Thorium

[Dixel]

Screenshot with proper file structure, only for the loader usage. @AstroSkipper

[mockingbird]

8 hours ago, VistaLover said:

... But why is Thorium reported to be Google Chrome 109 there? For consistency, weren't you supposed to test the same Thorium variant but on different OS (XP vs 7) ? 

It's been like this for ages -- years...  A minor version update won't produce a different result...  In my opinion, Chrome GPU rasterization is hit and miss when it comes to GPUs.  For 4th generation Core CPUs (haswell) IGP, I turn it off.  I leave it on for the Radeon HD7770 even though it's twice as slow as software rasterization. 

[AstroSkipper]

1 hour ago, Dixel said:

Screenshot with proper file structure, only for the loader usage. @AstroSkipper

Thanks for documenting!  The file and folder structure which is expected by your app was clear to me from the very first, though. However, the structure of my Thorium installation is different and isn't supposed to be changed. 

[Sampei.Nihira]

Chalkboard seems to me to be undeveloped.
For online testing, wouldn't it be better to use Motion Mark?

https://browserbench.org/

 

Edited May 17 by Sampei.Nihira

[66cats]

1 hour ago, Sampei.Nihira said:

https://browserbench.org/

Results make a bit more sense in this one.

Win 7 with GPU: 

Win 7 w/o GPU: 

XP (w/o GPU): 

[mockingbird]

On 5/16/2024 at 10:37 AM, 66cats said:

Mine (both on the same box).
<snip>
P.S. Not sure what that benchmark tests, but scrolling without HW acceleration feels choppy (~20fps) on this box, smooth/fluid (screen refresh rate) with HW acceleration on.

check out "chrome://flags" and "chrome://gpu".  In flags, you can disable rasterization manually without touching the acceleration settings in Chrome and then verify the result in chrome://gpu.

FYI, I ran some more tests to demonstrate what I was saying before about benefiting from disabling hardware rasterization on older GPUs.  In all cases, the CPU is a 4th gen Haswell.  For the iGPU, the tests were run on Windows 7.  For the 1660, it was run on Windows 10.

iGPU with rasterization enabled:



iGPU with rasterization disabled:



GeForce 1660 with rasterization enabled:



The GTX1660 is about 5 times faster in Chrome than the integrated GPU and about twice as fast as the GCN 1st Gen Radeon HD7770 (the HD7770 test was done with Thorium AVX on a 3rd gen Ivy Bridge though, but that probably doesn't make any difference).

Who said software rendering was dead?

[66cats]

19 minutes ago, mockingbird said:

In flags, you can disable rasterization manually without touching the acceleration settings

But the results don't change, nearly identical to switching "Use HW acceleration when available":

#enable-gpu-rasterization disabled: 

#enable-gpu-rasterization enabled: 

 

[NotHereToPlayGames]

26 minutes ago, mockingbird said:

Who said software rendering was dead?

Definitely NOT me!  I disable "HW Acceleration" on all of my HWA-capable computers.

[AstroSkipper]

18 hours ago, AstroSkipper said:

And so that everyone has something to laugh about , I am solemnly publishing the result of my Thorium test on my old Windows XP sweetheart:

And since you can set records both upwards and downwards, here is my result from the graphics test on the motion mark website in Thorium on my old machine:

Edited May 17 by AstroSkipper
Correction

[66cats]

22 minutes ago, NotHereToPlayGames said:

I disable "HW Acceleration"

Not sure what the Chalkboard benchmark tests, but MotionMark results seem a bit more reflective of reality. Why else would HW acceleration be a part of all modern browsers?

Edited May 17 by 66cats

[mockingbird]

4 minutes ago, 66cats said:

Not sure what the Chalkboard benchmark tests, but MotionMark results seem a bit more reflective of reality. Why else would HW acceleration be a part of all modern browsers?

Chalkboard is a very old (>10 years) benchmark released by Microsoft for IE10...  HWACCEL (notify me if you want this benchmark) is an even older benchmark released by Mozilla ages and ages ago when they started experimenting with DirectX 9 support for Firefox.  HWACCEL is great for testing whether layer acceleration is working and I find it useful for checking whether MyPal is functioning correctly (MyPal is now what I consider to be stable and also supports DirectX9 acceleration, but it is far, far slower than Thorium on XP).

Chalkboard and HWACCEL are not what I would consider to be benchmarks today, but rather sanity checks to see if the browser is functioning correctly since they are experimental by nature.  If Chalkboard is underperforming in Thorium or HWACCEL in MyPal then it's an easy way to know right away.

[UCyborg]

Chalkboard looks like an odd case that happens to be slower in Chromium with GPU acceleration.

I don't need benchmarks to tell me Chromium works much fluently with GPU acceleration on my PC, real life examples work just fine to prove it, from smoothness of auto-scrolling, animation on websites, looking/moving around in Google Street, without GPU it feels like 15 FPS at best vs. 60 with GPU, in-browser videos drop frames at various intensity, intensity depends on the OS, typically it's worst on XP, forget about 60 FPS videos in any case (I only use AVC1 or VP9 on YouTube).

MotionMark 1.3, there I get 8 to 9 points on XP, 14 - 16 points on 11 with --disable-gpu parameter passed to the browser, 480 to 484 points otherwise.

My PC was initially going to be pure AMD Dragon, perhaps CPU falls a bit short on CPU intensive tasks, but the platform generally works well for the purpose it was designed for. It comes from era before Google took over the web. Their engineers don't strike me as being good at optimizing software rendering. I dribbled with Android x86 in the past, it's practically unusable on both of my computers without real GPU drivers active. While it works fine on pure AMD laptop I have as-is, that wasn't the case on my desktop that later got ATI GPU replaced with NVIDIA GPU and NVIDIA was totally unsupported at the time. Nouveau generally doesn't perform too well anyway.

Though Android is normally never run without GPU, I don't think there's a smartphone out there where the OS doesn't have appropriate drivers baked in. On Windows, you could probably still get things done with GPU driver disabled if you really wanted.

[AstroSkipper]

In the meantime, I have made various performance improvements to Thorium, on the one hand by selecting suitable flags for my hardware, and on the other hand by generally using a mobile user agent from Android with the help of the ingenious User-Agent Switcher and Manager extension. Today, I have found an additional flag to reduce the FPS when it comes to media which is --max-gum-fps=30. Here is a screenshot from MotionMark which runs now at 45fps and not 60fps.



 

[Sampei.Nihira]

Always the usual advice,from a former professor who used XP until the year 2021 :

https://www.wilderssecurity.com/threads/what-is-your-security-setup-these-days.111264/page-1637#post-2979834

 

Reduce the number of extensions you use.
If you use too many extensions all the possible benefits of your customizations will be cancelled out and/or will be greatly reduced:

https://www.debugbear.com/blog/chrome-extensions-website-performance

Edited May 18 by Sampei.Nihira

[AstroSkipper]

Personally, I use very few extensions in Thorium. And those which are only used from time to time have been disabled by me.