Tested third party hardware for Pioneer

We’ve had a bit of trouble getting a Pioneer V1.2 board to boot properly, so we thought we’d document at least the particular modules it worked with:

ECC RAM: Samsung M393A1K43BB1-CTD
DDR4 rank 1 Synchronous Registered (Buffered) 2667MT/s 1.2V width 72 bits (ECC)
(info probed with dmidecode on donor server)

SD Card: Raspberry Pi branded 16GB microSD
Kernel messages:
[ 5.955450] mmc1 id:0 start:0 end:255 width:255
[ 5.955460] mmc1 listsel:0 tuntap:127
[ 5.955596] mmc1: new ultra high speed DDR50 SDHC card at address 59b4
j[ 5.957094] mmcblk1: mmc1:59b4 USD 15.0 GiB
[ 5.960928] evm: Initialising EVM extended attributes:
[ 5.968894] mmcblk1: p1 p2 p3
Markings:
front: 023386 2965
back: 9181aa dg 09ss2

We had several SD cards that weren’t correctly identified (reported SD032 and read all 0s), and one Raspberry Pi SD card that got stuck in readonly mode, which shows as (ro) in kernel messages.

The DIP switches did not need to be set.

2 Likes

This DDR doesn’t seem to be on the support list.

SG2042 DDR Support List

Exactly. If you check that list, there’s only one module on the supported list for this board (Pioneer v1.2, Longsys RER432A032G7-WFS100), and that one doesn’t seem to be available to purchase. Thus I report the one I got working.

On the plus side, we’ve confirmed more SD cards function with newer bootloaders.

1 Like

I can confirm the following work, from my setup:

RAM: Micron 4x32GB (128GB total) 2Gbx8 PC4-3200AA-RE4-12 Registered ECC
NVMe: Crucial T500 2TB Gen4 NVMe M.2
*SD: Samsung 32GB EVO Plus UHS-1 Class 10 Micro SDHC
**GPU: Dell AMD Radeon RX 580 8GB (JTPTC, CN-0JTPTC)

+ intermittent boot issues after several days of use, appear to be SD timeouts. Not convinced its my SD card, still investigating. See related forum posts.
++ the card does not turn the screen on during boot, so I don’t see boot messages (Plymouth-start), but the desktop shows up once gnome starts. Unplugging and replugging the monitor after the right point during boot will trigger the boot screen to turn on. Most likely a minor driver or related boot config issue.

3 Likes

Strange. I could NOT get:
Samsung 16GB 1Rx4 PC4-3200AA-RC2-12-DC1 (M393A2K40DB3-CWEBQ)
RAM to work. Not single in any slot…

But I WAS able to get:
Micron 16GB 2Rx8 PC4-2666V-RE2-12 (MTA18ASF2G72PDZ-2G6E1)
to work.

Also:
SanDisk SDSQUA4-064G-AN6MA (64GB Sandisk Ultra)

SATA drive detected in all 5 SATA ports. They are numbered 0-4 (scsi bus) or ata1-ata5 (ahci driver) from the serial connector (ata1 aka 0:0:0:0) to the ATX power connector (ata5 aka 4:0:0:0). Note that sda, sdb, etc are dynamically assigned based on discovery order.

Micron 5200 MTFDDAK480TDC SATA drive

I am booted without a GPU and will be testing some network cards next.

The serial consoles seem to be 3.3V signalling but the included cable with the motherboard is 5V. Doesn’t seem to hurt the motherboard. But that does mean that if you have other FTDI or similar USB to pin-header “TTL” or “Digital” signal devices from Arduino or ESP projects you could probably use them.

2 Likes

Unfortunately, the Micron RAM won’t boot with more than 1 stick installed.

And there doesn’t seem to be a way to confirm that the memory controller was set to 2666 instead of 3200…

I have Micron RAM working in all 4 slots (4 x 32 GB). I’ll pull a stick and get the part number.

Micron MTA18ASF4G72PDZ-3G2E1TI
32GB 2RX8 PC4-3200AA-RE4-12

I have 4 of these, for 128 GB total. They seem to work perfectly. I have done a full linux kernel compile with all 64 cores in parallel, no memory issues whatsoever. 1 stick seems to work fine too (slot closest to PCIe), but I have not tried 2 or 3 sticks, only 1 or 4.

3 Likes

Ah, so your previous post that these RAM were x4 was incorrect?

If so, that’s good news as it would mean that we might confidently say that it does require “x8” Registered DIMM and that give others more confidence on choices of RAM that work.

Does seem to require x8 modules from my testing. Finally got it to boot with an 8GB Samsung 1Rx8 module. I’ve got another generic 2Rx8 module on the way that I’m going to test on Sunday.

