AMD Catalyst 8.5 For Linux

[Porter]

I've had no problems under Ubuntu, and with the latest driver I eliminated all of the extra xorg.conf entries and let the defaults take over... it seems to work great.

It's interesting that people have such different experiences, I can only guess it's due to differences in implementation for each of the different video cards that the driver supports.

[bridgman]

Also different OS distributions, different buses, and different chipsets, and different amounts of "tweaking the OS" since initial installation.

I realized (sadly) today that there are more different Linux distros than there are X developers in the world, even including the Linux driver teams from Intel, NVidia and ATI/AMD in the count. That just ain't right.

[Redeeman]

bridgman:
I am extremely concerned by the fact that the drivers seems to only work on SOME distributions..

Unless some distribution does something really weird(which ubuntu probably does), it should just work.. no crap, no testing needed, just work, as well as it should on suse and fedora, and it indeed seems to for nvidia, Please please explain to me why this is different for AMD?

It does not matter how many distributions there are, the kernel, and X, and applications, are the same deal, the only difference is some provided tools and minor(only in some cases even exists) patches to change trivial things. in short, NOTHING that should EVER be relevant to the fglrx driver..

and most of those "distributions" which there is, is nothing more than remasters of the major distributions, which simply includes stuff presetup, or differently bundled applications, or SOME minor thing, hell, i've seen lots of "distros" which were nothing more than modified ubuntu to have different background image after install!

[mtippett]

bridgman:
I am extremely concerned by the fact that the drivers seems to only work on SOME distributions..

Redeeman,

If you can provide some contrasting examples, we can work through some of deltas between distributions.

NVidia have slightly different approaches but typically fall down in similar positions (except the DRI/GL interfaces since they completely replace that - this lowers the switching cost to NVidia and makes it work easier, but considerably increases the switching cost to anything else).

There are a few areas that could be called stress points with the distribution. In general these are the kernel, X ABI and more recent DRI/GL interfaces. Distributions or users going bleeding edge (.99 or .90 X versions), -rcxx kernels, complicate the matter more. Then hardware/kernel interplay is also a big factor.

Realistically, a lot of the posts of problems leave out critical information that would allow users to see trends. AGP vs PCIe vs PCI, AGP Chipset, CPU, Memory, kernel, distribution, etc. All of those have an impact on the system.

A lot of posts just say "my 3450 doesn't work". That could be any 1 of 3 product families (PCI, PCIe, AGP), on any number of distributions. Then config files and other critical information is ignored, or if it is presented, then it contains clearly cut and pasted items from forums across the internet. I am not saying the driver is perfect, but a lot of the time there is precious little information to understand the issues or establish trends. (Trends are needed to determine if it is broken hardware for a single user, to ensure internal repeatability).

Realistically, NV News Linux Forum shows lots of issues with NV drivers. Graphics drivers are complex beasts (we currently support 301 different pieces of hardware within the drivers), and as John says there are more distributions than developers.

Regards,

Matthew

[Raven3x7]

I' dont recall having any such probs in Arch. I'll update QW to 1.6 and test it later today.

[Redeeman]

Redeeman,

If you can provide some contrasting examples, we can work through some of deltas between distributions.

NVidia have slightly different approaches but typically fall down in similar positions (except the DRI/GL interfaces since they completely replace that - this lowers the switching cost to NVidia and makes it work easier, but considerably increases the switching cost to anything else).

There are a few areas that could be called stress points with the distribution. In general these are the kernel, X ABI and more recent DRI/GL interfaces. Distributions or users going bleeding edge (.99 or .90 X versions), -rcxx kernels, complicate the matter more. Then hardware/kernel interplay is also a big factor.

Realistically, a lot of the posts of problems leave out critical information that would allow users to see trends. AGP vs PCIe vs PCI, AGP Chipset, CPU, Memory, kernel, distribution, etc. All of those have an impact on the system.

A lot of posts just say "my 3450 doesn't work". That could be any 1 of 3 product families (PCI, PCIe, AGP), on any number of distributions. Then config files and other critical information is ignored, or if it is presented, then it contains clearly cut and pasted items from forums across the internet. I am not saying the driver is perfect, but a lot of the time there is precious little information to understand the issues or establish trends. (Trends are needed to determine if it is broken hardware for a single user, to ensure internal repeatability).

Realistically, NV News Linux Forum shows lots of issues with NV drivers. Graphics drivers are complex beasts (we currently support 301 different pieces of hardware within the drivers), and as John says there are more distributions than developers.

Regards,

Matthew

i dont believe there even exists pci or agp 3450.
and almost all of these distributions are simply a "respin" of the major dists, but with changed background image, this does not count in my eyes as "another distribution", or at the very least, not in the eyes of fglrx.

About nvidia, yes, i agree that their drivers are extremely far from bug free, they are in fact what i'd call quite crappy, but with them, i know i get the same treatment whether i go fedora, suse, debian, or gentoo, and this is something i have put to test many times.

I guess you are somewhat missing my point.

The thing i have a hard time accepting, is that "this does not work on distro gah", this to me is useless, if you instead can provide me with information such as "this driver will work with X versions blalblaba, provided you have kernel versions lalalala, with hardware tralalall", then thats fine.

[mtippett]

i dont believe there even exists pci or agp 3450.

