Prism trouble

[lunatic]

Hi,

I just picked up a two-brick Prism and have some trouble getting it to work.
It's a standard configuration where one brick has an I/O board and the other brick has two FireGL cards.

There are several problems:
First, I can't get an answer from the main bricks serial port. I do get to the L1 on the brick with the graphics cards. config shows both bricks:

first, when it boots up it says:

Code: Select all

ALERT: VRM not OK: type CPU C 12V, number 2


however it seems to work.

001c04-L1>power vrm
VRM Type Location Present Okay
-------------- -------- ------- -------
CPU A 12V 1 passed passed
CPU C 12V 2 passed FAILED
NODE 2.5V on board passed passed
NODE CPU A on board passed passed
NODE CPU C on board passed passed

when it boots up it disables 4GB of memory:

Code: Select all

EFI Boot Manager ver 1.10 [14.62]

Partition  0:                           Enabled Disabled
    CBricks          2         Nodes          4        0
    RBricks          0         CPUs           4        0
    IOBricks         2         Mem(GB)        8        4


How can I find out why it's disabled?

I also can't get an output on the serial port of the main brick with the I/O board. Is this normal behavior?
I can only access the L1 on the brick with the graphics cards.

The systems seems to detect only two of its four SATA disks:

Code: Select all

EFI version 1.10 [14.62] Build flags: EFI64 Running on Intel(R) Itanium Processor EFI_DEBUG
EFI IA-64 SDV/FDK (No BIOS ) [Jun 15 2006 14:26:15] - INTEL

Copyright (c) 2000-2005 Broadcom Corporation
Broadcom NetXtreme Gigabit Ethernet EFI driver v8.1.1


Seg: 1 Bus: 0 Dev: 1 Func: 0 - SGI IOC4 ATA detected: Firmware Rev 83
  Ata(Primary,Master) MATSHITADVD-ROM SR-8178 PZ21

Seg: 1 Bus: 0 Dev: 3 Func: 0 - Vitesse Serial ATA detected: Firmware Rev 1
  Sata(Pun1,Lun0) ST3200827AS      3.AAD                4ND212Z3
  Sata(Pun2,Lun0) ST3200827AS      3.AAD                4ND21QHM


Is it true that disk drives can only be used when a PCI I/O module is installed?
In that case, I wonder why this Prism has four disks installed.

I don't have a proper monitor at hand right now. If possible I would like to install the original SuSE Linux on the system. Does anybody know if it can be installed with serial console, but still later use the graphics?

thanks,
Dennis

[PymbleSoftware]

when it boots up it disables 4GB of memory:

Code: Select all

EFI Boot Manager ver 1.10 [14.62]

Partition  0:                           Enabled Disabled
    CBricks          2         Nodes          4        0
    RBricks          0         CPUs           4        0
    IOBricks         2         Mem(GB)        8        4


How can I find out why it's disabled?

I know nothing about prisms but in older IP53 machines when it disables memory there is something it doesn't like about the RAM, dust on connectors or not seated properly or the like . I solved this on my systems several times by reseating the RAM and inspecting for any dust on connectors. Failing that shuffle the RAM between bricks to make sure its not the board but the RAM itself. If bad RAM throw out or put in a anti-static bag with the label "suspect".

R.

[jeffwa]

First, I can't get an answer from the main bricks serial port. I do get to the L1 on the brick with the graphics cards. config shows both bricks:

Are you using the serial port or the USB connector? I would see if you can get into the L1 from the working brick. If you can, there is an option in there for setting the L1 communications behavior (options are something like AUTO, SERIAL, USB, etc.) Unfortunately I can't remember for the life of me what the command is but I know this setting can be a PITA if set incorrectly and you're trying to use the USB port.

first, when it boots up it says:

Code: Select all

ALERT: VRM not OK: type CPU C 12V, number 2


however it seems to work.

001c04-L1>power vrm
VRM Type Location Present Okay
-------------- -------- ------- -------
CPU A 12V 1 passed passed
CPU C 12V 2 passed FAILED
NODE 2.5V on board passed passed
NODE CPU A on board passed passed
NODE CPU C on board passed passed

That's going to be the power pod on the CPU, which is a bummer. However, there is a slim chance a reseat will clear things up. The power pod utilizes an edge connector on the IA CPU board - maybe it's just loose, or better yet maybe the cable is disconnected It's been awhile since I've replaced one, but I think you can remove the 2 (or 4..I forget) screws from the pod and slide it back away from the CPU. Otherwise just uncam the CPU, remove all the screws and pull the whole works out.

If that doesn't work, start looking for a new power pod (probably easier to find a CPU w/power pod) or just disable the CPU from the L1 and ignore the warning.

when it boots up it disables 4GB of memory:

Code: Select all

EFI Boot Manager ver 1.10 [14.62]

Partition  0:                           Enabled Disabled
    CBricks          2         Nodes          4        0
    RBricks          0         CPUs           4        0
    IOBricks         2         Mem(GB)        8        4


How can I find out why it's disabled?

As previously suggested, reseat the memory. What's the memory config? 4GB in each brick? If so and the disabled 4GB is all in one brick something is fishy. Remember the memory is in banks, so one bad DIMM and you'll drop the whole bank.

