So this summer I thought it was time to upgrade from a quad-opteron build that I was using as workstation/hobby server. I settled for a Supermicro H11DSi-NT and two Epyc 7551, that I bought on Ebay. Finding compatible RAM for a price that was in line of the rest of the build was the biggest headscratcher but after searching for MEM-DR416L-HL02-ER26 (something that I believe is compatible according to the motherboard website) on Ebay I found something called Hynix 16GB 1Rx4 PC4-2666V RDIMM DDR4-21300 ECC REG that the seller said was compatible with MEM-DR416L-HL02-ER26.
Now that I have gotten all of the components I cannot get the system to post. It turns on, the fans spin at full speed and don’t stop, if a GPU is connected the fans will turn for a short while then turn off for a few seconds and back on again, this happens two or three times and after that the fans on the GPU will spin without pausing again. I have tried the built in VGA port, two different GPU’s and disabling the built in graphics when trying the GPU’s.
I tightened the CPU’s with a torque screwdriver to just barely 1.6 nm which should be very close to what the manual said.
Another weird thing that happened is that initially the IPMI module was working fine and the LED was blinking but while using it in browser mode the module suddenly went offline and the LED stopped blinking and hasn’t worked since (and no LED since).
All of the components are new and at first I was skeptical towards the RAM actually being compatible but with the weird IPMI behavior I am leaning towards the motherboard being faulty? All of the components were bought as new on Ebay.
So now I’m not quite sure how to proceed, does anyone know if first gen Epyc is picky about RAM? Any help and tips appreciated.
re-sit the cpu. plug in 2 sticks of ram in the 2 the recomended slots for a 2 stick config (found in the motherboard manual), plug in the gpu.
boot the system… (fans should spin up to 100%)
wait for the memory to train.
wait for post (fans should quieten then run at at even rate).
if it doesn’t post, reset the cmos/bios/eufi. reboot and let it train up again.
hopefully it will boot.
also dont mix and match timings. use the ram that has the same timings only. sell the rest if you bought it… then buy more of the stuff you kept. with the same timings and ic config.
EPYC is not really sensitive to RAM to at least boot. performance is a separate issue though.
have you tried a single CPU single RAM config with each CPU separately to see if a CPU or seating issue has occurred. usually that is the issue i have had, this socket is super sensitive to alignment issues and that dang frame is not always perfect.
The IPMI behavior is weird, but just for a reference, my H11SSL-i + 7551P has similar behavior: no GPU output, fan at full speed, CPU fan ramping up and down when a GPU is plugged in, etc.
In my case, I also couldn’t get output from a GPU, and disabling onboard VGA via jumper also doesn’t help. Since the board was bought off eBay, the seller also didn’t reset the IPMI settings, and I couldn’t access it. I needed to use the VGA port to enter the BIOS. Sadly, the VGA port on the board also doesn’t work with a VGA adapter. I had to buy a cheap VGA monitor to get an output.
After I used ipmicfg to do IPMI factory reset, it has been working great since. (I have not tested GPU output, since I don’t really need it.)
i would try it now your up and running.
if it works, then great (it likely will now you got it running)
if it doesn’t you know you still have issues. which sux but at least you know.
Thank you everyone for your replies, unfortunately I still haven’t gotten the thing to post yet. The CPU’s came new in the box (and were sold as new) with the AMD sticker intact so they should not be vendor locked. I kinda got the impression that someone in the US had a firesale on 7551 and several smaller sellers are selling them for around 100 USD.
I have tried reseating the CPU’s several times and adjusting the torque slightly, I have tried resetting the CMOS and using only two ramsticks (one per socket). The motherboard is not in a case while trying all of this. All 16 ramsticks are new and identical.
At this point I suspect the motherboard being faulty.
with a cpu installed but NO RAM installed you should get a post code. if you just get the fans spinning but no post code, that would confirm some sort of board issue.
It didn’t change anything, do you know if the fan should slow down after posting? I want to try another VGA cable but I can’t find one which is ironic since I have received several in the last few years when buying monitors…
I contacted the seller on ebay who agreed that it might be a faulty motherboard due to the IPMI light disappearing.
I have 7302P and H11SSL-F. I had various problems getting things working.
Older PSU would not let CPU post. New good quality PSU required.
While I was trying to decide if I had PSU issues or memory issues, 32GB 2133 LRDIMMs, tried BIOS upgrades, up and down and back again before I could get it to work.
IPMI licence hash issues, eventually solved with GitHub, which allowed better access to BIOS and BIOS config.
Was seeing messages that correlated with BIOS broken, need to send back to SMC for fix. Hash allowed rewrite via IPMI.
Very educational.
I picked up three of these boards from a reputable Vendor late last year and had them stored few a month or two while I was moving.
I’m running 8GB ECC REG dimms, validated 7B12 64Core cpus on two of the three and a pair of 7551’s. All CPUs and memory were from STH forum members and validated/known working.
I’ve had similar issues as OP with these boards. The IPMI malfunctions or stops functioning, I found 2 of the three had damaged pins from form manufacturing for CPU socket 2 and the other one had damaged pins for Socket 1.
All three were returned to Supermicro for testing under RMA and replaced with new boards, since them repairing the sockets did not resolve the issues.
that being said.
OP, carefully reseat the CPUs, try one at a time, one socket/dimm at a time.
Trace the Problem Determination through to figure out if you have a viable board or a bad one.
as others have said, you may have Vendor locked CPUs.