EPYC3251D4I-2T build

Notice: Page may contain affiliate links for which we may earn a small commission through services like Amazon Affiliates or Skimlinks.

Aquatechie

Member
Oct 29, 2015
35
10
8
There's a new BIOS (1.20) and BMC Firmware (01.30.00) available since a few days.

EDIT: @chn already posted this information in another thread
Has any one encountered an issue with the System Inventory data being blank post-upgrade on this release of BIOS & BMC firmware? No Processor, Memory Controller, PCIe Device / Function, or Storage data is shown. Fan information, however, is being reported?
 

nasi

Member
Feb 25, 2020
64
20
8
Has any one encountered an issue with the System Inventory data being blank post-upgrade on this release of BIOS & BMC firmware? No Processor, Memory Controller, PCIe Device / Function, or Storage data is shown. Fan information, however, is being reported?
I would like to check that on my system but unfortunately my IPMI is not reachable. Server is running since a while without any issue.
Anyone else's IPMI seems to turn off? Anything I can do to prevent that?
 

cardinaldebiere

New Member
Jun 17, 2019
3
0
1
Has any one encountered an issue with the System Inventory data being blank post-upgrade on this release of BIOS & BMC firmware? No Processor, Memory Controller, PCIe Device / Function, or Storage data is shown. Fan information, however, is being reported?
I have exactly the same behavior in System Inventory (BMC is 1.30.00 and Bios is P1.20), only fans information is being reported. I cannot tell if it was the case with any previous BMC or Bios revisions.
 

Aquatechie

Member
Oct 29, 2015
35
10
8
I have exactly the same behavior in System Inventory (BMC is 1.30.00 and Bios is P1.20), only fans information is being reported. I cannot tell if it was the case with any previous BMC or Bios revisions.
Oh, it definitely worked in releases prior to the latest currently posted on Asrock's site. I'm somewhat glad to see I'm not alone. I do have to admit that this has been one rather buggy board right out the gate.
 

nagl23

New Member
Sep 29, 2021
8
2
3
Hi guys! I'm building the same machine (except the Case having a larger 350W PSU) - It's almost done, one more adapter cable is missing to complete the build.

I have one question about he CSE-721TQ-350B case. Behind the drives is a built-in SAS controller connecting the 4 disks. I have plugged in 4 SATA cables (using OcuLink connector) and 2 SATA power cables. This looks like the minimum required to make it work, but there's a lot of other pins and ports on the SAS controller board. I have no idea if I should touch them or connect anything else to the motherboard?

I'm attaching a photo of the SAS board and a docs page that is very rudimentary - I have connected numbers (1) and (4).
 

Attachments

csp-guy

Active Member
Jun 26, 2019
372
153
43
Hungary, Budapest
Hi guys! I'm building the same machine (except the Case having a larger 350W PSU) - It's almost done, one more adapter cable is missing to complete the build.

I have one question about he CSE-721TQ-350B case. Behind the drives is a built-in SAS controller connecting the 4 disks. I have plugged in 4 SATA cables (using OcuLink connector) and 2 SATA power cables. This looks like the minimum required to make it work, but there's a lot of other pins and ports on the SAS controller board. I have no idea if I should touch them or connect anything else to the motherboard?

I'm attaching a photo of the SAS board and a docs page that is very rudimentary - I have connected numbers (1) and (4).

Hi Nagl23!

You have sideband connector to this backplane.

The sideband cables are used to connect the controller to a managed backplane or disk enclosure, so that it can monitor the controller as well as the disks.

I've got IBM M1015 SAS SATA controller to my backplane. The oculink cable you are using has not equipped with sideband cable.
 

nasi

Member
Feb 25, 2020
64
20
8
I've never used a backplane yet so noob-question incoming: do you need sideband connector to get backplane and drives working? Or is it just nice to have?
 

nagl23

New Member
Sep 29, 2021
8
2
3
You have sideband connector to this backplane.

The sideband cables are used to connect the controller to a managed backplane or disk enclosure, so that it can monitor the controller as well as the disks.

I've got IBM M1015 SAS SATA controller to my backplane. The oculink cable you are using has not equipped with sideband cable.
Thanks for the explanation!

Please confirm my understanding is correct:

On the "ASRock EPYC3251D4I-2T" board there is a "OCuLink x4 Connector" port with included cable, but this cable is without a sideband - this port doesn't support it. So I can connect disks directly, but not throught the backlpate controller that is built into the SuperMicro case.

There the "IBM M1015 SAS SATA" (or similar one) controller comes - it goes into the PCIe slot and provides 2x Mini SAS 36P SFF-8087 ports. Those ports support cables with "4x SATA and sideband" like this one: Amphenol Mini-SAS SFF-8087 36Pos Straight to 4 x SATA Latch Cable - Pactech

Connecting the backplate controller the IMB PCIe contoller using this cable (4x sata + sideband) will make the 4 drives in the case slots work (with monitoring).
 

nagl23

New Member
Sep 29, 2021
8
2
3
I've got the IBM M1015 SAS SATA controller, but for now I have another issue - I can't make the system to power up.

