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

Re: 3d on alpha?

On Wed, Sep 04, 2002 at 11:17:30AM +0200, Idar Tollefsen wrote:
> > "pci_device_override" is necessary (and understood) ONLY on MIATA (PWS).
> I'll take you're word on the understood part, but the necessary part...
> The way pci_device_override is described seems to be just what I
> need to get the card working.

On MIATA, if you put a PCI card into one of the 64-bit slots that the
console does not recognize (it has a very small table), it will complain
and NOT LET YOU BOOT. It will, most likely, have already tried to run
the BIOS emulator, so that the complaint can be seen... ;-}

"pci_device_override" allow you to tell the console, "trust me", and it
will allow the unknown (to it) card to be in the 64-bit slot.

Such trust is not needed on LX164.

BUT, if the BIOS emulator actually crashes while attempting to run the
code on the video card you have installed, no amount of workarounds or
"trust mes" will make much difference... :-\

> The GVX1 is the same as the VX1 in that they both use the GLINT R3
> chip. In addition, the GVX1 has a GLINT Gamma 1 geometry processor
> which the VX1 is missing. They both have 32MB of RAM (different types
> tough, IIRC). Indeed, the two cards are so similar it's a wonder one
> works and the other one doesn't.

Ah, yes, I've seen a few of similar cards (Permedia-2 + Gamma), and
there have also been some problems with them.

Obviously the BIOS emulator has trouble running the BIOS code on the
GVX1 and not the VX1. With different chipsets on the different boards,
ie different things to get set up, it's no wonder (to me) that the
BIOSes are different...

> One interesting thing to note; I contacted 3DLabs about this, and they
> claim that the VX1 that Compaq supports for it's Alpha machines are
> based on the Permedia 3. That's the one and only communication I've been
> able to get from 3DLabs, which menas I've been unable to find out whether
> the Permedia 3 now has been renamed to GLINT R3, or if it at some point
> existed, or maybe still exists, another version of the VX1. Of course,
> a third option is simply that the 3DLabs representative was wrong.

AFAIK, GLINT R3 and Permedia-3 are the same. There is an AGP Permedia-3
which appears to identify itself as GLINT R4, but they are pretty much
the same, too.

> The symptoms when the card is installed are as you describe. There's
> about half a second of disk activity (I assume when the harddisk is
> spinning up), the keyboard light flashes once or twice and then
> the machine simply dies out. The screen isn't receiving any signals,
> the machine has no life, no activity, no nothing.

It seems likely that the BIOS emulator is crashing for some reason.

> I've tried different PCI slots, different monitors and tested the
> card in a x86 machine to verify that it works. The card works, but
> just not in the Alpha.

Well, x86 doesn't have a BIOS emulator to crash... ;-}

> I've heard claims that the SRM has hardcoded in device ID and vendor ID
> and will only boot on "approved" devices? 

Well, Tru64 and VMS are sold with a small selection of peripherals, and
they have made no attempt to support everything, so their console (SRM)
has done likewise.

> And that this was a ploy to make it impossible for "unauthorised"
> Alpha resellers to run Digital Unix, later True64?

Not necessarily; the set of peripherals, though small, is mostly
available retail, though there are notable exceptions.

> Indeed, isn't this what pci_device_override is suppose to
> circumvent?


MIATA was the first shipping box based on PYXIS core logic, which has
flaws. Those flaws are fixed by additional circuitry on later MIATA
mobos, and on other PYXIS platforms, like LX and SX.

Those flaws are exposed by certain driver practices, and can be
worked around by certain driver practices, or by placing the
PCI card in one of the 32-bit slots behind the PCI-PCI bridge.

SRM is only trying to be sure that a known PCI card with a driver
(under Tru64 or VMS) that knows what NOT TO DO is in a 64-bit slot.

> However, I've had no problem with the Diamond FireGL 1000 Pro
> that's in the machine now, and I somehow doubt that the FireGL actually is
> "authorised" to run Digital Unix or True64. Or maybe it's not the card
> itself, but the Permedia 2 chip that is "authorised"?

Yes, the table consists of vendor/device ID pairs, and Permedia-2 is
in there.


Jay A Estabrook                            HPTC - LINUX support
Hewlett-Packard Company - MRO1-2/K15       (508) 467-2080
200 Forest Street, Marlboro MA 01752       Jay.Estabrook@hp.com

Reply to: