Origin 200 + Graphics

21 views
Skip to first unread message

SkyWriter

unread,
Oct 10, 2003, 2:29:08 PM10/10/03
to
What are the full range of Graphics options that can go into
an O200? I know an SI can, but SSE (double high) can't. how
about Vpro? or personal Video?

Greg Douglas

unread,
Oct 13, 2003, 1:06:13 AM10/13/03
to

SkyWriter wrote:
> What are the full range of Graphics options that can go into
> an O200? I know an SI can, but SSE (double high) can't.

The Origin 200 Gigachannel Expander (Xtown) has 5 single width
XIO slots. It can take SI or SE, but not the dual Raster Engine
boards (SSI, SSE).

> how
> about Vpro?

I seem to recall that Vpro is a dual height (takes two adjacent
slots), like A-D, or B-C in an Octane. I could be wrong on this,
but I know VPro has an oddbal latching mechanism.

> or personal Video?

With that wingo board (dual height adjacent, with the Kapton cables
going between the two), I don't think so, because all the slots on
the Giga Xpander face the same way, and Personal Video requires
adjacent (back to front) slots.

--
Greg Douglas
Reputable Systems
http://www.reputable.com

M.C.D. Roos

unread,
Oct 13, 2003, 6:45:31 AM10/13/03
to
>> how
>> about Vpro?
>
>
> I seem to recall that Vpro is a dual height (takes two adjacent
> slots), like A-D, or B-C in an Octane. I could be wrong on this,
> but I know VPro has an oddbal latching mechanism.

Vpro graphics (or at least the V10 version) uses 1.5 slots, but of
course you can't do anything with the remaining half free xio slot.
There also is a different xio carrier for vpro thus.

greetings,
Michiel

SkyWriter

unread,
Oct 13, 2003, 8:06:42 AM10/13/03
to
"M.C.D. Roos" wrote:

thanks guys! i'm gonna be happy with this beast either way :)


Ian Mapleson

unread,
Oct 13, 2003, 9:59:49 AM10/13/03
to
Greg Douglas <gdou...@reputable.com> wrote:
> The Origin 200 Gigachannel Expander (Xtown) has 5 single width
> XIO slots. It can take SI or SE, but not the dual Raster Engine
> boards (SSI, SSE).

Could it thus take SI+Texture or SE+Texture?

Ian.

SGI Depot: http://www.futuretech.blinkenlights.nl/advert/ian.html
Email: mape...@yahoo.com
Home: +44 (0)131 477 1142
Mobile: 07743 495403 (usually off; leave a message and I'll call back)

SGI/Future Technology/N64: http://www.futuretech.blinkenlights.nl/
Doom Help Service (DHS): http://www.gamers.org/dhs/
BSc Dissertation (Doom): http://www.gamers.org/dhs/diss/

Wolfgang Szoecs

unread,
Oct 13, 2003, 12:16:02 PM10/13/03
to
In article <3F8A32C5...@reputable.com>,

Greg Douglas <gdou...@reputable.com> writes:
>
>
> SkyWriter wrote:
>> What are the full range of Graphics options that can go into
>> an O200? I know an SI can, but SSE (double high) can't.
>
> The Origin 200 Gigachannel Expander (Xtown) has 5 single width
> XIO slots. It can take SI or SE, but not the dual Raster Engine
> boards (SSI, SSE).

well - it clearly shows, that none of you has ever tried this out :-)
(I just ran about this two weeks ago - but now I know it for sure)

the O200's PROM is a stripped down PROM ! (as compared to the O2000) !
In other words: gfx-support is missing ! (due to space issues)
that's also the reason, why you can't have an O200 based Onyx2.

But this also means, that you cannot ever have an MGRAS based O200 !
(while O2000 should work !)

So whatever SGI-specific gfx-boards you plug into an O200
I seriously doubt that it will work ...
AFAIK the only exception is the PSItech PCI board !

...sorry to steal all your illusions.

Wolfgang

Tony 'Nicoya' Mantler

unread,
Oct 13, 2003, 1:04:12 PM10/13/03
to
In article <bmej42$4kn$1...@bongo.munich.sgi.com>,
wo...@engr.sgi.com (Wolfgang Szoecs) wrote:

[...]
: well - it clearly shows, that none of you has ever tried this out :-)


: (I just ran about this two weeks ago - but now I know it for sure)
:
: the O200's PROM is a stripped down PROM ! (as compared to the O2000) !
: In other words: gfx-support is missing ! (due to space issues)
: that's also the reason, why you can't have an O200 based Onyx2.
:
: But this also means, that you cannot ever have an MGRAS based O200 !
: (while O2000 should work !)
:
: So whatever SGI-specific gfx-boards you plug into an O200
: I seriously doubt that it will work ...
: AFAIK the only exception is the PSItech PCI board !
:
: ...sorry to steal all your illusions.

Speaking of strange video configurations, it occurs to me that SGI now has a
very workable PCI(-X) graphics system, and a lot of older machines with PCI
slots... Any chance of the two of these coming together? (or even a shoehorn
version?)

Performance would obviously suffer a bit in the slower slots, but it'd be nice
to have a simple way to add another 3-6 million pixels to an Octane or O2.


Cheers - Tony 'Nicoya' Mantler :)

--
Tony 'Nicoya' Mantler -- Master of Code-fu -- nic...@ubb.ca
-- http://nicoya.feline.pp.se/ -- http://www.ubb.ca/ --

Wolfgang Szoecs

unread,
Oct 13, 2003, 2:51:03 PM10/13/03
to
In article <nicoya-DDAAC4....@shawnews.wp.shawcable.net>,

Tony 'Nicoya' Mantler <nic...@ubb.ca> writes:
> In article <bmej42$4kn$1...@bongo.munich.sgi.com>,
> wo...@engr.sgi.com (Wolfgang Szoecs) wrote:
>
> [...]
>: well - it clearly shows, that none of you has ever tried this out :-)
>: (I just ran about this two weeks ago - but now I know it for sure)
>:
>: the O200's PROM is a stripped down PROM ! (as compared to the O2000) !
>: In other words: gfx-support is missing ! (due to space issues)
>: that's also the reason, why you can't have an O200 based Onyx2.
>:
>: But this also means, that you cannot ever have an MGRAS based O200 !
>: (while O2000 should work !)
>:
>: So whatever SGI-specific gfx-boards you plug into an O200
>: I seriously doubt that it will work ...
>: AFAIK the only exception is the PSItech PCI board !
>:
>: ...sorry to steal all your illusions.
>
> Speaking of strange video configurations, it occurs to me that SGI now has a
> very workable PCI(-X) graphics system, and a lot of older machines with PCI
> slots... Any chance of the two of these coming together? (or even a shoehorn
> version?)

sounds easy - but unfortunately it isn't.
No - this won't come together. ;-)

> Performance would obviously suffer a bit in the slower slots, but it'd be nice
> to have a simple way to add another 3-6 million pixels to an Octane or O2.

... it's simply not that easy or simple.
XFree86 and Xsgi are mutually quite exclusive.
(Also the O2 has a completely different pci implementation.)

XFree86 would probably compile/work without hw-OpenGL on those machines.
(at least as long as the drivers are endian-clean etc.)
you could use the pciba based hw-access - and that's what we used
initially to do some bring-up work etc.

OTOH - direct-rendering HW-accellerated OpenGL requires nearly always
a kernel driver part. --- and that's then the heavy part...
Nearly all XFree86 drivers are extremely x86 and/or Linux based...
Porting those to IRIX is pretty hard work... - it's not just a recompile.
(as would be to port an SGI-gfx driver to XFree86)

And you cannot have more than one PCI-gfx-board per PCI bus.
[ the legacy-memory and legacy-IO spaces will overlap. ]
(this is a design limitation of PC-gfx-boards and the PC-architecture)
So an Octane could only take one more board ... --- not worth the effort.

