2.6.15-rc1 - NForce4 PCI-E agpgart support?

2.6.15-rc1 - NForce4 PCI-E agpgart support?

[Mark Knecht]

Hi,
   I downloaded and built 2.6.15-rc1 as a test assuming Ingo will
release -rt support for this one of these days. (No rush Ingo!) It
booted on my AMD64 machine and is running fine AFAICT.

   One thing I was expecting to see was agpgart support for the
NForce4 chipset. Is this something that's coming or am I missing where
the configuration is done?

   I have a PCI-Express based Radeon and would like to get better
performance. I'm presuming that agpgart support is part of that
solution? (As it was on earlier architectures?)

Thanks,
Mark

Re: 2.6.15-rc1 - NForce4 PCI-E agpgart support?

[Arjan van de Ven]

On Wed, 2005-11-16 at 06:50 -0800, Mark Knecht wrote:
> Hi,
>    I downloaded and built 2.6.15-rc1 as a test assuming Ingo will
> release -rt support for this one of these days. (No rush Ingo!) It
> booted on my AMD64 machine and is running fine AFAICT.
> 
>    One thing I was expecting to see was agpgart support for the
> NForce4 chipset. Is this something that's coming or am I missing where
> the configuration is done?
> 
>    I have a PCI-Express based Radeon and would like to get better
> performance. I'm presuming that agpgart support is part of that
> solution? (As it was on earlier architectures?)

I'm pretty sure PCI-Express and AGP are mutually exclusive....

Re: 2.6.15-rc1 - NForce4 PCI-E agpgart support?

[Mark Knecht]

On 11/16/05, Arjan van de Ven <arjan@infradead.org> wrote:
> On Wed, 2005-11-16 at 06:50 -0800, Mark Knecht wrote:
> > Hi,
> >    I downloaded and built 2.6.15-rc1 as a test assuming Ingo will
> > release -rt support for this one of these days. (No rush Ingo!) It
> > booted on my AMD64 machine and is running fine AFAICT.
> >
> >    One thing I was expecting to see was agpgart support for the
> > NForce4 chipset. Is this something that's coming or am I missing where
> > the configuration is done?
> >
> >    I have a PCI-Express based Radeon and would like to get better
> > performance. I'm presuming that agpgart support is part of that
> > solution? (As it was on earlier architectures?)
>
> I'm pretty sure PCI-Express and AGP are mutually exclusive....

Ah, of course! My bad... They are different buses and connectors. I
was really thinking more of the 'gart' part of the agpgart.

Is there any requirement/need/value for something like a PCI-E-gart?
Or does this relocation requirement go out the window somehow when a
graphics device moves to PCI-Express?

Thanks,
Mark

Re: 2.6.15-rc1 - NForce4 PCI-E agpgart support?

[Alistair John Strachan]

On Wednesday 16 November 2005 15:09, Mark Knecht wrote:
> On 11/16/05, Arjan van de Ven <arjan@infradead.org> wrote:
> > On Wed, 2005-11-16 at 06:50 -0800, Mark Knecht wrote:
> > > Hi,
> > >    I downloaded and built 2.6.15-rc1 as a test assuming Ingo will
> > > release -rt support for this one of these days. (No rush Ingo!) It
> > > booted on my AMD64 machine and is running fine AFAICT.
> > >
> > >    One thing I was expecting to see was agpgart support for the
> > > NForce4 chipset. Is this something that's coming or am I missing where
> > > the configuration is done?
> > >
> > >    I have a PCI-Express based Radeon and would like to get better
> > > performance. I'm presuming that agpgart support is part of that
> > > solution? (As it was on earlier architectures?)
> >
> > I'm pretty sure PCI-Express and AGP are mutually exclusive....
>
> Ah, of course! My bad... They are different buses and connectors. I
> was really thinking more of the 'gart' part of the agpgart.
>
> Is there any requirement/need/value for something like a PCI-E-gart?
> Or does this relocation requirement go out the window somehow when a
> graphics device moves to PCI-Express?

Yes, you don't need it with PCIe.

-- 
Cheers,
Alistair.

