CentOS - Oct 2005 - Opteron + Nvidia + Centos 4.2 i386 = Thud

Dear All,

Just returned from vacation and was delighted to see that Centos 4.2 is now on 
the mirrors. Unfortunately the new i386 version is giving us problems on our 
dual Opteron workstations with Nvidia graphics cards...

Basically, nine times out of ten, the boxes will hang when they try to re-load 
X...  and just sit there with the keyboards Caps and Scroll lock LED's 
flashing. The only way out is the reset button.

The working fixes I have so far found are to either use a non-smp kernel 
(2.6.9-22.EL) or to use the VESA X driver. Neither is a proper solution as 
the machines are used for 3D modeling.

Non working fix attempts have included:
passing acpi=off to the kernel
trying Nvidia driver versions x86-1.0-7676, x86-1.0-7667 and x86-1.0-6629
Installing Centos from scratch as well as upgrading Centos 4.1 to 4.2 (via 
up2date) 
Trying three different Nvidia cards (Geforce 6800U, 5900 and a MX200) 
Trying two different motherboards (Tyan thunder K8W and Tiger K8W - both are 
AMD 8000 chipset based) 
Trying two different pairs of Opteron 240's and Opteron 246's.

As things used to work more or less perfectly with i386 Centos 4.1 
(2.6.9-11.ELsmp and the Nvidia 7667 driver) I'm starting to feel a little 
stupid - any chance that someone could put me out of my misery and suggest 
some solutions? 

Many thanks

Jake
-- 
Dr J. Grimmett
Computer Systems Manager
Division of Molecular Structure
National Institute for Medical Research
The Ridgeway
Mill Hill
London, NW7 1AA
Tel: 0208 816 2028
Fax: 0208 906 4477

On Thu, 27 Oct 2005 at 3:02pm, Jake Grimmett wrote
> Basically, nine times out of ten, the boxes will hang when they try to
re-load
> X...  and just sit there with the keyboards Caps and Scroll lock LED's
> flashing. The only way out is the reset button.
>
> The working fixes I have so far found are to either use a non-smp kernel
> (2.6.9-22.EL) or to use the VESA X driver. Neither is a proper solution as
> the machines are used for 3D modeling.
Out of curiosity, would it be possible to try x86_64 4.2?  Our graphics 
workstation here (with similar hardware to yours) runs the x86_64 version 
(although we haven't moved to 4.2 yet) quite well.  And, Nvidia's x86_64
drivers contain 32bit compat libraries, so standard x86 apps get the 
hardware acceleration as do x86_64 ones.

-- 
Joshua Baker-LePain
Department of Biomedical Engineering
Duke University

On Thu, 2005-10-27 at 15:02 +0100, Jake Grimmett wrote:
> Dear All,
> 
> Just returned from vacation and was delighted to see that Centos 4.2 is now
on
> the mirrors. Unfortunately the new i386 version is giving us problems on
our
> dual Opteron workstations with Nvidia graphics cards...
> 
> Basically, nine times out of ten, the boxes will hang when they try to
re-load
> X...  and just sit there with the keyboards Caps and Scroll lock LED's 
> flashing. The only way out is the reset button.
> 
> The working fixes I have so far found are to either use a non-smp kernel 
> (2.6.9-22.EL) or to use the VESA X driver. Neither is a proper solution as 
> the machines are used for 3D modeling.
> 
> Non working fix attempts have included:
> passing acpi=off to the kernel
> trying Nvidia driver versions x86-1.0-7676, x86-1.0-7667 and x86-1.0-6629
> Installing Centos from scratch as well as upgrading Centos 4.1 to 4.2 (via 
> up2date) 
> Trying three different Nvidia cards (Geforce 6800U, 5900 and a MX200) 
> Trying two different motherboards (Tyan thunder K8W and Tiger K8W - both
are
> AMD 8000 chipset based) 
> Trying two different pairs of Opteron 240's and Opteron 246's.
> 
> As things used to work more or less perfectly with i386 Centos 4.1 
> (2.6.9-11.ELsmp and the Nvidia 7667 driver) I'm starting to feel a
little
> stupid - any chance that someone could put me out of my misery and suggest 
> some solutions? 
> 
Jake,

There is a new kernel that addresses this specific issue that has just
been released. (2.6.9-22.0.1.EL)

I am building it now, but the kernel updates take quite some time.
Should be done later to day though.


-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
URL:
<http://lists.centos.org/pipermail/centos/attachments/20051027/5ce91b97/attachment-0001.sig>

Hi Johnny,

Many thanks - your suggested fix of a new kernel works perfectly! My Opterons 
are happy once more :)

Also - thanks Joshua for your advice: 
I last tried x86_64 with Whitebox 3, and had many problems with VMD, 
InsightII, Xeasy and MolMol - I'll have a go at 64bit again soon - has your 
64bit experience been painless ? 

Jake

Problem was:
>Centos 4.2 i386 version caused smp Opteron workstations with Nvidia graphics
> cards to, nine times out of ten, hang when they try to re-load 
> X...  and just sitting  there with the Caps and Scroll lock LED's 
> flashing. only way out is the reset button.
> Jake,
>
> There is a new kernel that addresses this specific issue that has just
> been released. (2.6.9-22.0.1.EL)
>
> I am building it now, but the kernel updates take quite some time.
> Should be done later to day though.
>
>
-- 
Dr J. Grimmett
Computer Systems Manager
Division of Molecular Structure
National Institute for Medical Research
The Ridgeway
Mill Hill
London, NW7 1AA
Tel: 0208 816 2028
Fax: 0208 906 4477