Wolfgang

Tony 'Nicoya' Mantler

unread,
Oct 13, 2003, 6:13:58 PM10/13/03
to
In article <bmes6n$e3e$1...@bongo.munich.sgi.com>,
wo...@engr.sgi.com (Wolfgang Szoecs) wrote:

: In article <nicoya-DDAAC4....@shawnews.wp.shawcable.net>,


: Tony 'Nicoya' Mantler <nic...@ubb.ca> writes:

[...]
: > Speaking of strange video configurations, it occurs to me that SGI now has a

: > very workable PCI(-X) graphics system, and a lot of older machines with PCI
: > slots... Any chance of the two of these coming together? (or even a shoehorn
: > version?)
:
: sounds easy - but unfortunately it isn't.
: No - this won't come together. ;-)

Since when is programming supposed to be easy? ;)


: > Performance would obviously suffer a bit in the slower slots, but it'd be nice

: > to have a simple way to add another 3-6 million pixels to an Octane or O2.
:
: ... it's simply not that easy or simple.
: XFree86 and Xsgi are mutually quite exclusive.
: (Also the O2 has a completely different pci implementation.)

How much difference is there with the O2, from a driver perspective? From what
I've read in the documentation, the O2 should be a simple recompile from code
that works on origin-style machines. In fact, it should be even more forgiving
(no limits on 32-bit DMA maps, DMA mapping funcs are no-ops, etc).


: XFree86 would probably compile/work without hw-OpenGL on those machines.


: (at least as long as the drivers are endian-clean etc.)
: you could use the pciba based hw-access - and that's what we used
: initially to do some bring-up work etc.

I'm pretty sure that the Radeon (R100) driver in particular has gotten a decent
workout in linux/ppc, so it should be relatively endian-safe and legacy-free.


: OTOH - direct-rendering HW-accellerated OpenGL requires nearly always

: a kernel driver part. --- and that's then the heavy part...
: Nearly all XFree86 drivers are extremely x86 and/or Linux based...
: Porting those to IRIX is pretty hard work... - it's not just a recompile.
: (as would be to port an SGI-gfx driver to XFree86)

Well, XFree (with the exception of the closed drivers) isn't exactly an icon of
3d performance, so if someone other than SGI were to do the port, I doubt they'd
cry about not getting accelerated 3d.


: And you cannot have more than one PCI-gfx-board per PCI bus.


: [ the legacy-memory and legacy-IO spaces will overlap. ]
: (this is a design limitation of PC-gfx-boards and the PC-architecture)
: So an Octane could only take one more board ... --- not worth the effort.

It should be possible to disable the VGA spaces. They shouldn't be needed for
anything other than VESA crap and VGA text mode (?), neither of which would be
used in the context of an SGI.

I'm pretty sure I've heard of PC users with multiple PCI vid cards.

SkyWriter

unread,
Oct 13, 2003, 6:21:09 PM10/13/03
to
Wolfgang Szoecs wrote:

> sos


> ...sorry to steal all your illusions.
>
> Wolfgang

sorry, maybe i missed something? who cares about prom, really now..
it's nice to see the console geegaws, but not necessary.

the question really is: does irix kernel negate the configuration?

Greg Douglas

unread,
Oct 14, 2003, 2:08:11 AM10/14/03
to

Wolfgang Szoecs wrote:
> well - it clearly shows, that none of you has ever tried this out :-)
> (I just ran about this two weeks ago - but now I know it for sure)

Wolfgang,
I just tried it about 2 days ago. SI and SE gfx works fine in the
Origin 200 Gigachannel Xpansion box. Honest. Interestingly enough,
the welcome screen says... "Welcome to " and that's it.

>
> the O200's PROM is a stripped down PROM ! (as compared to the O2000) !
> In other words: gfx-support is missing !

> But this also means, that you cannot ever have an MGRAS based O200 !

> So whatever SGI-specific gfx-boards you plug into an O200
> I seriously doubt that it will work ...

> ...sorry to steal all your illusions.

Doubt and reality are two very different things.

It's not an illusion.