'No sense being pessimistic, it probably wouldn't work anyway.'
Third year Computer Science undergraduate.
1F2 55 South Clerk Street, Edinburgh, UK.

Re: 2.6.15-rc1 - NForce4 PCI-E agpgart support?

[Mark Knecht]

On 11/16/05, Alistair John Strachan <s0348365@sms.ed.ac.uk> wrote:
> On Wednesday 16 November 2005 15:09, Mark Knecht wrote:
> > On 11/16/05, Arjan van de Ven <arjan@infradead.org> wrote:
> > > On Wed, 2005-11-16 at 06:50 -0800, Mark Knecht wrote:
> > > > Hi,
> > > >    I downloaded and built 2.6.15-rc1 as a test assuming Ingo will
> > > > release -rt support for this one of these days. (No rush Ingo!) It
> > > > booted on my AMD64 machine and is running fine AFAICT.
> > > >
> > > >    One thing I was expecting to see was agpgart support for the
> > > > NForce4 chipset. Is this something that's coming or am I missing where
> > > > the configuration is done?
> > > >
> > > >    I have a PCI-Express based Radeon and would like to get better
> > > > performance. I'm presuming that agpgart support is part of that
> > > > solution? (As it was on earlier architectures?)
> > >
> > > I'm pretty sure PCI-Express and AGP are mutually exclusive....
> >
> > Ah, of course! My bad... They are different buses and connectors. I
> > was really thinking more of the 'gart' part of the agpgart.
> >
> > Is there any requirement/need/value for something like a PCI-E-gart?
> > Or does this relocation requirement go out the window somehow when a
> > graphics device moves to PCI-Express?
>
> Yes, you don't need it with PCIe.
>
> --
> Cheers,
> Alistair.

Thanks Alistair.

So, should I be able to see better grapohics performance on my Radeon
PCI-E device with 2.6.15-rc1? Are there setups I should test for you
guys? (I'm not a developer.)

Thanks,
Mark

Re: 2.6.15-rc1 - NForce4 PCI-E agpgart support?

[Alistair John Strachan]

On Wednesday 16 November 2005 18:25, Mark Knecht wrote:
> On 11/16/05, Alistair John Strachan <s0348365@sms.ed.ac.uk> wrote:
> > On Wednesday 16 November 2005 15:09, Mark Knecht wrote:
[snip]
> > >
> > > Is there any requirement/need/value for something like a PCI-E-gart?
> > > Or does this relocation requirement go out the window somehow when a
> > > graphics device moves to PCI-Express?
> >
> > Yes, you don't need it with PCIe.
> >
> > --
> > Cheers,
> > Alistair.
>
> Thanks Alistair.
>
> So, should I be able to see better grapohics performance on my Radeon
> PCI-E device with 2.6.15-rc1? Are there setups I should test for you
> guys? (I'm not a developer.)

I think the latest drm tree (which might be part of -rc1, I haven't checked 
the changelogs) includes support for several PCIe radeons. Your best bet 
would probably be to compile in DRM (kernel side), check dmesg that it's 
detected your card, then download the latest snapshot of X11R6 6.9/7.0 and 
build it.

The alternative is ATI's proprietary driver which probably already supports 
your card.

-- 
Cheers,
Alistair.

'No sense being pessimistic, it probably wouldn't work anyway.'
Third year Computer Science undergraduate.
1F2 55 South Clerk Street, Edinburgh, UK.

Re: 2.6.15-rc1 - NForce4 PCI-E agpgart support?

