Messages

211

GPU Compute / Accelerators / Re: Current support distributions for Navi AMD

« on: September 07, 2020, 06:47:13 am »

On Debian buster, it should be possible with updated kernel and mesa packages from the buster-backports catalog.

This shows how to search the current versions of mesa packages in buster-backports, nobody backported them at the time of writing this reply.

If I was using a Navi card myself, I would prepare the necessary backports and find a way to make those packages available conveniently.  For now, I decided to start with the RX 580 and wait for the Big Navi cards to arrive.  I don't feel it is a good use of money to buy an RX 5700 today when the Big Navi might come in 2 weeks.

OpenBSD commented that amdgpu is currently being adapted for the powerpc64 port, that list might be the best place to get further details about Navi.

On any distribution, developers are sometimes willing to make an extra effort to support things like this if the manufacturer or another user can provide free or subsidized access to the hardware.

212

Operating Systems and Porting / Re: installers not booting

« on: September 07, 2020, 03:23:06 am »

The OpenBSD memory issue mentioned above has been resolved by a patch.  I created separate topic for any dedicated discussion about OpenBSD 6.8

213

Operating Systems and Porting / OpenBSD 6.8, first official release, testing, outstanding issues

« on: September 07, 2020, 03:22:08 am »

According to the PPC page on the OpenBSD web site, OpenBSD 6.8, which is imminent, will be the first official release to support PPC64

Has anybody else tried their PPC64 installer snapshots recently?

After my attempt to run it last week, they included a patch to support machines with larger amounts of memory and the installer progresses further but it still has a limit somewhere.

The next problem I found was that it becomes stuck when I have the LSI 9207-4i4e HBA in the machine.  If I remove the HBA, the OpenBSD installer boots all the way to the first question:

Code: [Select]

Welcome to the OpenBSD/powerpc64 6.8 installation program.
(I)nstall, (U)pgrade, (A)utoinstall or (S)hell?

If anybody else has time to test it, especially if you have accessories like a HBA or NIC, that can help the OpenBSD 6.8 release gain traction or at least help qualify the list of supported hardware on the wiki.

214

GPU Compute / Accelerators / Re: Which GPU would *you* choose?

« on: September 06, 2020, 07:57:54 am »

Another observation I made when installing the Sapphire Radeon RX 580 Pulse: this particular card has heatsink on the back of the card

In a Blackbird, the 8x slot is on the fan side of the GPU, so the rear heatsink on the RX 580 will not clash with another card.

In a Talos II Lite, there are only two slots and the 8x slot is behind the 16x slot, opposite of the Blackbird layout: therefore, if you put something like a HBA with a front heatsink into the 8x slot, it may be too close to the rear heatsink of the GPU

I tried putting the LSI 9207-4i4e behind the Sapphire Radeon RX 580 Pulse and I found they don't touch but the gap is less than 2mm

The only way around that appears to be choosing another RX 580 variation or using one of the other strategies described above, e.g. a PCIe riser or an 8x GPU

215

Firmware / swapping flash chips, ordering spares

« on: September 06, 2020, 03:43:52 am »

I looked at the wiki pages Updating Firmware and Debricking the BMC

The latter suggests "Remove the BMC SPI flash chip and reflash it with a flash programmer"

Rather than reflashing the existing chip, it may be safer to keep one additional flash chip on site.  Any new images can be flashed to the spare flash chip, it is swapped into the socket and if it doesn't work, the old chip is swapped back in.

Can anybody clarify the exact type of chip to order?  Are both flash chips identical or different part numbers?

Has anybody seen any useful video that shows the correct way to remove the chip?

I have a CH341A flash programmer, what type of cable or adapter is needed to insert the flash chip, or does it fit directly into one of these programmers?

216

Operating Systems and Porting / Re: testing a new machine, memory test, etc

« on: September 06, 2020, 01:56:42 am »

Yes, I had decided to try that but it is not perfect

memtester can't test 100% of the memory because some of the memory is always occupied by the operating system and memtester runs as an application in user space

It also requires you to spend time installing the OS before you can run the test.  The ideal scenario is that you run the test before spending time on installation.