IRIS 1% uname -R
6.5 6.5.21m
IRIS 2% hinv
1 225 MHZ IP27 ProcessorCPU: MIPS R10000 Processor Chip Revision: 3.4
FPU: MIPS R10010 Floating Point Chip Revision: 3.4
Main memory size: 256 Mbytes
Instruction cache size: 32 Kbytes
Data cache size: 32 Kbytes
Secondary unified instruction/data cache size: 2 Mbytes
Integral SCSI controller 0: Version QL1040B (rev. 2), single ended
Disk drive: unit 1 on SCSI controller 0
Integral SCSI controller 1: Version QL1040B (rev. 2), single ended
CDROM: unit 6 on SCSI controller 1
IOC3/IOC4 serial port: tty1
IOC3/IOC4 serial port: tty2
IOC3 parallel port: plp1
Graphics board: SI
Integral Fast Ethernet: ef0, version 1, module 1, slot io1, pci 2
Origin PCI XIO board, module 1 slot 7: Revision 4
Origin 200 base I/O, module 1 slot 1
IOC3/IOC4 external interrupts: 1
IRIS 3% uptime
9:26am up 7 mins, 3 users, load average: 1.06, 0.77, 0.38
IRIS 4%

Now, granted, this may not be an SGI SUPPORTED configuration, but I
assure you it works.

Wolfgang Szoecs

unread,
Oct 14, 2003, 7:23:41 AM10/14/03
to
In article <nicoya-D7E567....@news.sc.shawcable.net>,

Tony 'Nicoya' Mantler <nic...@ubb.ca> writes:
> In article <bmes6n$e3e$1...@bongo.munich.sgi.com>,
> wo...@engr.sgi.com (Wolfgang Szoecs) wrote:
>
>: In article <nicoya-DDAAC4....@shawnews.wp.shawcable.net>,
>: Tony 'Nicoya' Mantler <nic...@ubb.ca> writes:

[...]

>: ... it's simply not that easy or simple.


>: XFree86 and Xsgi are mutually quite exclusive.
>: (Also the O2 has a completely different pci implementation.)
>
> How much difference is there with the O2, from a driver perspective? From what
> I've read in the documentation, the O2 should be a simple recompile from code
> that works on origin-style machines. In fact, it should be even more forgiving
> (no limits on 32-bit DMA maps, DMA mapping funcs are no-ops, etc).

theoretically right - OTOH - the PCI implementation still has some
limitations, which won't make it funny.
And then the business sense - there's no rational justification for that.
OTOH - if someone wants to do it - the necessary XF86 source-bits are there.

>: XFree86 would probably compile/work without hw-OpenGL on those machines.
>: (at least as long as the drivers are endian-clean etc.)
>: you could use the pciba based hw-access - and that's what we used
>: initially to do some bring-up work etc.
>
> I'm pretty sure that the Radeon (R100) driver in particular has gotten a decent
> workout in linux/ppc, so it should be relatively endian-safe and legacy-free.

legacy-free - for sure not.
Most PC-video-boards still require x86 BIOS execution to initialize
the board into a known state.

[...]

>: And you cannot have more than one PCI-gfx-board per PCI bus.
>: [ the legacy-memory and legacy-IO spaces will overlap. ]
>: (this is a design limitation of PC-gfx-boards and the PC-architecture)
>: So an Octane could only take one more board ... --- not worth the effort.
>
> It should be possible to disable the VGA spaces. They shouldn't be needed for
> anything other than VESA crap and VGA text mode (?), neither of which would be
> used in the context of an SGI.

that's not 100% true.
For executing the BIOS - you need the VGA part to be on.

>
> I'm pretty sure I've heard of PC users with multiple PCI vid cards.
>

right. but those boards are exceptions !
Also - such setups are in most cases only for the SAME type of board.
(you cannot mix different brands...)

The main point here is, that most PC video boards REQUIRE that the
VGA-BIOS ran and initialized the card into a state from which XF86 etc.
can start doing setup work.
And as the legacy-mem and legacy-io addresses are fixed, there can't be
more than one per bus...

But I know - There are some boards (AFAIK Matrox is one) which are indeed
an exception from this - where the 3d-gfx-engine can be inited without
a prior BIOS run.
Multiple such boards won't cause conflicts if you turn off the
VGA part from all but the first one.

For boards which require a BIOS init - there is no real chance to have
several of them on one bus - as you would need to init each of them,
and this step causes conflicts.
You could of course do some extra-ugly hackery to avoid it,
but the remaining stuff (like XF86) would also need to support that...

Anyways - it's not really worth the effort.

Wolfgang

Wolfgang Szoecs

unread,
Oct 14, 2003, 7:26:32 AM10/14/03
to
In article <3F8B4FF3...@mrnutty.com>,

ok - agreed.

