stars
- Roblan11: TFTV logo
- Flags: https://github.com/lipis/flag-icons

Download Links:
Theme Download from userstyles.world

HQ Icons Download from userstyles.world

WIP and any improvement/suggestions are welcome. Maybe a tampermonkey script in the future too, idk.

Enjoy!

In 2018 I thought TF2 only really used about 3 cores, but I figure I test again to double check.

Using a 5800X and I am assigning TF2 to specific cores (not HT ones) on game launch.

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

2 cores: +50% performance
3 cores: +11% performance
4 cores: +2% performance

Scaling past 3 cores is very small. Pretty much all gains are gone past 4 cores. So TF2 is still heavily reliant upon single core performance.

A safe bet is probably at least a 6 core Zen 3 or Alder Lake CPU, ie a ryzen 5 5600(x) or a core i5 12400(f), because having a couple of cores spare for multi tasking, or other games/apps that use 6 cores, is nice to have. Dual channel memory is a must with atleast 3200MHz and decent timings.

According to this video, DDR4 vs DDR5 is pretty much the same for CSGO, therefore probably TF2.

I believe the program referenced by number 5 in this document is what he was using: https://docs.google.com/document/d/1v96FmpiOn__WLnygQnrW8Mr1wpIwIJvqdhkvjCpdVaY/edit. iirc pazer was the one who made the fix for corrupt i63 demos also, so it is probably what kaidus was using.

Testing Ryzen 5000 with benchmark_test

Mobile
CPU: Ryzen 7 5850U
RAM: 3200MHz CL22 (JEDEC basically)
GPU: Vega integrated
OS: Windows 10 21H2
CFG: Mastercomfig low, DX8, fullscreen optimizations off
Resolution: 1280x720@60

Both battery and plugged in.

Battery power. 9W CPU package. 3.5GHz cores

4812 frames 23.062 seconds 208.66 fps ( 4.79 ms/f) 31.850 fps variability

Battery power. 25W CPU package.

4812 frames 17.692 seconds 271.98 fps ( 3.68 ms/f) 44.306 fps variability

Wall power. 10W CPU package.

4812 frames 21.026 seconds 228.86 fps ( 4.37 ms/f) 34.879 fps variability

Wall power. 25W CPU package. 4.1GHz cores.

4812 frames 15.939 seconds 301.90 fps ( 3.31 ms/f) 51.681 fps variability

Considering some gaming laptops cant hit 300fps in this game, this is pretty impressive. Still CPU bound though, so the 6000U refresh probably wont help much. 5700G is probably very similar to this, probably a bit higher.

This is also with the Radeon Pro drivers, so not entirely sure how the gaming drivers would compare.

Desktop
CPU: 5800X stock
RAM: 3200MHz CL14-14-14-32
GPU: RX580 stock
OS: Windows 10 21H2
CFG: Mastercomfig low, DX8, fullscreen optimizations off

1440P

4812 frames 10.276 seconds 468.27 fps ( 2.14 ms/f) 61.150 fps variability

720P

4812 frames 10.118 seconds 475.61 fps ( 2.10 ms/f) 61.432 fps variability

4.85GHz on the cores. Maybe +200MHz would net +5% more performance.

In your own source, it uses 97th percentile FPS numbers and not frame times for measuring. 99th percentile or 1% lows FPS is the defacto way to measure a games smoothness, because people know what 30FPS is as opposed to 33.33ms each frame . AVG Frames per Second = AVG Frametime (or Seconds per Frame). Check out this video for more info.

Also, AVG frametime and AVG FPS effectively communicate the same amount of information because they are the same metric. They don't measure input latency directly either, because that depends on your mouse, monitor etc.

I observed 5% improvement in FPS going to directx 8 on a ryzen 6 core and RX 580, which is modern imo. Also on a GTX 960M laptop with a 5700HQ, it improved performance by 10+%. Definitely didn't make my game run worse than DX9 when I used it.

My experience on Linux has been positive. It ran TF2 faster than Windows by a noticeable amount, and it felt smoother ingame also. I got 255FPS on Manjaro Linux, whilst Windows 10 got 234FPS. It feels much better than 20 more FPS in-game though.

I use Lutris and all I did was enable Feral Gamemode and ACO shaders for my AMD GPU and most games just work. Lutris is a must have for any Linux gamer in my view, and gets most games running with the best settings in 1 click. Games like Quake Live, Aim Trainers and general stuff like Microsoft Office ran perfectly using Lutris.

