• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

Benchmarking Unity graphics performance in WebGL

There are other Windoze system in the house. I have a Radeon RX 580 graphics card. I'm most certainly not planning on moving hardware components around trying to get this stupid benchmark to run.
My youngest sister has an i7 8700k + RX 580 8GB in the system I built for her so I can ask her what result she gets in this test.
I suspect it will work fine with a RX 580 8GB (Polaris architecture) and also with AMD Vega GPUs.

I've now done the test on i3wm and it seems to score higher than the results I posted for the LXQt setup. But I am also comparing ROSA Fresh (Yandex & Chrome) with Void Linux (Opera) so it is not an apples to apples comparison. But it is my impression that light window managers (bspwm, spectrewm, Openbox, dwm, i3, Fluxbox, PekWM, StumpWM) get higher results in this test:

2023-08-17-160818_1920x1080_scrot.png

Hardware: Intel i3-3240 + 8GB DDR3 @1600MHz dual-channel + NVIDIA GTX 650 1GB + EVO 850 500GB
Software: Void Linux + i3wm + Opera (latest stable version) + proprietary Nvidia driver

There are also persons who question the usefulness of this benchmark. If you look up WebGL games you are going to see that a very large portion of those games use the exact same Unity technology that is being tested here. You will see the Unity logo while loading many WebGL games.
 
photo_2023-08-20_16-11-21.jpg

iPhone 12 mini and Safari. There were still some apps open.
I do think this iPhone could score around 79000 if all other apps were closed during the benchmark.
 
7.png

Hardware: Intel i3-3240 + 8GB DDR3 @1600MHz dual-channel + NVIDIA GTX 650 1GB + EVO 850 500GB
Software: ALT Sisyphus + LXQt + Chromium 115.0.5790.110 + proprietary Nvidia driver

If you open about seven light apps in an iPhone 12 mini and then play a WebGL game you have the same performance as on an eleven year old Intel i3-3240 + NVIDIA GTX 650 1GB where you don't open any other apps.

The Intel i3-3240 is a CPU that will still be sufficient for office work and light gaming in 2023. After all, the iPhone 12 and iPhone 12 mini are currently still being sold and they are not 'super cheap'.

My impression is also that ALT Sisyphus is slightly faster in this benchmark (and also in WebXPRT 4) than the average Linux distro.
Not as fast as Clear Linux, but faster than 85%
 
Last edited:
screen71.png

You can see the hardware and software in the output of neofetch.

I replaced only the CPU, motherboard and RAM in my PC and now I had two problems with this test. Although with the old hardware I did not experience any problems.

With the CryptoHash Script, I got an error message.
The 2D Physics Spheres kept running indefinitely, as if the test was not made sufficiently demanding.

The problems reported by other users may be CPU related, and not GPU related as I previously thought.

Vivaldi scores better than other Chromium-based browsers so I did the test with a recent Vivaldi snapshot.

Update: The 2D Physics Spheres works on Vivaldi snapshot.

screen76.png


2D Physics Spheres didn't work in a different browser so I figured it wouldn't work overall. But specifically in Vivaldi it works fine. It also scores pretty well in this test.

There is only one specific test that does not work on my system in Vivaldi and that is the CryptoHash Script.
 
GPNjgnb.png


Hardware: Intel 12600KF (stock) -- Kingston 6000 MHz CL36 -- Sapphire RX 7600 -- BIOSTAR B760MZ-E PRO -- Antec P6 -- Xilence XP550 -- ARCTIC i35 -- EVO 850 500GB
Software: OpenBSD -- Chromium version 120.0.6099.224 -- bspwm -- open source GPU driver

Vivaldi is faster than Chromium in this benchmark. OpenBSD is doing very well.
This test suggests that OpenBSD is quite capable of playing Unity WebGL games in Chromium.
 
Error memory 2 test win11 google chrome firefox edge

Capture d'écran 2024-01-28 020229.png
 
I tried running this but when the teddy bear thing came up, it went for a bit then a msg popped up about out of memory & if I'm the developer to fix this. All that on a system with 32GB of system memory but using iGPU from i7-11700k.
 
I tried running this but when the teddy bear thing came up, it went for a bit then a msg popped up about out of memory & if I'm the developer to fix this. All that on a system with 32GB of system memory but using iGPU from i7-11700k.

It is a benchmark that may have been last updated in 2015. There are no fixes for possible issues with new CPUs.

I think it is the CPU and not the GPU that is causing these types of problems, as evidenced by my own experiments.

In your case, you can reopen the benchmark, then uncheck 'Animation & Skinning'. When you start the benchmark it will skip the part that gives the error.
 
It is a benchmark that may have been last updated in 2015. There are no fixes for possible issues with new CPUs.

I think it is the CPU and not the GPU that is causing these types of problems, as evidenced by my own experiments.