> the question really is: does irix kernel negate the configuration?

hmm ... good question ... - no the kernel does not negate it. it should work.

wolfgang

Wolfgang Szoecs

unread,
Oct 14, 2003, 7:29:00 AM10/14/03
to
In article <3F8B92C...@reputable.com>,
Greg Douglas <gdou...@reputable.com> writes:

> Doubt and reality are two very different things.

sometimes, yes :-))

> It's not an illusion.
>
> IRIS 1% uname -R
> 6.5 6.5.21m

[...]

> Now, granted, this may not be an SGI SUPPORTED configuration, but I
> assure you it works.

hmm --- indeed interesting...

wolfgang

Colin Anderson

unread,
Oct 14, 2003, 11:09:29 AM10/14/03
to
Greg Douglas <gdou...@reputable.com> wrote in message news:<3F8B92C...@reputable.com>...

> Wolfgang Szoecs wrote:
> > well - it clearly shows, that none of you has ever tried this out :-)
> > (I just ran about this two weeks ago - but now I know it for sure)
>
> Wolfgang,
> I just tried it about 2 days ago. SI and SE gfx works fine in the
> Origin 200 Gigachannel Xpansion box. Honest. Interestingly enough,

Neato! Have you tried multiple gfx cards?

Craylink a second O200 and have yourself a "quad processor Octane",
with internal CD-ROM no less!

Rainer M. Canavan

unread,
Oct 14, 2003, 12:00:53 PM10/14/03
to
Wolfgang Szoecs wrote:
> In article <nicoya-D7E567....@news.sc.shawcable.net>,
> Tony 'Nicoya' Mantler <nic...@ubb.ca> writes:
>
>>In article <bmes6n$e3e$1...@bongo.munich.sgi.com>,
>> wo...@engr.sgi.com (Wolfgang Szoecs) wrote:
>>
>>: In article <nicoya-DDAAC4....@shawnews.wp.shawcable.net>,
>>: Tony 'Nicoya' Mantler <nic...@ubb.ca> writes:
[...]
>
> right. but those boards are exceptions !
> Also - such setups are in most cases only for the SAME type of board.
> (you cannot mix different brands...)

Yes, you can. I've had a Matrox and an Nvidia card in a PC once, on the
one and only PCI bus in that box.

> The main point here is, that most PC video boards REQUIRE that the
> VGA-BIOS ran and initialized the card into a state from which XF86 etc.
> can start doing setup work.
> And as the legacy-mem and legacy-io addresses are fixed, there can't be
> more than one per bus...

Not more than one, but you can turn them on, run the bios (in an
emulator for non x86 boxen), and then off again one at a time. XFree86
does run the bios on server initialization for secondary and other
cards, while the primary one is disabled.

> But I know - There are some boards (AFAIK Matrox is one) which are indeed
> an exception from this - where the 3d-gfx-engine can be inited without
> a prior BIOS run.
> Multiple such boards won't cause conflicts if you turn off the
> VGA part from all but the first one.
>
> For boards which require a BIOS init - there is no real chance to have
> several of them on one bus - as you would need to init each of them,
> and this step causes conflicts.

You said it above yourself: turn on one at a time, let the bios have a
got at it... The only requirement is that at maximum one of the cards
should require legacy VGA crap to work after initialization.

> You could of course do some extra-ugly hackery to avoid it,
> but the remaining stuff (like XF86) would also need to support that...
>
> Anyways - it's not really worth the effort.

rainer

Wolfgang Szoecs

unread,
Oct 14, 2003, 3:20:31 PM10/14/03
to
In article <bmh6jn$m7id4$1...@id-204520.news.uni-berlin.de>,

"Rainer M. Canavan" <usenet2...@canavan.de> writes:

>> And as the legacy-mem and legacy-io addresses are fixed, there can't be
>> more than one per bus...
>
> Not more than one, but you can turn them on, run the bios (in an
> emulator for non x86 boxen), and then off again one at a time. XFree86
> does run the bios on server initialization for secondary and other
> cards, while the primary one is disabled.

yeah - in some simple setups this works, but as soon as you think
about running two independent Xservers etc. --- you need some
more synchronization/locking work... --- quite a nightmare.