217

Operating Systems and Porting / testing a new machine, memory test, etc

« on: September 05, 2020, 02:53:07 pm »

What burn-in test routines do people use when making a new machine or modifying the machine substantially?

For example, on x86 machines, I frequently run memtest86 from a USB stick before installing the OS.  It is not available for OpenPOWER.  It would be really useful to have something like this accessible in petitboot.

218

Operating Systems and Porting / Re: installers not booting

« on: September 05, 2020, 03:04:33 am »

I had a closer look and the VGA disable jumper was not correctly attached.  After fixing it, the X display started working.

The current Debian config on the machine is buster with the following from backports:

Code: [Select]

bind9-dnsutils                           1:9.16.6-2~bpo10+1                 
bind9-libs:ppc64el                       1:9.16.6-2~bpo10+1                 
firmware-amd-graphics                    20200721-1~bpo10+1                 
firmware-linux-nonfree                   20200721-1~bpo10+1                 
firmware-misc-nonfree                    20200721-1~bpo10+1                 
linux-image-5.7.0-0.bpo.2-powerpc64le    5.7.10-1~bpo10+1                   
linux-image-powerpc64le                  5.7.10-1~bpo10+1                   
wireless-regdb                           2020.04.29-2~bpo10+1

although after discovering the jumper issue, it is not clear if I really needed those things from backports.

Looking closely at the log, I notice the amdgpu driver in Debian was compiled for Xorg 1.20.3:

Code: [Select]

X.Org X Server 1.20.4
...
[  2101.612] (II) Module amdgpu: vendor="X.Org Foundation"
[  2101.612]    compiled for 1.20.3, module version = 18.1.99
[  2101.612]    Module class: X.Org Video Driver
[  2101.612]    ABI class: X.Org Video Driver, version 24.0


This appears to arise from the sequence in which the official Debian packages were compiled and uploaded to the Debian archive.

Overall, my original issue with the Debian installer is resolved, I'll put another comment here after getting feedback about the OpenBSD installer.

219

Operating Systems and Porting / Re: installers not booting

« on: September 04, 2020, 05:26:30 pm »

For the Debian installer boot issue, I tried updating the system with the PNOR file here and now the LSI 9207-4i4e is working

For the OpenBSD installer issue, I've queried it on the OpenBSD ppc mailing list

In Debian, X is able to start on the VGA but it is not working through the RX 580 GPU.  The screen flickers, sometimes it says there is no signal, nothing happens.

X logs show it is sees both the GPU and the monitor, it obtains the monitor ID string and supported resolutions.

I tried installing the jumper on the motherboard to disable the regular VGA, that has not helped.

I've tried various other things, for example
- disabled wayland in gdm config
- apt install firmware-linux-nonfree firmware-amd-graphics
- update the Debian kernel from buster-backports, 5.7.0-0.bpo.2-powerpc64le

Can anybody make any suggestions about how to get the GPU and display functional?

220

GPU Compute / Accelerators / Re: Which GPU would *you* choose?

« on: September 04, 2020, 12:44:14 pm »

The Wikipedia article on the Radeon RX series has a nice table

A key point that jumps out at me there: some of these are 8x cards, e.g. RX 560.  Most manufacturers produce cards that have the same size connection as a 16x card but in fact they are only wired for 8x.  Some people talk about cutting open the end of their 8x slot and using it for one of these cards.

As these cards are very cheap and as people only have 2 slots in their Blackbirds and Talos II Lite, this could be an interesting hack to keep the 16x slot free for some other application.

As a bonus, those cards use a lot less power, generating less heat and noise.

I suspect modifying the slot will void the warranty, maybe Raptor could consider supplying motherboards with a slot that is modified before delivery.

221

Operating Systems and Porting / Re: installers not booting

« on: September 04, 2020, 11:10:21 am »

I had tried changing the option in petitboot configuration from hvc0 to tty1

I also tried pressing 'e' on the Expert mode option and manually adding console=tty1 to the command line

Neither of these things made any difference.

