Reply With QuoteAh, ok I didn't fully understand that. Re-read stuff and it's a bit clearer now, thanks.Originally Posted by Michael
Phoronix
These systems were purchased for the Phoromatic Ubuntu Tracker and so I was in the process of setting up the Lucid build when I decided to quickly run these tests.
Senior Member
Ah, ok I didn't fully understand that. Re-read stuff and it's a bit clearer now, thanks.
Senior Member
I don't understand why many people perceive this benchmark as unfair. Or why substituting the OSS driver by Fglrx would make it any different. This is not a race, and even less one with strict rules limiting the equipment of different competitors to somehow level the playing field. These tests give a quick overview of what you can expect from two similar pieces of hardware where the major difference is the graphics card. These, in turn, are controlled by their respective drivers, with two options for the AMD chip: one was impossible to benchmark since there was no support at the time the tests were made, the other is under development but in a working state. Honestly, how could this be any more representative of the current state of the AMD graphics stack? I personally found the results very interesting, and I'm positively surprised seeing that the performance difference wasn't any bigger. But then of course, I don't have any sentimental attachment to such a thing as a device driver, my ego isn't affected by the results of a graphics driver benchmark, and I don't take what is meant to be an informative test as a competition with losers and winners.
Moderator
Well said, it is what it is, if people don't like how the blob blows away the free solution with lesser capable hardware then they should be contributing to make the free solution better. Hiding the facts doesn't improve anything.
Phoronix Member
I'm sorry, but this is lamest test I have seen on phoronix, seriously...
How can one compare the performance of unfinished work with proprietary work? To show that OS ATI driver still suck? Hmm too unfair...
Prop vs prop, OS vs OS, eg. apples to apples and oranges to oranges. That will show attitude to Linx for both patties and openness as well, fair - VERY fair... If that can't be done, just don't compare, state the facts or test 2D, not 3D which is the main point of competition for 3D accelerators and tests which are just 3D...
This doesn't give any value at all, except, yeah ATI still doesn't support xorg 1.7, we have to live with OSS... That is kinda suggestion for total newbies in linux, I thought phoronix is about more advanced users, bubuntu is for newbies...
I have tried using X 1.6.3 + catalyst and nvidia and latest kernel for NV, both work and 3D is playable for both, NV is better of course. NV power management sux, ATI seems to be better...
Can U plz repeat the tests comparing apples to apples and orages to oranges? It would be awesome, thanx in advance.
Moderator
How is it unfair? It shows the performance delta between two different solutions. What don't you get about that? What don't you get that opensource drivers have a long ways to go to catch their counterparts even when used with superior hardware capabilities. If anything it's extremely" fair " as it shows real results that people can prepare for with the various solutions.
Senior Member
It's unfair because it's incomplete. It sure does show a performance delta, but it doesnt show everything that should be... Whether it was meant to be a quick test or not, incomplete is still incomplete. It doesnt show at least 2 other solutions that may or may not be better for you.
Senior Member
I think some people care more about maximizing their performance than they care about how it is done. With that said, there is no harm in including alternative drivers in the benchmarks for the people that insist that whether or not the drivers are open source takes precedence over the drivers' performance.
Senior Member
I agree. People that want the open source drivers to improve should work on them. It is unreasonable to expect others to do this for them when there are usable alternatives available. Nvidia provides a visible source wrapper for its binary drivers that its customers can modify to keep the driver working with future kernel versions, which eliminates the primary benefit of an open source version and with it, much of the incentive to make an open source version. Unlike the people working on Gnome and Hurd, the general public is not inclined to reinvent the wheel simply because they dislike some minor aspect of it when it does not get in the way of their use of it.
AMD Linux
You probably know this but the fglrx driver has the same source wrapper, called the Kernel Compatibility Layer, which gets compiled during install.
I would argue that the KCL only covers one small benefit of the open source drivers, not the primary one(s).
Posting Permissions