What do you think about Phoronix's tests?

Ryzen 5 4500U Six-distro and Windows benchmark test

I think the test doesn't reflect the truth. What do you think? How can Ubunutu, as known as a very bloated Linux distro, can win in overall? I've been following Photonix's tests and Manjaro usually takes the middle, or more the last places. I don't mind it but how is this possible?

What is truth and what is oppinion ? Sorry but this thread is useless.
Enjoy your weekend

I haven't reviewed that specific test and probably won't spend my time that way. However, my general feeling from reading many of his benchmarks is that they often not very useful.

It isn't that they are wrong. He is good at generating data. It is more that they can be misleading because the setup of the tests and what is being tested is often questionable.

He tries to benchmark some fairly broad things and benchmarking is definitely a complicated subject when the tests are broad.

1 Like

I'm following him for a long time just for my curiosity. In general, tests are often on Ubuntu machines. I just thought that "hm, is this test fair totally? he is using different kernels." Besides, most of people who I know think Ubuntu is a very bloated distro and it isn't as powerful as others like Mint, Manjaro or Arch...

Let's just take the very test as an example. Here is the summary.

For the open-source, cross-platform Xonotic first person shooter, Windows 10 was faster than most of the Linux distributions tested when at 1080p with low quality settings. But Ubuntu 20.04 and Fedora Workstation 32 both managed to come out ahead.

Questions that would come to my head would be

  • Why is it 30% faster on Ubuntu/Fedora. Is there a driver missing on the other distros? Is it kernel optimization? As a reviewer you need to investigate that for the benchmarks to be relevant/useful.
  • Another questions might be, who cares? Why do a benchmark where the lowest performers run at ~200 FPS. Does it matter if a game runs faster than that?
  • There is little information about how/what was installed with the various ditros and how the test were setup.

Testing without investigation provides no context for the results and thus limits the usefulness.

Is it fair? I suppose that depends how you define fair. It isn't "apples to apples" because no work was done to normalize things. On the other hand, it is probably representative. That is how those distros performed the way he set them up.

Do I think he is manipulating the tests to make Ubuntu perform better? I don't have any way of knowing that for certain, but I doubt that is the case.

I am not sure that is really a fair statement. For one thing, last I checked Mint used Ubuntu kernels so performance between Mint and Ubuntu should be very close. Ubuntu might even fair a bit better due to gnome usually being a bit more optimized than cinnamon.

Beyond that, most of the major Linux distros are equally "powerful" with the proper configuration.

1 Like

Forum kindly sponsored by