Bug in Intel Atom C2000 series processors?

[Evan]

Yes the C0 stepping is band new in 2017.
Maybe pickup a 2nd board and make sure it's fixed and just keep it as a spare.
It's still not likely it fails.

Do note that certain usages like how Cisco uses them makes the issue more prevalent (at least reading between the lines)

 

[Black Ninja]

I am really hoping to find out what's really has been done to fix the issue with Atom C2000 if anything. I smell some conspiracy here, and I'll tell you why:

I gave supermicro support team my serial number of one of the atom C2000 I have and they said they do RMA replacement despite the board is out of warranty. But I didn't send it to them , I order brand new C2000 instead, that came from them on Oct 2017.
Then I gave the new serial number to super-micro support team and got this response : "Please issue a ticket for ZMxxxxxxx RMA and put “C2000 RMA” in the note section. Please let me know the serial # for the rest of the boards you have."

Do you guys see what is wrong with this picture ?

P.S. Anybody who got a replacement with new "fixed" board should think twice now. Did you get anything different really, but different board so you don't complaint about the bug. According to supermicro it don't matter if you get the new stepping C0 as far as the bug is concern. They told me over the phone that there is no way for me to tell if the bug is there or not , unless I gave him serial number so they can tell me.

 

[Drewy]

One of my 2758 SM boards has just starting farting " NMI: IOCK error (debug interrupt?) for reason 71 on CPU 0." every 10 seconds or so.
Symptom of impending doom ?

 

[Derek Piper]

I found this page googling for the NMI: IOCK error. I'm getting it too, and also on an Intel Atom C2550 board.


NMI: IOCK error (debug interrupt?) for reason 61 on CPU 0
NMI: IOCK error (debug interrupt?) for reason 71 on CPU 0

Hardware bug huh? ... great :/

 

[Drewy]

Same as mine, so could be.
I ended up sending my board back to SM. They didn’t seem as helpful as they had done when I approached them some months earlier. I got the board returned reasonably quickly. Can’t see any obvious modifications. They said they’d just updated the firmware. However it now works, so go figure....

 

[leonroy]

Had two boards fail within two months of each other.

One an A1SRi-2758F (8 core) lasted almost 6 years, the other an A1SRi-2558F (4 core) lasted just shy 3 years.

Interestingly enough the 8 core arrived with a cooling fan in the server. The 4 core had no cooling fan. When we queried this the reseller was adamant that the server did not need one. Now that they've repaired the 4 core unit they sent it back with a cooling fan fitted.

 

[ullbeking]

The 4 core had no cooling fan. When we queried this the reseller was adamant that the server did not need one. Now that they've repaired the 4 core unit they sent it back with a cooling fan fitted.

I _think_ the difference comes down to what chassis you are using. Some SM boards use 1U chassis in such a way that the CPU can be passively cooled, but if you are using arbitrary system then they can be passively cooled. OR... maybe the 255[08] really doesn need a fan.

 

[sfbayzfs]

I finally sent my C2000 boards in for the RMA after removing them all from service - Supermicro is still fixing them, when I asked, they said turnaround is 7-10 business days.

 

[weust]

I think my A1SRi-2758F died too. I can access BMC/IPMI just fine, but booting shows nothing on virtual console or VGA port.
Hadn't used it in a while, at least half a year.
When I power on the fans (PSU and case thingy) work for about a minute or two, then it goes off and tried again.

Mine is from early 2015 and I will contact my reseller first. Hopefully I can get a new one.
Was planning it to be my new OPNsense firewall/router. It was my Hyper-V 2012 R2/2016 and ESXi server before.

 

[mmo]

I think my A1SRi-2758F died too. I can access BMC/IPMI just fine, but booting shows nothing on virtual console or VGA port.
Hadn't used it in a while, at least half a year.
When I power on the fans (PSU and case thingy) work for about a minute or two, then it goes off and tried again.

Mine is from early 2015 and I will contact my reseller first. Hopefully I can get a new one.
Was planning it to be my new OPNsense firewall/router. It was my Hyper-V 2012 R2/2016 and ESXi server before.

if you have no luck with the reseller, you can directly RMA it with Supermicro. I've RMA'd all my C2xxx boards with Supermicro without any issues.

 

[weust]

I will try that, but this is a good reseller and was told they take it up with Supermicro.
For some reason there is a Dutch part of Supermicro with support, etc.

Just need to send my serial number per mail tonight. Didn't have that with me to work, obviously.

 

[weust]

Update: Supermicro says my motherboard is Out Of Warranty, so no luck.
Can imagine is after more then 4 years.

I bought a AMD based Supermicro board as a replacement. Fast enough for OPNsense.

 

[leonroy]

Update: Supermicro says my motherboard is Out Of Warranty, so no luck.
Can imagine is after more then 4 years.

My board was from 2014, Supermicro Netherlands swapped it no issue whatsoever. I find it best to deal with Supermicro directly, not the reseller.

 

[weust]

My board was from 2014, Supermicro Netherlands swapped it no issue whatsoever. I find it best to deal with Supermicro directly, not the reseller.

When did you swap it?

The idea I have now is that once I go to Supermicro directly, they will just lookup the serial number and see it's been looked up before. Same answer.

 

[leonroy]

I contacted Supermicro two weeks ago. Replacement's actually out for delivery today.

 

[weust]

Did you call or send and email?
I can always try.

 

[leonroy]

Did you call or send and email?
I can always try.

email.

 

[weust]

Alright. I will send them an email. Thanks.

 

[sfbayzfs]

If your C2000 board comes up as out of warranty, double check the serial number, it's probably a type-o on your part. (which happened to me)

When I got my boards back, among other things, the reports say "ECO rework done by RR14".

 

[Jyrki]

Have you guys looked those reworked boards really carefully? I assume there would be one or two tiny smd resistors added somewhere. I would rather rework solder those myself than wait 2+ weeks for board returning back.

Edit: I looked around docs a bit. I woud guess JTMP1 (TPM header/Port 80 header) has resistor from pin 1 to +3.3V. +3.3V can be found from pin 9 of the same connector, but I would assume it is also available closer of pin 1 somewhere on the board. I don't have photo of back side of PCB and mine is installed in case and running. I would guess resistor is either 100ohm, 8.2kohm or 10kohm.


Edit2,3: Those wondering steppings reported by software. Reported 0 = stepping A0/A1, Reported 8 = stepping B0. I don't know what stepping C0 reports. Also stepping C0 is reported as 8. There is another field showing it is C0 SoC. I assume because cpu part of SoC is unmodified.