Do you have the possibility to compare with a nvidia card?
Originally Posted by sundown
It's not the same!
With my nvidia card (good ol' Geforce 7900GS) Firefox seems to be faster than in Windows XP (Vista is slower) even when compiz is active and with smooth scrolling activated, just unbelieveable.
Today I updated my 3d driver install scripts to work with Ubuntu hardy too, but what do you think I found out: 5-4 was silently updated to
label="ATI Proprietary Linux Driver-8.493.1"
DATE="Thu May 22 17:10:33 EDT 2008"
The first (and hotfixed) driver was:
label="ATI Proprietary Linux Driver-8.493"
DATE="Fri May 16 11:25:45 EDT 2008"
Why dont you call it 8-5-1 or whatever? Even nvidia would raise a 3rd value for minor updates. As this crappy driver still has the modelinebug and and test that every month of course I did not have 8.493.1 as target for auto remove modelines, which of course caused some trouble. I absolutely dislike hidden updates with different revisions!
That seems odd; let me ask about it. I agree that would cause problems for distributions.
It was a non-functional change, removing the textual description of unreleased ASICs.
Originally Posted by Kano
Well it would be nice when you tell me that until the modeline bug is fixed, as I still hope that this will work in the near future again and use something like:
case $VER in 8.16.20|8.18.6|8.18.8|8.19.10|8.20.8|8.21.7|8.30.3 |8.31.5|8.33.6|8.34.8|8.35.5|8.36.5|8.37.6|8.38.6| 8.38.7|8.39.4|8.40.4|8.41.7|8.42.3|8.433|8.443.1|8 .452.1|8.455.2|8.471|8.476|8.493|8.493.1) RMML=1 ;; esac
Maybe I should use a whitelist... This is the blacklist needed for Debian etch - otherwise the Xserver usually crashes with one or more specified modeline.
Do you have a fully described bug report on this? In general, the modelines should only be needed when your monitor is broken or doesn't follow standards. This allows scope and impact to be understood.
Originally Posted by Kano
A bug that appears in one application, but no others, or an issue with one particular monitor will generally appear lower on the priority list. The ability to reproduce an issue by removing as many barriers as possible is of critical importance.
PTS is a great way to get application problems reproducable, startup problems can only be supported old school - full bug report, with secondary confirmation of reproducability and full system information. The higher the number of full system information that can be provided to support the bug report the clearer it will be where the problem lies and clarifies the ROI of resolving an issue.
fglrx is the ONLY driver which crashes the xserver when you specify additional modelines. And you need em especially when you use non standard res on a CRT - because you want to use 1152x864 @ 100 Hz (funny thing to try: you can select this 100 Hz in the amdccle but when you check the monitor it is never 100 Hz) or 1280x960 @ 85 Hz. It is impossible to use those. Just TFT you usually only use with the highest res stored from EDID - which usually works. A driver has to use modelines when specified and use auto generated (defaults) + EDID if not. There should be no discussion if it is really needed or not, the xserver has to do this because it is standard.
Update: I went AMD/ATI. :facepalm:
Well, I also recently bought a new computer, with AMD 780g (ASRock A780FullDisplayPort + Phenom 9550). While the processor is somewhat sluggish (really my fault for not going for the Black Editions, but it was an upwards of €40 difference), I'm quite content with the performance of the on-board HD3200.
Sure, it could be better, and it _is_ much better on Windows (I shortly dual-booted to check the support on Windows, and I must say, I wasn't all that impressed - sound is better supported on Linux, even with the official AMD drivers). Videowise, AVIVO leaves the processor pretty much bored at 20-30% while playing an intensive h264 video, while on Linux the usage peaks to 70-100%, resulting partly in stuttering and forcing me on some files to fall back to the teary -vo xv on mplayer.
But, I still don't regret my choice of an AMD/ATI board. Sure, things might have been overall better if I chose an Nvidia chip-set or even chose Intel over AMD. But:
1. I'm glad I've supported a company that at least tries to be open-source friendly. "It just works, and works well" isn't an argument, or better, isn't the only facet to consider. Binary blobs are known to be buggy, and are not really renowned for their security.
2. Linux is still, honestly said, an exotic platform on the Desktop for all purposes. If you believe in GNU/Linux and want it to thrive, you have every reason to sacrifice performance for open-source friendliness and buy AMD. If you want cutting edge performance on the Desktop, there is already an OS where it's possible, where you don't have to whine about performance or stability and things just work.
Ultimately, being patient will bring about the desired change -- and the situation is improving all the time, imho.
Tags for this Thread