AMD Radeon Catalyst: Windows 7 vs. Ubuntu 12.04 LTS

[log0]

To anyone proposing different configurations/distros.

Go ahead grab pts and run the benchmarks. It is stupid simple. Show that your configuration makes a significant difference(not just a few percent) and I am sure it will be considered.

[F i L]

@F i L

Go to http://openbenchmarking.org/ and do that yourself. Its already been explained 100 times why he does it the way he does it.
No matter what distribution he used or settings there would be more like you posting here crying he did not use a certain flag here
or tweak a setting there. Its about default installs, and arch's default is the prompt.

I don't appreciate your ad hominem implication. I'm not "crying" by suggesting a logic reason for someone who spends there time running a website, which boast Linux benchmarking as key attraction, to actually benchmark with more comparable systems. I personally don't have the time or resources to benchmark a multitude of machines, so your cheeky suggestion to "do it myself" is out of the question, this is the reason I browse Phoronix.com in the first place.

And if it's been explained 100 times already with the arguments you just gave, then I find those reasons to be in error. No one expects the absolute best-possible combination of settings in the benchmarks, but that's no excuse for using a provably slower default system when there are other common Linux systems with much better "default" performance (Ubuntu Studio, Linux Mint, Elementry, Pear OS, Arch, ect...). Using Arch is a logical choice because it's designed to be lightweight, fast, up-to-date, and easily customizable.

Remember, I never said that benchmarking a default Ubuntu system was a bad idea, only that it should be second to benchmarking a better Linux setup.

-----

On a completely unrelated note (and not directed at soupbowl): I've ran a OpenGL vs DirectX comparison between Linux and Windows 7/8 recently on two dual-booted systems. The nVidia Card (280GT I think) got significantly better performance with GL on Arch Linux (100+ fps more) vs DX on Win8; however the AMD Radeon HD 4870 O.C. got significantly (as in ~half) the performance on Ubuntu 12.04 that it did on Win7. This was my own code, which largely tests pixel shader ops and render targets (barely any polygons), so take all that with a grain of salt.

[F i L]

@F i L

Go to http://openbenchmarking.org/ and do that yourself. Its already been explained 100 times why he does it the way he does it.
No matter what distribution he used or settings there would be more like you posting here crying he did not use a certain flag here
or tweak a setting there. Its about default installs, and arch's default is the prompt.

I don't appreciate your ad hominem implication. I'm not "crying" by suggesting a logic reason for someone who spends there time running a website, which boast Linux benchmarking as key attraction, to actually benchmark with more comparable systems. I personally don't have the time or resources to benchmark a multitude of machines, so your cheeky suggestion to "do it myself" is out of the question, this is the reason I browse Phoronix.com in the first place.

And if it's been explained 100 times already with the arguments you just gave, then I find those reasons to be in error. No one expects the absolute best-possible combination of settings in the benchmarks, but that's no excuse for using a provably slower default system when there are other common Linux systems with much better "default" performance (Ubuntu Studio, Linux Mint, Elementry, Pear OS, Arch, ect...). Using Arch is a logical choice because it's designed to be lightweight, fast, up-to-date, and easily customizable.

Remember, I never said that benchmarking a default Ubuntu system was a bad idea, only that it should be second to benchmarking a better Linux setup.

-----

On a completely unrelated note (and not directed at soupbowl): I've ran a OpenGL vs DirectX comparison between Linux and Windows 7/8 recently on two dual-booted systems. The nVidia Card (280GT I think) got significantly better performance with GL on Arch Linux (100+ fps more) vs DX on Win8; however the AMD Radeon HD 4870 O.C. got significantly (as in ~half) the performance on Ubuntu 12.04 that it did on Win7. This was my own code, which largely tests pixel shader ops and render targets (barely any polygons), so take all that with a grain of salt.

[EDIT: forums is fucking up, third tries a charm]

[devius]

What I'd like to see is more graphics cards. Can we expect the same performance differences with lower specced cards? Oh, and BTW it seems that nvidia has more tests that favor Ubuntu. I'd like to see Qaridarium unravel that one

[kraftman]

Ubuntu Studio is also LTE

It has a low latency kernel

Kernel that can be installed for testing

MS WOS has a 300 Hz kernel, Ubuntu a 100 Hz one, and Ubuntu Studio a 1000 Hz

Ubuntu has 250Hz kernel. Suspend composition should give better results under Ubuntu, but it's disabled for now.

[davgarcia]

Ubuntu has 250Hz kernel. Suspend composition should give better results under Ubuntu, but it's disabled for now.

Moreover, Ubuntu has CONFIG_NO_HZ and CONFIG_HIGH_RES_TIMERS switched on:

$ cat /boot/config-$(uname -r) | grep -E "HZ|TIMERS"
CONFIG_RCU_FAST_NO_HZ=y
CONFIG_NO_HZ=y
CONFIG_HIGH_RES_TIMERS=y
# CONFIG_HZ_100 is not set
CONFIG_HZ_250=y
# CONFIG_HZ_300 is not set
# CONFIG_HZ_1000 is not set
CONFIG_HZ=250
CONFIG_MACHZ_WDT=m

Which means it can use much higher resolution timers. Have a look here:

http://www.advenage.com/topics/linux-timer-interrupt-frequency.php

All in all, I think Ubuntu default settings make for a good enough gaming platform (there's room for improvement though).

[ohdiego]

Is there any step by step guide for AMD CATALYST installation for Precise Pangolin, I used to have no problems on Lucid but now in Precise I cant find the way to install it correctly

[baffledmollusc]

To be honest, I'm pleasantly surprised how much on par the Windows and Linux drivers versions are. I wouldn't have expected that, and removes one reason companies might trot out for not porting games to Linux.

To put in my two cents regarding the benchmarks on stock vs tweaked configurations, I much prefer seeing stock results. Sure, I could tweak my system six ways to Sunday, but, frankly, I really can't be bothered. I suspect most people fall into that category too, or the the category where they have no clue *how* to tweak their system. I suspect the number of people who bother optimizing their system is a very small percentage, so a stock comparison gives the most information about how drivers will perform for most people.

That said, out of academic interest, I would always be curious to see just how much performance could be obtained by different kernel configs, window managers, XOrg config file tweaks...

[curaga]

We need an AMD APU test with the high performance profile before we can conclude that.

[smitty3268]

then the concluding is the same with the sentence: if you want want to be a "hacker" who tweak your driver you better buy a Intel system.

Why? The intel defaults are much better, so they wouldn't need to tweak the driver then.