I purchased a Pioneer 1.3 bare board, and I am at a loss for how to get this thing to boot.
I have tried various combinations of the following RAM and boot media:
1 stick of NEMIX RAM MR25600-324 32GB RDIMM in the bottom slot
1 stick of the officially blessed Longsys 32GB 2Rx8 PC43200AA-RE3-14 in the bottom slot
4 sticks of the Longsys RAM
These all with the fedora-disk-gnome-workstation_livecd-f38-20231010-033114.n.0-fix.raw.xz image, which most people report working:
16GB Raspberry Pi branded MicroSD card
64GB Samsung Pro Endurance MicroSD card
128GB SanDisk Extreme Pro MicroSD card
256GB SanDisk Max Endurance MicroSD card
256GB TeamGroup High Endurance MicroSD card
128GB Forsee EMMC module on MicroSD adapter
128GB Forsee EMMC module directly attached to motherboard
No storage at all
I have also tried these things with and without a Radeon R5 240 GPU and no other PCIe devices attached.
Upon powering on, similarly to others here, there is a solid green light near the front panel header and a flashing green light near the USB3 header. With a serial adapter attached to each of the MCU and RISC-V headers, I do get some output on the MCU and am able to issue various commands successfully:
i have the exact same issue, bare board v1.3, no output from rv header but mcu works, can’t get a boot to start, and used one of the same RAM as stated supported list
i could guess from others it could be the RAM but since your ram is the official one, i guess that is not the issue… there must be something non-obvious that we are missing…?
EDIT:
from what i gather in this thread,2Rx4 ram won’t likely work, and someone who had the same issue used a 8GB Samsung 1Rx8did the trick, so I’m gonna order that one and see what happens…
Yeah there’s definitely some issue with different geometries not being tested/supported by their controller. But more importantly, where are the docs or any guidance at all about this stuff?
yes, one thing i have noticed on that ram list is the marking for tested setup was v1.1 or v1.2 or evb (evaluation board?) but not v1.3 so there is some info gap right there it would seem…
They also have stale links to the old AVL in the memory installation instructions. I mean, nobody is asking for perfection but some guidance around qualified parts (especially those actually obtainable from US sources) would be the absolute minimum bar I would expect from someone claiming a “workstation” or “server” board and not an embedded experiment.
I am having the same issue with the v1.3. I changed my PSU and I had the 64GBx2 RAM and it does not post. I just ordered the Samsung RAM to test. I hope that works. Better labeling on the board would be helpful as well.
I am sorry, I am new to this. That is why I got the board, so I can learn. Not sure how to get any output over serial. If you don’t mind, can you give some guidance. Would be appreciated.
There are two different serial connections you can use. Port 23 (MCU, left) will basically tell you if the machine is attempting to initialize. Port 23 (RISC-V, right) will let you interact with the system and is most-likely the connection you want to check first, it has linux kernel debug output and will also be where you can do login without a display connected.
In your motherboard box, you should have gotten a serial cable that looks like this
Once you connect, you won’t see any output. At this point you can reboot your RISC-V board and start seeing if any debug output comes up.
You can leave the cable plugged in for subsequent reboots. One thing you might want to do between reboots is to right click your terminal and click “Reset Terminal” otherwise the text might get messed up sometimes.
RDIMM and the “x8” organization seem to be required. Verify that on any future parts you find in a search.
I would caution you that the speed of 2666 worked for me as a single DIMM, but trying to use 4 of these slower than 3200 speed DIMM’s did not. I had to purchase 3200 speed DIMM to get 4x DIMM configuration to work.
From the first picture, it doesn’t look like you have the RAM in the correct slot.
Recommend you try with a SINGLE DIMM in the slot closest to the PCI Express connector (with your GPU installed? and leave the other three DIMM slots empty.
Look up the part number of your DIMM on the web and confirm that it is REGISTERED (RDIMM) and has a 1Rx8 or 2Rx8 organization and NOT “x4”
If it’s “x4” you will need to find a x8 DIMM for further testing.
Thank you for the response. The RAM I was using was a DDR4 3200MHz 2Rx4 1.2V. I ordered an 8GB 1Rx8 RAM to test. I hope that works. Just waiting for it to come in.
Based on my previous success with a Micron OEM part rebranded and resold, I ordered more. Of course, the new order was a slightly different part number.
Same Micron RAM. Same specs. There are 4 different revisions of the “same” DIMM made by Micron.
MTA18ASF2G72PDZ-3G2R1 (works)
MTA18ASF2G72PDZ-3G2E1 (doesn’t work :sadness:)
According to the datasheets online the specs are identical, the EEPROM on the DIMM are identical. The difference is a revision of the PCB, a different “register” chip and a different memory chip die revision. None of which would be expected to significantly change the configuration or performance. They all fail even in single DIMM configuration. They DO work in another server and I’m going to push them with some memory tests overnight.
sigh… MILK-V, why is your memory training and memory controller so picky?
I think it’s possible that there are three different motherboards listed in this table. The Pioneer, the EVB and a “x4 EVB” where the EVB and the Pioneer only support x8 DIMM’s and the “x4 EVB” is able to support x4 DIMM’s.
The “Hardware” column may be more important that it seems at first glance.