[Mark Knecht]

On 11/16/05, Alistair John Strachan <s0348365@sms.ed.ac.uk> wrote:
> On Wednesday 16 November 2005 18:25, Mark Knecht wrote:
> > On 11/16/05, Alistair John Strachan <s0348365@sms.ed.ac.uk> wrote:
> > > On Wednesday 16 November 2005 15:09, Mark Knecht wrote:
> [snip]
> > > >
> > > > Is there any requirement/need/value for something like a PCI-E-gart?
> > > > Or does this relocation requirement go out the window somehow when a
> > > > graphics device moves to PCI-Express?
> > >
> > > Yes, you don't need it with PCIe.
> > >
> > > --
> > > Cheers,
> > > Alistair.
> >
> > Thanks Alistair.
> >
> > So, should I be able to see better grapohics performance on my Radeon
> > PCI-E device with 2.6.15-rc1? Are there setups I should test for you
> > guys? (I'm not a developer.)
>
> I think the latest drm tree (which might be part of -rc1, I haven't checked
> the changelogs) includes support for several PCIe radeons. Your best bet
> would probably be to compile in DRM (kernel side), check dmesg that it's
> detected your card, then download the latest snapshot of X11R6 6.9/7.0 and
> build it.
>
> The alternative is ATI's proprietary driver which probably already supports
> your card.

Thanks. I'll see if this old guitar player can get all of that done.

Cheers,
Mark

Re: 2.6.15-rc1 - NForce4 PCI-E agpgart support?

[Lee Revell]

On Wed, 2005-11-16 at 11:54 -0800, Mark Knecht wrote:
> > The alternative is ATI's proprietary driver which probably already supports
> > your card.
> 
> Thanks. I'll see if this old guitar player can get all of that done.

Mark,

You should really decide whether you're more interested in 100% xrun
free audio performance OR better video performance and pursue one or the
other - if you try to work on these in parallel you'll find it's one
step forward, two steps back.  There have been many cases in the past
where video drivers ended up doing evil things that would ruin reliable
audio performance to get 0.1% better numbers on some Windows benchmark,
then the same bad behavior got ported over to Linux.  I'd be especially
cautious with the Radeon driver as much of it seems to be reverse
engineered.  And if you read the "X spinning in the kernel" thread you
see that apparently these GPUs can "crash" (!) in which case you seem to
be screwed.

Lee

Re: 2.6.15-rc1 - NForce4 PCI-E agpgart support?

[Mark Knecht]

On 11/16/05, Lee Revell <rlrevell@joe-job.com> wrote:
> On Wed, 2005-11-16 at 11:54 -0800, Mark Knecht wrote:
> > > The alternative is ATI's proprietary driver which probably already supports
> > > your card.
> >
> > Thanks. I'll see if this old guitar player can get all of that done.
>
> Mark,
>
> You should really decide whether you're more interested in 100% xrun
> free audio performance OR better video performance and pursue one or the
> other - if you try to work on these in parallel you'll find it's one
> step forward, two steps back.  There have been many cases in the past
> where video drivers ended up doing evil things that would ruin reliable
> audio performance to get 0.1% better numbers on some Windows benchmark,
> then the same bad behavior got ported over to Linux.  I'd be especially
> cautious with the Radeon driver as much of it seems to be reverse
> engineered.  And if you read the "X spinning in the kernel" thread you
> see that apparently these GPUs can "crash" (!) in which case you seem to
> be screwed.
>
> Lee

Lee,
   I've been reading that thread with interest. Thanks.

   Also, at this point I have no reason to believe that I'm not 100%
xrun free with the right set of apps, meaning I cannot really say any
xrun I see today is kernel induced. (They may be...I just cannot say
that.) For instance, I've yet to see an xrun in days of using Ardour
and running Jack from a terminal. However running Jack from QJC and
using Aqualung results in one or two every few days. This is a low
level issue from my perspective.

   My interest is, in general, audio only. However, on the odd
afternoon I have no qualms about rebooting into a different kernel or
a different video driver that might support a bit of game playing.
That's all that's going on here. Someone recently said there might be
PCI Express support in 2.6.15 so I was just looking.

   That said, over the last couple of years I have done a couple of
jobs that are sound track related. (Pro Tools, Vegas, etc. under XP)
I'm keeping my eyes open for a tool set that might allow me a similar
opportunity one of these days under Linux. (No rush at all)

   Note: It may not be obvious, but I'm young(-ish), retired, and
pretty much get to do what I want now. The only thing that stops me
from doing this on two machines is space and noise concerns.

   Thanks for all your inputs. You've been a great help.

cheers,
Mark