AMD Catalyst 9.5 Driver For Linux Released

[phoronix]

Phoronix: AMD Catalyst 9.5 Driver For Linux Released

If you closely follow the Phoronix Forums you already know that the Catalyst 9.5 Linux driver is available for download. In fact, it has been available since this past Friday on their web server, but it was not officially announced and linked to from their driver web-site until now...

http://www.phoronix.com/vr.php?view=NzI3OQ

[kensai]

Been using the driver since Friday, all working fine, in fact this is 9.4 second edition because there is not much to see here, a bit of improvement here and there and a few megabytes up in size. That tells they are not removing legacy code yet.

<RANT>
Why disable legacy code if you have not removed it, for two drivers releases yet, as far as I can tell people with older cards could still have used 9.5 if only the legacy code was enabled. You are keeping it there, so why not enable it, until you remove it completely?

Still the same old, we support 3 outdated distributions but we don't support Linux. Well, that is why NVIDIA owns this market because they support Linux. I mean, no support for kernel 2.6.29 yet? And kernel 2.6.30 is just a few weeks away?

These guys are still lost, a half-baked open source driver, with a lot and I mean a lot of lack on 3D features. And a badly supported proprietary driver, which can;t be even updated for newer kernels when the need be.

OMG, I am so desperate to buy a new laptop, and never ever commit the same mistake of buying anything with ATI again. I love AMD but if the laptops come in this bad combo I can't buy.
</RANT>

[he_the_great]

While I'm not yet using 9.5, but using the packaged 9.4 drivers in Debian Sid, it has been working well for the 2.6.29-AMD64 kernel. I can't speak for other configurations.

[RagingDragon]

Been using the driver since Friday, all working fine, in fact this is 9.4 second edition because there is not much to see here, a bit of improvement here and there and a few megabytes up in size. That tells they are not removing legacy code yet.

<RANT>
Why disable legacy code if you have not removed it, for two drivers releases yet, as far as I can tell people with older cards could still have used 9.5 if only the legacy code was enabled. You are keeping it there, so why not enable it, until you remove it completely?

Still the same old, we support 3 outdated distributions but we don't support Linux. Well, that is why NVIDIA owns this market because they support Linux. I mean, no support for kernel 2.6.29 yet? And kernel 2.6.30 is just a few weeks away?

These guys are still lost, a half-baked open source driver, with a lot and I mean a lot of lack on 3D features. And a badly supported proprietary driver, which can;t be even updated for newer kernels when the need be.

OMG, I am so desperate to buy a new laptop, and never ever commit the same mistake of buying anything with ATI again. I love AMD but if the laptops come in this bad combo I can't buy.
</RANT>

For what it's worth, I'm using your Catalyst 9.5 pkgbuild from AUR, and it works for me too.

[Kano]

@he_the_great

Does working great mean you like the hundrets of

[30761.185566] [fglrx:firegl_find_any_map] *ERROR* Invalid map handle!<3>[fglrx:drm_vm_close] *ERROR* map not found -> inconsistent kernel data!!! vma_start:0xb7de2000,handle:0xd0100000

errors when you run dmesg?

[RagingDragon]

While I'm not yet using 9.5, but using the packaged 9.4 drivers in Debian Sid, it has been working well for the 2.6.29-AMD64 kernel. I can't speak for other configurations.

There are unofficial patches in the wild - I have Catalyst 9.5 running on 2.6.29 as well (Arch 64). The issue is that these patches are a pain for package maintainers, and that it would make more sense for a single person/group to maintain the patches (i.e. AMD) rather than many people/groups (i.e. each distribution) maintaining them separately.

[he_the_great]

@he_the_great

Does working great mean you like the hundrets of

[30761.185566] [fglrx:firegl_find_any_map] *ERROR* Invalid map handle!<3>[fglrx:drm_vm_close] *ERROR* map not found -> inconsistent kernel data!!! vma_start:0xb7de2000,handle:0xd0100000

errors when you run dmesg?

I don't run dmesg, so yes.

[Kano]

Then run it...

[Rip-Rip]

I mean, no support for kernel 2.6.29 yet? And kernel 2.6.30 is just a few weeks away?

Maybe AMD will forget 2.6.29, and jump directly to 2.6.30, who kwows.

[Kano]

The problem with ATI is: they work too slow. Currently you need a KERNEL patch that exports one symbol to get it running together with other adjustments. So if they really need that symbol they should test it NOW and not when the rc phase is over.