I think Kano got it right here. I am not saying that the test is useless, simply because it compares different drivers.
Originally Posted by Kano
Actually, the article reinforced it to me again that my ION purchase was right. I truly believe that ATI has better hardware per $, but what does that matter when it can't run right. Included in that is the fact that their proprietary driver doesn't work on 10.04 yet (a month before release!).
I would love to use ATI's hardware in my HTPC setups, but since they don't even have XV acceleration yet, I am not interested.
But it would have been nice to have 1080P/720P video tests included. I already know the outcome, but it still would keep one up to date (even if there is no change).
As a free software advocate, I liked this benchamrk because it shows that an incomplete open driver is useable. It can't compete yet ofcourse but it can play some games and that's terrific. Most important at the moment is to have workable drivers and only when this is a fact we should ask for maximum performace.
I use ATI. Are you saying that the opensource driver lacks xv. Opengl makes a pretty good alternative renderer. Mplayer is about the best video player around. Compile it for your processor extension and watch it fly.
I think gtrawoger was saying that Xv on fglrx has some tearing.
The open drivers have supported Xv for a long time, unless you're talking about Evergreen where acceleration including Xv is WIP.
I registered to complain becuase this article was so bad!
I am a regular phoronix reader and am suprised at the misleading article you have posted. There is absolutely no legitimate justification for the way you handled this benchmark. Proprietary NVIDIA vs. Open-Source ATI is not a fair comparison, and the lack of fglrx, nv, and nouveau, etc. make your intentions quesitonable.
You have the responsibility of giving people information that is not misleading. fglrx is the equivalent to nvidia's proprietary driver and may even outperform it.
Many people will assume ION is better than 4330 wihtout the knowledge to realize your test was seriously rigged.
In fairness, this was one of those "I have the hardware, and I've already installed Lucid, what can I do with it *today* ?" articles.
I imagine there will be a proprietary-vs-proprietary follow-up at some point.
Phoronix has invested a lot of time in automating benchmark suites. All you had to do was install fglrx and rerun the same tests, and you would have had a much more informative and less misleading benchmark. The issue here is that anyone with any knowledge of how these things work could have predicted the results of this test. "Proprietary driver beats open-source driver badly in 3D accelleration!" The only variable is how many times faster is the proprietary one. So the people who follow open source hardware drivers learn nothing, and those who are new to these things think "ION beats 4330 in linux" as they can't always understand the specific circumstances that lead to this result.
Originally Posted by bridgman
Yes and no - the tests were run a couple of days before we released the proprietary driver for Lucid.
Now that the driver is out, it's possible to run a proprietary vs proprietary benchmark on Lucid.
Maybe Michael will run an fglrx vs Nouveau comparison just to keep everything completely fair and balanced
Don't be silly flrgx didn't work on Lucid last week how could he have possible have tested it! Its not like he could have used a different distro or anything. ;-)
Originally Posted by psperl
Yes, that's true, fglrx not supporting the newest xserver is indeed annoying. That problem should end very soon though and this test should have been run on Ubuntu 9.10, which is the current release version of Ubuntu that most people are using, which fglrx supports.