>> But I know - There are some boards (AFAIK Matrox is one) which are indeed
>> an exception from this - where the 3d-gfx-engine can be inited without
>> a prior BIOS run.
>> Multiple such boards won't cause conflicts if you turn off the
>> VGA part from all but the first one.
>>
>> For boards which require a BIOS init - there is no real chance to have
>> several of them on one bus - as you would need to init each of them,
>> and this step causes conflicts.
>
> You said it above yourself: turn on one at a time, let the bios have a
> got at it... The only requirement is that at maximum one of the cards
> should require legacy VGA crap to work after initialization.

...but like written above - giving this guarantee of exclusivity
(only one board on the same bus at the same time has VGA turned on)
is not possible to guarantee without some kernel-support.
Still the above example:
Try to start two Xservers (one for each board) simultanously.
...or switch back to a VT in some unfortunate moment, etc ....

By having only one board per bus - you're avoiding all this easily.

Wolfgang

Wolfgang Szoecs

unread,
Oct 14, 2003, 3:34:15 PM10/14/03
to
In article <66ce502.03101...@posting.google.com>,

co...@beyondboxes.com (Colin Anderson) writes:
> Greg Douglas <gdou...@reputable.com> wrote in message news:<3F8B92C...@reputable.com>...
>> Wolfgang Szoecs wrote:
>> > well - it clearly shows, that none of you has ever tried this out :-)
>> > (I just ran about this two weeks ago - but now I know it for sure)
>>
>> Wolfgang,
>> I just tried it about 2 days ago. SI and SE gfx works fine in the
>> Origin 200 Gigachannel Xpansion box. Honest. Interestingly enough,
>
> Neato! Have you tried multiple gfx cards?

The mgras driver supports up to three cards per system in Octanes and IP27.
(and a max of two in IP22/IP28 --- but more aren't possible there anyways)

> Craylink a second O200 and have yourself a "quad processor Octane",
> with internal CD-ROM no less!

...sounds indeed funny.

wolfgang

Timo Kanera

unread,
Oct 14, 2003, 3:49:25 PM10/14/03
to
In article <66ce502.03101...@posting.google.com>,
co...@beyondboxes.com (Colin Anderson) writes:

> Neato! Have you tried multiple gfx cards?
>
> Craylink a second O200 and have yourself a "quad processor Octane",
> with internal CD-ROM no less!

yeah, I always thought Octane was way too silent anyway ;-)

so long,
Timo

--
Timo Kanera <ti...@kanera.de> . GPG Key-ID: 1024D/30CDB412

SkyWriter

unread,
Oct 14, 2003, 5:31:25 PM10/14/03
to
Wolfgang Szoecs wrote:

> In article <66ce502.03101...@posting.google.com>,
> co...@beyondboxes.com (Colin Anderson) writes:
> > Greg Douglas <gdou...@reputable.com> wrote in message news:<3F8B92C...@reputable.com>...
> >> Wolfgang Szoecs wrote:
> >> > well - it clearly shows, that none of you has ever tried this out :-)
> >> > (I just ran about this two weeks ago - but now I know it for sure)
> >>
> >> Wolfgang,
> >> I just tried it about 2 days ago. SI and SE gfx works fine in the
> >> Origin 200 Gigachannel Xpansion box. Honest. Interestingly enough,
> >
> > Neato! Have you tried multiple gfx cards?
>
> The mgras driver supports up to three cards per system in Octanes and IP27.
> (and a max of two in IP22/IP28 --- but more aren't possible there anyways)

three? cool, i haven't had three head machine since my old sparc 10.
i thought octane was limited to two.

Colin Anderson

unread,
Oct 15, 2003, 12:05:00 AM10/15/03
to
ti...@kanera.de (Timo Kanera) wrote in message news:<5j0u51-...@tankgirl.local.kanera.de>...

> In article <66ce502.03101...@posting.google.com>,
> co...@beyondboxes.com (Colin Anderson) writes:
>
> > Neato! Have you tried multiple gfx cards?
> >
> > Craylink a second O200 and have yourself a "quad processor Octane",
> > with internal CD-ROM no less!
>
> yeah, I always thought Octane was way too silent anyway ;-)

Hehe, who said it had to be practical? :) These sorts of crazy hacks
are just plain neat. If SGI had made a coal-powered, cast iron
workstation, I'd probably have one under my desk right now!