Some ideas:

1) The PSU is working, the fan is turning on when I short-circuit green cable in the 24-pin connector.

2) The power button cable 2-pin connector from the CBL-0084L cable is ok (it's short-cicuted when I press the button).

3) Power connectors to the board - I've connected both 8-pin and 4-pin. The one with 8-pin is connected using a 24-pin to 8-pin adapter.

4) Not everything is connected to the "System Panel Header" port on the mother board, can this be the issue here? I'm using CBL-0084L cable to split the 16-pin cable from the SuperMicro chassis. The only thing I'm not able to connect is POWER LED as it has 3-pin connected instead of 2-pin that is in the port. The rest is connected accordingly to "FreeNAS_Mini_Motherboard_Removal_and_Replacement_Guide_WEB.pdf" which shows similar connection. See the attached docs.

What can be wrong here?
 

Attachments

itronin

Well-Known Member
Nov 24, 2018
1,234
794
113
Denver, Colorado
4) Not everything is connected to the "System Panel Header" port on the mother board, can this be the issue here? I'm using CBL-0084L cable to split the 16-pin cable from the SuperMicro chassis. The only thing I'm not able to connect is POWER LED as it has 3-pin connected instead of 2-pin that is in the port. The rest is connected accordingly to "FreeNAS_Mini_Motherboard_Removal_and_Replacement_Guide_WEB.pdf" which shows similar connection. See the attached docs.

What can be wrong here?
your "additional" front panel connectors are indicator LED's and not required.

power led - while this "fix" won't help your power on issue, you can change that 3 pin power led position and at least solve that issue. Use a paperclip or a straight pin, or the tweezers from a swiss army knife to relocate the green connector to the middle position in the housing. You'll hang the the empty pin hole to the left of your power led 2 pin position. If the empty pin position overhang bothers you then you can always acquire a 2 position housing or 2 one position housings, remove the pins from the three position housing and insert into whichever housings you acquired.
NB: on the FP diagram the the green circle for the power led is "hot" . Humorously enough this corresponds to the green wire on your 3 position FP connector.

you may have checked these but in my experience the most common type sof "my system won't start" in a new build are due to:

1) card incompatibility - did you try removing the M1015 to see if that solved your power on issue?

2) a short against the bottom of the motherboard. remove the motherboard and make sure there are not any "spare" standoffs underneath the motherboard in non standoff positions.

While you have the motherboard out of the system chassis you might consider performing a bench build. place the motherboard on a non conductive surface, cable it up and see if it fires up. If your power cables won't reach but you have a spare ATX power supply you may be able to use that. Try bench build tesitng without the M1015 installed and with the M1015 installed as well.
 
  • Like
Reactions: nagl23

Aquatechie

Member
Oct 29, 2015
35
10
8
I think I see your problem. In the image power_cables.jpg, the white power connector labeled "P2" looks out of place. My board came with a ATX 24 pin-to-4 pin adapter that is plugged in where your P2 is currently connected. P2 is plugged into the rightmost half of the 8-pin connector where the yellow, purple & black wires are. This allows me to power the board on/off using the chassis power button.

Don't know if this may help any, but here's the legend I made for myself creating a wiring harness to connect my Supermicro chassis to my Asrock board. What model chassis are you using if you don't mind my asking?
 

Attachments

nagl23

New Member
Sep 29, 2021
8
2
3
1) card incompatibility - did you try removing the M1015 to see if that solved your power on issue?
I don't have the card connected at the moment.

2) a short against the bottom of the motherboard. remove the motherboard and make sure there are not any "spare" standoffs underneath the motherboard in non standoff positions.
I'll check that tomorrow. :)


power led - while this "fix" won't help your power on issue, you can change that 3 pin power led position and at least solve that issue. Use a paperclip or a straight pin, or the tweezers from a swiss army knife to relocate the green connector to the middle position in the housing. You'll hang the the empty pin hole to the left of your power led 2 pin position. If the empty pin position overhang bothers you then you can always acquire a 2 position housing or 2 one position housings, remove the pins from the three position housing and insert into whichever housings you acquired.
NB: on the FP diagram the the green circle for the power led is "hot" . Humorously enough this corresponds to the green wire on your 3 position FP connector.
I was able to replace the cable in the 3-pin connector (with this helpful video:
). Now everything looks right, connected exactly like in the chart, but it still doesn't work.


I think I see your problem. In the image power_cables.jpg, the white power connector labeled "P2" looks out of place. My board came with a ATX 24 pin-to-4 pin adapter that is plugged in where your P2 is currently connected. P2 is plugged into the rightmost half of the 8-pin connector where the yellow, purple & black wires are. This allows me to power the board on/off using the chassis power button.

It might be the issue, the P2 comes from the PSU directly, maybe I need to derive both the 4-pin and 8-pin cables from the 24-pin one?