Things that use DLLs I dont think will work, like P-REC. But custom HUDS, matercomfig have worked just fine for me, so presumably most other things will work.

Wine+DXVK can run the Windows version of TF2, then I think stuff that uses DLLs, like P-REC, will work. But I witnessed a 10-20% performance hit, even in DX8 while Linux was in DX9. So I wouldnt bother with that and just use the ingame demo system.

I read somewhere (I think in the mastercomfig discord) that deleting the SDL. so file in the TF2 folders fixes the double mouse input problem with raw input. Seemed to work for me.

TF2 definitely feels better in Linux, but I cant measure the inpug lag differences. People online claim Linux has lower input lag, but not sure how trustworthy those posts are.

You can disable the compositor in Linux, which is basically like Windows Classic theme, and that removes vsync that you cant disable in Windows 10 anymore. Borderless-windowed gaming has similar input lag to fullscreen mode, it's great.

https://www.youtube.com/watch?v=nOWH7lXrx7Q

I can say that forcing vendor IDs for me when I was using NVIDIA optimus in Windows 7 with a driver that wasn't from the OEM helped my FPS a bit and allowed the game render correctly. I dont have the exact numbers on hand, but it helped.

For some reason the game was recognizing my dGPU as the Intel integrated one with the driver direct from NVIDIA. This was the github issue report that helped me: https://github.com/ValveSoftware/Source-1-Games/issues/633

On Windows 10 I didnt have the same issues though. Also apparently helps on some linux systems, but I never tested that.

your results show enabling and disabling HPET is functionally the same in regards to frame time.

from what i understand WIndows doesnt use HPET by default anymore anyway and keeping settings at default is best. As regards to input latency im not sure. I personally have seen no difference.

Recently I have grown tired of Windows BS and decided to try out Linux again. I figured since all the games I play are available on Linux, I might as well give it a go. I was pleasantly surprised.

Here is a comparison with Windows I did. Not completely in depth, but I ran the demo 6 times on each system. Windows was de-bloated and had process affinities 4,6,8,10. On Linux, TF2 was ran via Lutris with Feral Gamemode and ACO enabled (basically I tried to get the most FPS on both)

Testing setup:
CPU: AMD R5 2600X locked at 4.2GHz
RAM: 16GB 3200MHz CL14-14-14-28
GPU: AMD RX 580 at stock speeds
Resolution: 2560x1440
OS: Windows 20H1 (dx98) / Pop! OS 20.10 (dx92)
CFG: Mastercomfig medium-high tweaked

https://i.imgur.com/O5n6SB2.png

AVG Windows: 235.63FPS
AVG Linux: 244.47FPS

Linux wins by 9FPS! It also feels much smoother to play in-game too, but I have no way of measuring that. The only real issue is that fonts look a bit different, but it's not a big deal to be honest.

These findings are supported in other games like CS:GO. Considering you can get Office running via Wine perfectly, and Proton allows many Windows games to run, this might be a viable option for many people. Games like DOOM 2016 run flawlessly with about a 2FPS loss, despite not being supported on Linux. Pretty cool. If you have some free time try it out!

On the ETF2L page for P-REC the mirror link works just fine.

In case that doesn't work either, the archive.org version works too.

The 3800xt (according to GN CPU benchmarks) does about 128fps stock. I would expect out of a comet lake CPU you will get 5.1ghz max (silicon lottery). This is about 146fps.

this info is not confirmed. but to be safe, i will lower the numbers a bit from leaks.

If Zen 3 is 10% faster in games than the current Zen 2 CPUs, then that's 141fps. A 3% difference. Comet Lake CPUs you will have to manually OC and it will consume over 250W - and you get no PCI-e Gen 4 support.

And that's a conservative 10%. Most recent leaks suggest that Zen 3 beats Comet Lake in CPU bound games. And will probably use less power. Besides October 8th isn't that far away, so if you don't need a new CPU right now, might as well wait to see. I doubt 10700Ks will go out of stock in that time.

As for RDNA 2, there is a lot of conflicting information. The most realistic thing I've seen is that it will perform about the same as a 3080 and will consume less power for the top model. I doubt we will see another Vega again, but who knows. The only way to know is wait until 28th October. I just wouldn't count RTG out just yet.

Why do you think NVIDIA increased performance so much over Turing this time for the same prices? I can tell you with some certainty that NVIDIA isn't doing this out of the goodness of their heart.