Greg Douglas

unread,
Oct 15, 2003, 3:34:12 AM10/15/03
to

Colin Anderson wrote:
> Hehe, who said it had to be practical? :) These sorts of crazy hacks
> are just plain neat. If SGI had made a coal-powered, cast iron
> workstation, I'd probably have one under my desk right now!

They're working on one now. It will be "powered" (and I use
the term loosely) by Linux, of course.

--
gd.

Rainer M. Canavan

unread,
Oct 15, 2003, 6:03:54 AM10/15/03
to
Wolfgang Szoecs wrote:
> In article <bmh6jn$m7id4$1...@id-204520.news.uni-berlin.de>,
> "Rainer M. Canavan" <usenet2...@canavan.de> writes:

[...]


> yeah - in some simple setups this works, but as soon as you think
> about running two independent Xservers etc. --- you need some
> more synchronization/locking work... --- quite a nightmare.


OK, multiseat makes things difficult, multihead with only one
XServer (which is what most people will want to have) and as many
boards as you can put in your box isn't that bad.

rainer

SkyWriter

unread,
Oct 15, 2003, 7:57:14 AM10/15/03
to
Greg Douglas wrote:

if you ask me, and I know nobody would, but just suppose it happened,
linux is more 'over powering', than 'powering'. anyway to get anything
useful out of say google i always put '-linux' for one of the criteria.


Wolfgang Szoecs

unread,
Oct 15, 2003, 8:36:58 AM10/15/03
to
In article <3F8C95F2...@mrnutty.com>,
SkyWriter <skyw...@mrnutty.com> writes:

>> The mgras driver supports up to three cards per system in Octanes and IP27.
>> (and a max of two in IP22/IP28 --- but more aren't possible there anyways)
>
> three? cool, i haven't had three head machine since my old sparc 10.
> i thought octane was limited to two.

I actually have a three-pipe Octane. OTOH - it's a pretty noisy setup.
So some days ago I moved over to a V12+DCD2.
(not the absolute perfect subsitute, but waaay more quiet.)

wolfgang

Wolfgang Szoecs

unread,
Oct 15, 2003, 8:34:12 AM10/15/03
to
In article <bmj62b$mjohd$1...@id-204520.news.uni-berlin.de>,

"Rainer M. Canavan" <usenet2...@canavan.de> writes:

> OK, multiseat makes things difficult, multihead with only one
> XServer (which is what most people will want to have) and as many
> boards as you can put in your box isn't that bad.

and if then the Xserver hangs in one of the board's bios-init,
and you want to restart it - or switch back to console - bang ...
and - there are some more scenarios like these.
The ugly point is, that such lock-ups result in panics.
IMO it's not justifiable for a commercially marketed and supported system.
(sure - for unsupported home-user setups it's probably acceptable)

Wolfgang

Colin Anderson

unread,
Oct 15, 2003, 2:55:17 PM10/15/03
to
wo...@engr.sgi.com (Wolfgang Szoecs) wrote in message news:<bmjf1a$1amd$2...@bongo.munich.sgi.com>...

I've noticed that Octane2 (well, blue Octane2 dual R12K/400 V12 to be
exact) is quite a bit more quiet than, say, a dual R12K/300 MXE. At
what point did the Octane series become more quiet? Are all Octane2s
that much quieter?

I ask because I'm considering an original blue stock dual R12K/400
V8... and am hoping its as quiet as the dual R12K/400 V12s I've used.

Tony 'Nicoya' Mantler

unread,
Oct 15, 2003, 3:23:09 PM10/15/03
to
In article <66ce502.03101...@posting.google.com>,
co...@beyondboxes.com (Colin Anderson) wrote:

: I've noticed that Octane2 (well, blue Octane2 dual R12K/400 V12 to be


: exact) is quite a bit more quiet than, say, a dual R12K/300 MXE. At
: what point did the Octane series become more quiet? Are all Octane2s
: that much quieter?
:
: I ask because I'm considering an original blue stock dual R12K/400
: V8... and am hoping its as quiet as the dual R12K/400 V12s I've used.

