openvortex-dev
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Openvortex-dev] VIA fix?


From: Ryan Underwood
Subject: Re: [Openvortex-dev] VIA fix?
Date: Fri, 8 Aug 2003 15:42:09 -0500
User-agent: Mutt/1.5.4i

Hi,

On Fri, Aug 08, 2003 at 08:37:43AM -0500, Seg wrote:
> 
> setpci -d '12eb:*' 40.B=ff
> 
> And now it works fine. I know at least one other person here uses the
> fix, do you get similar problems without it, playback for a second or
> two then lockup? Does anyone have any idea what this actually does?

PCI latency -> 0xff means that the card is allowed to hog the bus for
potentially the entire bus cycle when it is a master.  If there is some
other silly card that is constantly interrupting the vortex's master
cycle, it may not have enough time to get any data transferred.

> Does it cause problems if done on non-VIA chipsets? Apparently this tweak was
> put in the later Windows drivers, those who have disassembled them, do
> you have any idea if it is just enabled no matter what or do they check
> for a VIA chipset?

It won't cause problems except that a card with a 255 latency may hog
the bus forever, starving everything else of bus time.  Normally with
the Aureal cards this doesn't really happen though, the card wil finish
its business on the bus and hop off.

Some video cards used to get set to very high latency by the windows
drivers as a benchmark cheat, since it increases the responsiveness of
the video, but with the drawback of your sound clicking and modem
hanging up, since they got little to no bus time. :)  A video card tends
to be much more busy than the vortex card though, so increasing the
latency timer can really starve the rest of the pci cards.
> 
> This really ought to be in the driver, maybe as a manual module option
> to begin with. Is anyone using a (newer) VIA chipset that *doesn't* need
> this fix? It'd be useful to know if newer VIA chipsets ever 'fixed'
> whatever is going on here.

It used to be in a patch to the sourceforge driver.  There was also
another fix for the VIA northbridge/VIA southbridge combo but the page
has disappeared, from the google cache:

This fix works by adjusting the bit called ?Extend PCI#2 Internal Master
for Efficient Handling of Dummy Requests? for the register ?PCI #2
Master Control? of the AGP controller in VIA north-bridges.  This should
be enabled for the best Vortex 2 compatibility on VIA motherboards.  The
default setting in the chip itself if Disabled, and the BIOS must set
this to 1, as per VIA?s recommendations.  Unfortunately, some
manufacturers don?t and hence you have to set in the follwing way.

<snip generic WCPREDIT instructions>

Type this in: Bus 0, Device 1, Function 0.  And, for Register put in 42
and in Data add 10 to the current value.

The link is here but it can also be dug up from the google cache:
http://myweb.tiscali.co.uk/chris.day/vortexfix00.html

setpci should also do that, just `setpci -d '12eb:*' 42.B` to get the
value, and `setpci -d '12eb:*' 42.B new_value` to set the value,
where new_value is the value that was read + 10.

HTH

-- 
Ryan Underwood, <nemesis at icequake.net>, icq=10317253




reply via email to

[Prev in Thread] Current Thread [Next in Thread]