I also can't get an output on the serial port of the main brick with the I/O board. Is this normal behavior?
I can only access the L1 on the brick with the graphics cards.

Not normal behavior, but as I mentioned earlier I believe there is an L1 option that can cause some grief. I'll try to dig that info up and post it later.

Is it true that disk drives can only be used when a PCI I/O module is installed?
In that case, I wonder why this Prism has four disks installed.

Correct, the SCSI (IO9) or SATA (IO10) controller is located on the IO9 or IO10 respectivly. No IO, no drives. 4 drives in a two brick/dual pipe config would be odd, not to mention impossible with the graphics options that take two PCI slots. The reason it would be odd in a two brick/two pipe config is that the IO9/10 runs at 66MHz, so PCI bus 0 will be limited to 66MHz, thus slowing down the graphics card on bus 0 (wait, now my memory is failing...did the ATI graphics options run at 66 or 133?). Regardless, this was not a standard config and

I don't have a proper monitor at hand right now. If possible I would like to install the original SuSE Linux on the system. Does anybody know if it can be installed with serial console, but still later use the graphics?

Yes, in fact you have to install via console because the graphics won't be initialized until you install Pro Pack on it.

[lunatic]

Are you using the serial port or the USB connector? I would see if you can get into the L1 from the working brick. If you can, there is an option in there for setting the L1 communications behavior (options are something like AUTO, SERIAL, USB, etc.) Unfortunately I can't remember for the life of me what the command is but I know this setting can be a PITA if set incorrectly and you're trying to use the USB port.

I did use the serial port CONSOLE where the L1 should normally answer.
Now, I have tried attaching a monitor and a USB keyboard and I actually get a picture. The box boots up normally into EFI.
Still, there is nothing on the CONSOLE of the main brick. I wonder maybe it's disabling that port because of the graphics?

That's going to be the power pod on the CPU, which is a bummer. However, there is a slim chance a reseat will clear things up. The power pod utilizes an edge connector on the IA CPU board - maybe it's just loose, or better yet maybe the cable is disconnected It's been awhile since I've replaced one, but I think you can remove the 2 (or 4..I forget) screws from the pod and slide it back away from the CPU. Otherwise just uncam the CPU, remove all the screws and pull the whole works out.

Well, it works, so I will keep it there, right now.

As previously suggested, reseat the memory. What's the memory config? 4GB in each brick? If so and the disabled 4GB is all in one brick something is fishy. Remember the memory is in banks, so one bad DIMM and you'll drop the whole bank.

Can the firmware tell me which bank of memory it has disabled?

Correct, the SCSI (IO9) or SATA (IO10) controller is located on the IO9 or IO10 respectivly. No IO, no drives. 4 drives in a two brick/dual pipe config would be odd, not to mention impossible with the graphics options that take two PCI slots. The reason it would be odd in a two brick/two pipe config is that the IO9/10 runs at 66MHz, so PCI bus 0 will be limited to 66MHz, thus slowing down the graphics card on bus 0 (wait, now my memory is failing...did the ATI graphics options run at 66 or 133?). Regardless, this was not a standard config and

Maybe someone placed the two disks in there as quick spare parts.
The graphics cards seem to be AGP cards.

Dennis

[lunatic]

Another problem:
When I try to boot the SLES 9 CD1, I can choose the type of installation and then it only says this:

Code: Select all

Uncompressing Linux... done
Loading initrd initrd...\


after this, the system just silently reboots.
The odd thing is that it has a Debian 6.0 installed, which seems to work fine.

[lunatic]

never mind, the SLES 9 service pack 3 CD1 boots fine. The plain SLES 9 CD1 doesn't boot.

[lunatic]

I just found out that the Prism specific drivers are on a third CD of the SGI Propack. I do have Propack 4 SP 4, but only CD 1 and 2.

[jeffwa]

I did use the serial port CONSOLE where the L1 should normally answer.
Now, I have tried attaching a monitor and a USB keyboard and I actually get a picture. The box boots up normally into EFI.
Still, there is nothing on the CONSOLE of the main brick. I wonder maybe it's disabling that port because of the graphics?

Wait, I thought you said the XG2N brick was the one you COULD get a console on? I suppose it doesn't really matter because you can install via keyboard/mouse + graphics (note that you'll get console only until disk 3 is installed). Which, could mean you're on to something about the console being disabled due to graphics. Perhaps they redirect the console to AGP?

Can the firmware tell me which bank of memory it has disabled?

Not that I know of. POD will tell you which node the disabled DIMM's are in. Serial dimm -v might as well, I can't remember. I'd also watch your console during boot, or look at the log on the l1. It just takes a couple minutes to reseat the DIMM's...as another poster noted, this is a common problem, especially with systems rattled by shipping.

The graphics cards seem to be AGP cards.

Doh! You're right, I forgot they put an dual port AGP riser in there. Even more reason why it would be impossible for there to be an IO option in a XG2N, there are no PCI slots!

I just found out that the Prism specific drivers are on a third CD of the SGI Propack. I do have Propack 4 SP 4, but only CD 1 and 2.

Correct, you will need the SGI ProPack Graphics & SGI ProPack Kernel Modules-rtgfx off disk 3 to get graphics in SLES