I found a few USB serial dongles but I haven't found my serial cable yet so I've tried ssh and obmc-console-client

For Debian, on the console, I found the kernel had crashed during mpt3sas driver initialization.  I saved the stack trace.  I removed the HBA and now it boots.  The HBA is a spare LSI 9207-4i4e that I had hoped to use temporarily.  It is based on LSI SAS 2308.  Is there any hope for this card working on Talos II?

For OpenBSD, I found a different problem, here is the output:

Code: [Select]

SIGTERM received, booting...
[   22.960137] kexec_core: Starting new kernel
Hello, World!
Copyright (c) 1982, 1986, 1989, 1991, 1993
The Regents of the University of California.  All rights reserved.
Copyright (c) 1995-2020 OpenBSD. All rights reserved.  https://www.OpenBSD.org

OpenBSD 6.8-beta (RAMDISK) #47: Thu Sep  3 19:35:45 MDT 2020
    deraadt@powerpc64.openbsd.org:/usr/src/sys/arch/powerpc64/compile/RAMDISK
real mem  = 137438953472 (131072MB)
panic: buf_mem_init: can't reserve VM for buffers

The operating system has halted.
Please press any key to reboot.

128GB and can't reserve VM for buffers... I didn't fill all the slots so I could add more I suppose.

This is the OpenBSD image I tried to use:

https://cdn.openbsd.org/pub/OpenBSD/snapshots/powerpc64/install68.iso

sha256sum install68.iso
ecfcaf79833358c94ecf8f7390346c8d4f167169f7eb987417a6b532bf1e12e7

222

General Discussion / Re: Fractal Design cases

« on: September 04, 2020, 10:56:55 am »

I tried the above, connecting the case fans through the Fractal case's fan hub.  The fan hub is getting power from the PSU via a SATA power socket and it has a 4-pin input that connects to motherboard FAN1

Visually, it looks like the fans speed up after power on and they slow down after some time has passed.  It sounds really quiet too.

Is there any way to verify the speed of the case fans definitively?  For example, can I force them up to full speed while leaving the CPU fans at minimum, to see if I can hear the case fans alone?

223

Operating Systems and Porting / installers not booting

« on: September 04, 2020, 09:22:43 am »

petitboot starts up and displays the menu

I've tried both the Debian 10 and OpenBSD USB sticks.  Neither of them appears to be working.  The install options (e.g. Debian/Expert mode) appear in petitboot, I select the option I want, the screen goes blank and then displays the message "SIGTERM received, booting..."

Leaving the machine like this for 5 or 10 minutes, nothing more happens, it just appears stuck like that.

224

General OpenPOWER Discussion / Re: Talos II arrived

« on: September 03, 2020, 05:19:15 pm »

Do you mean these products?

https://www.broadcom.com/products/storage/host-bus-adapters

For example:

9400-8i for NVMe+SATA, PCIe 3.1 x8, 8 internal ports, 2 HD Mini-SAS SFF8643

9500-8i for NVMe+SATA, PCIe 4.0 x8, 8 internal ports, 1 x8 SFF-8654 (SlimSAS)

Will the U.2 disks get full speed through either of these cards and their respective connectors?  Is there any reason to get the PCIe 4.0 version, or the older PCIe 3.1 card will be sufficient.

I'm thinking about workstation systems with 1 or 2 of the U.2 disks at most.

225

General Discussion / Re: Fractal Design cases

« on: September 03, 2020, 05:10:15 pm »

I went looking for more details about the fan hub integrated in the case.

It appears that I can connect the 4-pin wire from the hub to one of the zone 3 outputs on the Talos II

The hub will then supply power to all three case fans.

I will test this to see if they run at full speed all the time or if the Talos II is able to control their speed through the hub

If the fan hub can't control the speed then I'll replace the case fans with 4-pin fans and connect those directly to the motherboard.

Does that make sense?