In your case, you can reopen the benchmark, then uncheck 'Animation & Skinning'. When you start the benchmark it will skip the part that gives the error.
Ok, but I don't see the point anymore of running a test from 2015 in the first place for modern hardware. Modern as in post 2020.
 
Ok, but I don't see the point anymore of running a test from 2015 in the first place for modern hardware. Modern as in post 2020.

Many WebGL games were released in 2018 (or earlier) and never updated.


The newest benchmarks are sometimes (and rather frequently) less relevant than older benchmarks.
 
Many WebGL games were released in 2018 (or earlier) and never updated.


The newest benchmarks are sometimes (and rather frequently) less relevant than older benchmarks.
If one plays webGL games, then yes, but I don't so its no problem for me. :)
 
If one plays webGL games, then yes, but I don't so its no problem for me. :)

If you use the Internet on a daily basis, you may encounter old WebGL content on a regular basis.
Major browser vendors Apple, Google, Microsoft, and Mozilla are members of the WebGL Working Group. Google Maps’ terrain view uses WebGL.
A lot of websites in widely varying domains have WebGL content, shopping websites sometimes use it.
 
hDtniAS.png


Hardware: Intel 12600KF (stock) -- Kingston 6200 MHz CL36 -- Sapphire RX 7600 -- BIOSTAR B760MZ-E PRO -- Antec P6 -- Xilence XP550 -- ARCTIC i35 -- EVO 850 500GB
Software: Alpine Linux -- Vivaldi -- open source GPU driver

What strikes me is that on average this result is very close to the result of jiajo's i7-13700KF.

In my country the i7-13700KF is currently exactly 2.1 times more expensive than the i5-12600KF.
But in this WebGL test the difference between the two CPUs is only around 5.8%

A second thing that stands out is the big difference in particle performance vs. the iPhone 12 mini (which is as powerful as the standard iPhone 12)
The i5-12600KF is almost 13 times faster here, showing that ARM cannot compete in all areas with what is actually an architecture still very similar to Intel Nehalem.

ARM is maybe not spectacular compared to previously old CPU architectures.
 
dwMaWv5.png


Hardware: Intel 12600KF (stock) -- Kingston 6200 MHz CL36 -- Sapphire RX 7600 -- BIOSTAR B760MZ-E PRO -- Antec P6 -- Xilence XP550 -- ARCTIC i35 -- EVO 850 500GB
Software: Calculate Linux, KDE Plasma, Mesa open-source driver, XFS file system, Chrome unstable

In two tests I score slightly higher with the 12600KF than the result posted by jiajo (i7-13700KF).
- Instantiate & Destroy
- Physics Spheres

This is the first time testing it on this system, I can probably get higher results if I run the test again.
 
2024-05-06 11.37.35 files.unity3d.com 43d768e350e3.png


New mobo new cpu new ram same gpu (Driver 24.4.1) still crashes on cryptohash script same out of memory msg despite now having 32GB of system memory and 16GB of VRam and I'm not going to fck around trying to figure why anymore atleast particles works
so if all they do is add all then divide by total tests then I'd still get a score of 290,893 even without the stupid cryptohash crash
the Stupid thing is If I have Vivaldi's task manager open cryptohash runs fine and then immediately crashes on start of Animation and Skinning
 
Last edited:
CfJTY5U.png


Hardware: DELL Latitude E6540 and Intel® Core™ i7-4610M CPU @ 3.70GHz
Software: ALT KWorkstation 10 and Chrome (stable)
 
GEEI6Wn.png


I hadn't used my Calculate Linux partition for a while but today I did a full system update of Calculate. It seems to perform better after the update (in this WebGL benchmark).

I outperform the i7-13700KF from jiajo and the R7 7800X3D from Athlonite in the 'Instantiate & Destroy' and 'AI Agents' tests. I use the Intel 12600KF and RX 7600.
 
NjYCHRa.png


Hardware: Intel 12700KF (stock) -- G.SKILL RIPJAWS @3600 CL18 (stock) -- Sapphire RX 7600 -- ASRock B760M-ITX/D4 WiFi -- fractal design DEFINE NANO S -- bequiet! SYSTEM POWER 10 550W -- DeepCool AG500BK ARGB -- EVO 850 500GB
Software: Calculate Linux, PeKWM, Mesa open-source driver, XFS file system, Chrome 128.0.6559.0
 
4KTLrPE.png


Hardware: Intel 12700KF (stock) -- G.SKILL RIPJAWS @3600 CL18 (stock) -- Sapphire RX 7600 -- ASRock B760M-ITX/D4 WiFi -- fractal design DEFINE NANO S -- bequiet! SYSTEM POWER 10 550W -- DeepCool AG500BK ARGB -- EVO 850 500GB
Software: OpenBSD -current, bspwm, Mesa open-source driver, UFS file system, Chromium 128
 
Back
Top