There may not be a HD 3450, but the point was that there are different variants of lots of cards (the 3650 comes in PCIe and AGP, the 2400 comes in PCIe, AGP and PCI). (Visit some of the AIB sites to see the AGP ranges.

and almost all of these distributions are simply a "respin" of the major dists, but with changed background image, this does not count in my eyes as "another distribution", or at the very least, not in the eyes of fglrx.

I agree that a lot of the small market share distributions are tweaks on other distributions, but the leading distributions are all different in small user-invisible, but software-visible ways.

About nvidia, yes, i agree that their drivers are extremely far from bug free, they are in fact what i'd call quite crappy, but with them, i know i get the same treatment whether i go fedora, suse, debian, or gentoo, and this is something i have put to test many times.

Yes, they replace the entire 3D architecture (in an incompatible way for the non NV drivers). There are quite a lot of NV users who try ATI after installing the NV drivers and then complain that ATI doesn't work. Realistically as DRI2 and other changes come in, we will approach that invasiveness, but with an increased consistency.
I guess you are somewhat missing my point.

The thing i have a hard time accepting, is that "this does not work on distro gah", this to me is useless, if you instead can provide me with information such as "this driver will work with X versions blalblaba, provided you have kernel versions lalalala, with hardware tralalall", then thats fine.

As per our release notes we support Red Hat, Novell, Ubuntu and Red Flag.

Outside that we have general coverage, but we do not claim support. X version and kernel versions are a poor guide. Just with Ubuntu Intrepid, there is 74,000 lines of patches against the git release of code. That makes X for Ubuntu different in subtle and potentially fatal ways between the patches to Red Hat. Same for the kernel, and other parts of the system.

Regards,

Matthew

[PGHammer]

Ubuntu fails; openSuSE works?

There may not be a HD 3450, but the point was that there are different variants of lots of cards (the 3650 comes in PCIe and AGP, the 2400 comes in PCIe, AGP and PCI). (Visit some of the AIB sites to see the AGP ranges.



I agree that a lot of the small market share distributions are tweaks on other distributions, but the leading distributions are all different in small user-invisible, but software-visible ways.



Yes, they replace the entire 3D architecture (in an incompatible way for the non NV drivers). There are quite a lot of NV users who try ATI after installing the NV drivers and then complain that ATI doesn't work. Realistically as DRI2 and other changes come in, we will approach that invasiveness, but with an increased consistency.
I guess you are somewhat missing my point.



As per our release notes we support Red Hat, Novell, Ubuntu and Red Flag.

Outside that we have general coverage, but we do not claim support. X version and kernel versions are a poor guide. Just with Ubuntu Intrepid, there is 74,000 lines of patches against the git release of code. That makes X for Ubuntu different in subtle and potentially fatal ways between the patches to Red Hat. Same for the kernel, and other parts of the system.

Regards,

Matthew

That smarts!

However, the 8.5 Catalysts *do* create suitable packages under *at least* openSuSE 10.2/10.3/11.0 (for the first time *ever*, I have my X1650 PRO AGP doing proper 3D in Linux; in this case, under openSuSE 11.0 RC1, after repeated failures with the open-source radeonhd drivers in the same distro, and the ati, radeonhd, and these same proprietary drivers in Ubuntu 8.04). I used the same HOWTO that was written for earlier ATI Linux Catalysts for 10.2 and modified it (instead of building directly against 11.0, which, naturally, isn't supported, I built a 10.3-style RPM, but based on 11.0RC1's kernel-source).

The HOWTO is here:

http://blog.linuxoss.com/2007/04/20/opensuse-102-amdati-drivers-installation/

By the by, the hardware is an ASUS P4C800-E Deluxe (Intel 875P AGPset), P4 2.6C, ATI-branded (but sold by VisionTek) Radeon X1650 PRO AGP 512 MB. Now I gets to have some fun.....

[PGHammer]

Add openSuSE

I agree that a lot of the small market share distributions are tweaks on other distributions, but the leading distributions are all different in small user-invisible, but software-visible ways.



Yes, they replace the entire 3D architecture (in an incompatible way for the non NV drivers). There are quite a lot of NV users who try ATI after installing the NV drivers and then complain that ATI doesn't work. Realistically as DRI2 and other changes come in, we will approach that invasiveness, but with an increased consistency.
I guess you are somewhat missing my point.



As per our release notes we support Red Hat, Novell, Ubuntu and Red Flag.


Regards,

Matthew

Matthew, you can also add SuSE (including openSuSE 10.3 and 11.0 through RC1) to the Known Working distribution list (10.2 was already known to work). While 11.0 hasn't been released yet (and thus is not officially supported as a build target), as long as you have the proper tools and kernel-source (which you should if you remembered to add the base-development package via YaST [on installation] or YOU [post-install]), you can build a 10.3-style RPM which your pre-11.0 openSuSE will still install (which is exactly what I did). This was after repeatedly beating my head against a wall trying to get *any* sort of 3D working using open-source drivers included with this distribution of openSuSE (and with both open-source and these same Catalysts with Ubuntu 8.04). It's nice to have *something* work for a change the Right Way.

The original page (based on openSuSE 10.2 x64, but works with the IA32 flavors as well) can be found here: http://blog.linuxoss.com/2007/04/20/...-installation/

Now I get to really have some fun!

[Redeeman]

What i wish for it is to simply work for the upstream code. If distributions modify, that is for them to deal with, but I as an advanced user cannot use "support redhat/suse" at all, i dont use these distributions, what i can however use, is the knowledge that specific code will work, its then up to ME to make sure that software is actually like that, and not modified, and indeed _I_ will take care of this.