1 Like

So far I had several modules that didn’t work, including

two different kinds of ECC RDIMMs

  • Samsung 16GB 2Rx4 “PC4-2133P-RA0-10-MB1” (M393A2G40DB0-CPB2Q S)
  • Kingston 8GB 1Rx8 “PC4-3200AA-RD1-13” (KSM32RS8/8MRR)

and two different kinds of non-ECC UDIMMs

  • G.SKILL 8GB "4Gx2 DDR4-2400 (F4-2400C15D-8GRR)
  • Kingston 16GB probably 2Rx8 (KF3200C16D4/16GX)

So yea, not even every x8 ECC RDIMM works

1 Like

@milkv-caravan yes it appears so. I copy-pasted that original text from the eBay sale where I purchased them. Sorry for the confusion; I’ve updated that post. According to Micron, the part number I’m using is actually 2Gb x 8 (2RX8) Registered ECC.

Micron MTA18ASF4G72PDZ-3G2E1TI
32GB 2RX8 PC4-3200AA-RE4-12

hm sadly mine doesn’t boot with that memory either :confused:

I was so hyped by this board but that’s really fading away by now … and I sunk a lot of money into different memory by now >_<

We can add “AMD Radeon HD 7750” to the list of working GPU cards.

I had an old card in an old system that I transplanted and it’s detected and boots into graphics mode with Wayland in the 2023-10-10 Fedora 38 desktop image.

# lspci -v -s 1:0.0
0000:01:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Cape Verde PRO [Radeon HD 7750/8740 / R7 250E] (prog-if 00 [VGA controller])
        Subsystem: XFX Pine Group Inc. Device 3248
        Flags: bus master, fast devsel, latency 0, IRQ 66
        Memory at 4100000000 (64-bit, prefetchable) [size=256M]
        Memory at 4050000000 (64-bit, non-prefetchable) [size=256K]
        I/O ports at 1000 [size=256]
        Expansion ROM at 4050040000 [disabled] [size=128K]
        Capabilities: [48] Vendor Specific Information: Len=08 <?>
        Capabilities: [50] Power Management version 3
        Capabilities: [58] Express Legacy Endpoint, MSI 00
        Capabilities: [a0] MSI: Enable+ Count=1/1 Maskable- 64bit+
        Capabilities: [100] Vendor Specific Information: ID=0001 Rev=1 Len=010 <?>
        Capabilities: [150] Advanced Error Reporting
        Capabilities: [270] Secondary PCI Express
        Kernel driver in use: radeon
        Kernel modules: radeon, amdgpu

I received Milk-V Pioneer Box a few days ago.

It equips 4 sets of DDR4 DIMM.

longsys 32GB 2Rx8 PC4-3200AA-RE3-14 2232
RER432A032G7-WFS100

It booted pre-installed Fedora 38 on nvme0n1p3 without any trouble.
I resized /dev/nvme0n1p3 partition to maximum size using gparted.

$ df -hP
Filesystem Size Used Avail Use% Mounted on
devtmpfs 4.0M 0 4.0M 0% /dev
tmpfs 63G 0 63G 0% /dev/shm
tmpfs 26G 1.5M 26G 1% /run
/dev/nvme0n1p3 939G 17G 922G 2% /
/dev/nvme0n1p2 441M 323M 105M 76% /boot
/dev/nvme0n1p1 122M 44M 79M 36% /boot/efi
tmpfs 63G 8.0K 63G 1% /tmp
tmpfs 13G 136K 13G 1% /run/user/1000
tmpfs 13G 96K 13G 1% /run/user/1001

$ uname -a
Linux fedora-riscv 6.1.31 #1 SMP Thu Jun 15 01:30:00 CST 2023 riscv64 GNU/Linux

I would like to find out what 128GB DDR4 or 64GB DDR4 memory are verified to work.

I have application that I would like to run requires beyond 128GB DRAM.

Idle power consumption is around 90W.

Thanks.

Tammy

1 Like

Based on @kpatterson success with some micron modules I ordered very similar ones:

MICRON MTA18ASF4G72PDZ-3G2R

and finally got my board to boot. Tested with 1 and 2 modules and will also test with 4 once I get two more.

1 Like

Ordered from Newegg:

Actual module that arrived is Micron/Crucial brand part. I’ll get the micron part number tomorrow. 16GB for $37 and four DIMMs work. 2Rx8 3200MHz

Micron MTA18ASF2G72PDZ-3G2R1UI
16GB 2Rx8 PC4-3200AA-RE2-12

WORKS in single and quad DIMM configs.

We have two Milk-V Pioneers on the workbench.

We just tested Samsung M393A1K43BB1-CTD 8gb RAM, which DID work, and Samsung M393A4K40CB2-CTD 32gb RAM which did NOT work. Waiting on Micron MTA18ASF4G72PDZ-3G2R 32gb RAM, which we’ll test next. If anyone can suggest other 32gb RAM sticks that worked for them, we can try to get them to confirm.

We’ve also got Mellanox ConnectX-6 DX CX623106A DP/N F6FXM 2x100g Ethernet cards, which are recognized and don’t seem to cause any problems at boot time, but we haven’t yet tried passing packets through them. And we’ve got Intel QuickAssist 8970 cards, likewise, no obvious problems yet, but not yet tested for functionality.

I’ll update as we progress.