[mythtv-users] Problems with the VIA KM400 chipset

[Jarod Wilson]

On Jun 5, 2004, at 11:25, Tim Tait wrote:

> Jarod Wilson wrote:
>
>> On Jun 4, 2004, at 20:11, BP wrote:
>>
>>> Andy Long wrote:
>>>
>>>> Does anyone know if there are problems using the PVR-250 on a board 
>>>>  with the
>>>> VIA KM400 chipset?  I'm looking at getting the Abit VA-10
>>>> (http://www.newegg.com/app/viewProductDesc.asp?description=13-127 
>>>> -159&catalo
>>>> g=23&manufactory=BROWSE), but have heard that there are problems 
>>>> with  the
>>>> ivtv driver and via chipsets.  Is that just limited the the  
>>>> KT133/KT266, or
>>>> is it all Via chips?
>>>> Thanks.
>>>
>>>
>>> I've not had a single problem since I switched to my Abit VA-10 
>>> about  six months ago.  It's been running 24x7 with an occasional 
>>> reboot for  kernel updates.  I'm using the ivtv 0.1.9 driver patched 
>>> for 2.6 and  the 2.6.4 kernel.  My previous board (Chaintech 7NIF2 
>>> (NForce2)) was  nothing but trouble and acted like the reported VIA 
>>> 266 ivtv bug.
>>
>>
>> My MSI KM4M-L (KM400-based micro ATX) board has no problems at all. 
>> But  then neither does my KM266 board, and my Chaintech 7NIF2 has 
>> been  flawless (its been my master backend with two cards in it, 
>> rock-solid  for at least 8mo)...
>
> There were issues on many AMD mobo's with the Athlon "CPU Disconnect" 
> feature, particularly nForce2's. It has something to do with extra 
> delay and a race condidition if "CPU disconnect" is enabled. It caused 
> my FIC AU13 (who's BIOS does not loet you turn  that feature off) to 
> lock up during long builds. Turning off HD DMA reduced the frequency 
> but did not eliminate it. Even hitting "Caps lock" on/off during the 
> build could trigger it. The fix was simple - turn off "CPU Disconnect" 
> in the BIOS if possible, or via "athcool" if not. The downside was a 
> slightly hotter cpu. Turning of the APIC also works, if you habe 
> enough interrupts available. There were some kernel hacks to fix it 
> through long delays before reading interrupts (ick). Whatever the 
> cause/fix, I don't seem to have a problem anymore with the 2.4.26 
> kernel, though it may be too soon to be sure.

I should add that I run Red Hat distros on all these boards, which have 
APIC disabled, thus nary a problem with my nForce2 board. Not sure what 
impact has had on the stability of both my KM400 and KM266 boards 
versus other distros.

> Some or all of this may apply to Via chipsets. I also have an MSI KM2M 
> Combo-L running a samba server for a year w/o problems - but it 
> defaulted to "CPU Disconnect" off in the BIOS, and when I turn it on 
> things go to hell - though mostly it results in really lousy PCI IDE 
> controller performance, not lockups. I have not tried having it on for 
> 2.4.26 kernel.

That's the same KM266 board I have. Until today, it has been my test 
machine. Its getting replaced by the KM400 system, because the KM400 
system is getting replaced as my secondary Myth box by an EPIA M10000 
in a Travla C137 today. I'm gunning for diskless boot, a PVR-350 for 
output, and total silence...

-- 
Jarod C. Wilson, RHCE
jcw at wilsonet.com

Got a question? Read this first...
     http://catb.org/~esr/faqs/smart-questions.html
MythTV, Fedora Core & ATrpms documentation:
     http://wilsonet.com/mythtv/
MythTV Searchable Mailing List Archive
     http://www.gossamer-threads.com/lists/mythtv/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: PGP.sig
Type: application/pgp-signature
Size: 186 bytes
Desc: This is a digitally signed message part
Url : http://mythtv.org/pipermail/mythtv-users/attachments/20040605/57a85f4b/PGP-0001.pgp