I'm using this chassis: CSE-721TQ-350B (docs)
And this motherboard: EPYC3251D4I-2T (docs)
 

Attachments

nagl23

New Member
Sep 29, 2021
8
2
3

Attachments

Last edited:
  • Like
Reactions: Aquatechie and nasi

nagl23

New Member
Sep 29, 2021
8
2
3
AsRock was kind enough to send me a beta BMC firmware (L1.32.00) that adds fan control options to the IPMI webUI.

I had a ton of Card Side Temperature Upper Non-critical - going high - Asserted/Deasserted events in the IPMI Event log due to the fact that all my case fans would only spin up based on the CPU temperature sensor. So when the CPU sat at idle, the case fans weren't pushing enough airflow to keep that Card Side Temperature sensor cool. The onboard LAN sensors also got rather uncomfortably warm (70C) while idle, but at least that wasn't blowing up my IPMI log. With the case fans manually set to 80% (~1100RPM), things are much cooler and still quiet.

Here is the link to the 1.32 BMC firmware
The link is down. Can you please publish it again?


1.30.00 is still the newest version available on AsRock website: https://download.asrock.com/BIOS/Server/EPYC3251D4I-2T(01.30.00)BMC.zip
 

drdaeman

New Member
Nov 12, 2021
1
0
1
On the "ASRock EPYC3251D4I-2T" board there is a "OCuLink x4 Connector" port with included cable, but this cable is without a sideband - this port doesn't support it. So I can connect disks directly, but not throught the backlpate controller that is built into the SuperMicro case.
Hey, I'm building a network appliance with the same motherboard and case/enclosure and wonder about the same exact thing. I'm still waiting for the drives to arrive, but from what I understand about how everything should work, I believe I won't get activity and failure LEDs working properly without some sort of sideband connection. I don't think I want to buy and plug a PCIe controller just for this - my rationale is that motherboard has its own controller and I can find some better uses for the slot.

So... There is a SATA_SGPIO1 connector on the motherboard. It looks like it matches with SC721's backplate's (I think it has two versions, mine is with BPN-SAS-733TQ) JP51 sideband header. I've also compared it to Supermicro's own MBD-X11SCL-F motherboard I-SGPIO1 pinout and looks like it's either the same or mirrored across the long axis - a bit hard to tell because they don't not draw the pins but lists them by numbers and seem to mix numbering schemes.

I wonder switching backplate to SGPIO mode and connecting JP51 with SATA_SGPIO1 would get the LEDs working.

There are CBL-CDAT-0661 or CBL-0157L cables from Supermicro, I believe one is 1:1 and another is mirrored. But I cannot find pinouts for those cables, so I'd probably go with a DIY solution out of individual DuPont wires. This looks like a fairly simple serial bus, so I guess if I'd connect the pins according to their signals (clock to clock, data to data, grounds to grounds, etc), everything should be fine.

If someone had already tried something like this, has some success or failure stories, or knows a readily available cable that works in this scenario - please tell :)

Thanks!

---

Edit 2021-11-20: Okay, my drives had arrived, and it looks like at least the activity LEDs work even without the SGPIO cable. I wonder if there's a way to test a failure LED (fortunately or not, I don't have any faulty drives, hah)... But anyway, now I lean towards that maybe I was overthinking it and it's not needed at all.
 
Last edited:

nasi

Member
Feb 25, 2020
64
20
8
I've build a watercooled server based on this nice little mainboard.
  • Bitfenix Prodigy case with insulation material
  • Asrock Rack EPYC3251D4I-2T
  • 2x Kioxia SATA SSD 1.92TB
  • 2x Seagate IronWolf 6TB 5400rpm
  • Intel Optane 800p 58GB (not in use yet, with Corsair cooler)
  • external 12V PSU, MeanWell GST120A with max. 102W
  • Alphacool Eisstation DC-LT pump+reservoir (running on 7V)
  • modified Aquacomputer Cuplex Kryos CPU cooler
  • custom made Anfitec VRM cooler (you can order it by writing them an email, maybe it will be available in their onlineshop soon Anfi-tec )
  • Watercool HTSF2 1x120 LT radiator
Update on this: my Alphacool Eisstation water pump died and is now leaking water.
The system was running without water for an unknown time, I guess due to evaporation. Once I noticed the burning hot cpu cooler I added water while the system was still running. I think that was a big mistake. A few minutes later I noticed a lot of water in the basement of the case. Seems like the freshly added water became very hot in the first moments and some piece of plastics in the water pump couldn't stand this temperature and broke. Lesson learned for the future. But still interesting that CPU and VRMs where only passively cooled and system was still working for a while.

So now I've bought a Aquacomputer Aqualis D5 with "Micro" reservoir, have added a fill port and a drain port for easier service and have moved the HDD cage underneath the mainboard (where normally the ATX PSU would be) to make more space for the water pump. Now everything is running fine again, temperatures are in the same area as before and it's still very low noise.
The Aquacomputer D5 pump has a potentiometer to adjust it's speed. It's running very slow but that's enough for my needs and it's nearly inaudible.
Most time-consuming part of rebuild was to find a solution to run tubes around narrow corners (the case is quite small). I ended up with having almost no tube at some points.
DSC06018_small.JPGDSC06028_small.JPGDSC06023_small.JPG