I've found that my Octane's noise level is extremely dependant on the ambient
temperature, and no doubt dependant on how many toasty cards are installed.

With my current config (dual 250, SE) and room temperature around 25c, I find
that the machine is basically as quiet or quieter than my Indigo2 (R10k
MaxIMPACT), though the noise level climbs quite quickly as the temperature
approaches 30c.

It seems that at the higher temperatures, the normal moving-air-whooshing sound
is drowned out by a much louder monotonic bearing whine from the powersupply
fan. I suspect that switching to a sleeve-bearing fan would make this much more
bearable. (Or I could just keep the room cooler)

The system is under my desk, with a pillow leaning against the wall behind it to
muffle the higher-frequency whooshing noises (don't worry, the system is far
enough from the wall that it doesn't interfere with airflow), and the noise
level is well within acceptable levels for my ears.

As a somewhat surprising bonus, the Octane also draws less power than the
Indigo2, despite having 2 CPUs and a power-hungry fire-belching reputation.

Wolfgang Szoecs

unread,
Oct 15, 2003, 4:17:16 PM10/15/03
to
In article <66ce502.03101...@posting.google.com>,
co...@beyondboxes.com (Colin Anderson) writes:

> I've noticed that Octane2 (well, blue Octane2 dual R12K/400 V12 to be
> exact) is quite a bit more quiet than, say, a dual R12K/300 MXE. At
> what point did the Octane series become more quiet? Are all Octane2s
> that much quieter?
>
> I ask because I'm considering an original blue stock dual R12K/400
> V8... and am hoping its as quiet as the dual R12K/400 V12s I've used.

it's simplier:
if there's only ONE XIO board - then the pwr-supply won't run at hi-fan,
while if there's more than one XIO board => full speed (and quite loud)
so even an older 'green' single-pipe Octane is pretty quiet
compared to a dual/tri-pipe. (or other XIO add ons like Mscsi or DIVO)

Wolfgang

Timo Kanera

unread,
Oct 15, 2003, 4:42:54 PM10/15/03
to
In article <nicoya-79F4CE....@news.sc.shawcable.net>,

Tony 'Nicoya' Mantler <nic...@ubb.ca> writes:

> As a somewhat surprising bonus, the Octane also draws less power than the
> Indigo2, despite having 2 CPUs and a power-hungry fire-belching reputation.

E-Series mgras graphics seem to be a bit less power-hungry than I-Series and
R12k draws less than r10k. I measured a constant 200 watts for my (single)
r12k-300, 1GB Octane with ESI - 260-280 watts after I swapped the ESI for
an EMXI. A dual r10k-175 machine with SSI draws more than 310 watts..

Alexis Cousein

unread,
Oct 16, 2003, 5:48:18 AM10/16/03
to
SkyWriter wrote:
> if you ask me, and I know nobody would, but just suppose it happened,
> linux is more 'over powering', than 'powering'. anyway to get anything
> useful out of say google i always put '-linux' for one of the criteria.
>
[Grin] Mhh -- Shall I mention Linux in my signature?

SkyWriter

unread,
Oct 16, 2003, 7:40:35 AM10/16/03
to
Alexis Cousein wrote:

that kind of question, answers itself. :)

Alexis Cousein

unread,
Oct 16, 2003, 9:43:57 AM10/16/03
to
SkyWriter wrote:

Yes indeed. It's not because I sometimes bitch about things like,
say, the way the buffer cache works on Linux, pthreads pre-next gen
Pthreads, or fight with its rather more cavaliere attitude towards
library versioning (which forces me to use a "modules" environment
to sort out the mess), that I don't enjoy working on the Altix series.

Not to mention that there are no IRIX laptops, and that Linux
is as close as I'll get to IRIX on that machine (especially with
5Dwm and the other IMD4Linux bits).

--
Alexis Cousein Senior Systems Engineer
ale...@sgi.com SGI/Silicon Graphics Brussels
<opinions expressed here are my own, not those of my employer>
<SGI sells IRIX *and* Linux systems -- I'm not flaming either>

Reply all
Reply to author
Forward
0 new messages