Mechanical DESIGN FLAW in the x346 when adding 2nd processor!!

[Dimarc67]

Hi, All.

Just got in an order of seven x346 servers (8840-11U), each with a 2nd processor upgrade (13N0688). The processor upgrade, of course, comes with the necessary additional heat sink, and Voltage Regulator Module (VRM) card as well.

I installed the first processor upgrade without issue. As per the installation instructions: remove the air-baffle and dummy heat sink, install the processor, sink, and VRM, then replace the baffle.

Here's the issue: The clear, plastic air-baffle has a small groove molded into its underside to cradle the top-rear corner of the VRM. However, the VRMs that I received (all seven are identical) have a tiny soldered component on that corner of the card, and the component prevents the baffle from seating properly.

I first noticed this on the 2nd upgrade I was installing. I immediately assumed that perhaps the component had been soldered a little out of position. But then I found the exact same thing on the third upgrade I was installing. At this point, I went back to check the first installation I'd done. I found that the soldered component had been hit by the baffle here, too, BUT THE BAFFLE HAD SNAPPED THE COMPONENT RIGHT OF THE CARD. I found it lying lose on the motherboard next to the VRM card outlet.

I called IBM, and they're willing to send out a "Certified Engineer" to take a look. Maybe if I'm the first to report this, they'll give me something special... you know, like a free IBM mousepad or something.

Anyway, the quick fix to completely avoid this issue is to break off a 1/8" square piece of the baffle so it clears the errant component.

But in the meantime, I've got a (literally) broken VRM for which I have to wait for replacement. I guarantee there are/will be other x346's out there that have to go through this as well.

Just be careful installing the baffle!! It's easy to not notice that it's not seated properly...

Happy Trails.

David Marcus
David@Dimarc.com

 

[Dimarc67]

Just a quick follow up...

I couldn't wait for IBM to send an Engineer, so I got a quick replacement VRM from my reseller and "modified" the plastic baffle so it doesn't interfere.

I've since been told that IBM is aware of the issue and that they're engineers are passing the word along to look out for this, though I don't think the word is spreading well. It's such a little mechanical design issue.

I'm also told that a system with a VRM damaged by the baffle in this manner will have trouble booting to a RAID with a ServeRAID 7k. It's possible that if you're not using a 7k, you may not notice any specific problems. I discovered my broken VRM before the server's first boot (and I didn't have a 7k in it anyway), so I don't know what issues there may or may not have been.

A malfunctioning voltage regulator can cause very subtle issues in a variety of subsystems, all or none of which that could affect system performance and consistency. But who needs the headaches of wondering, much less tracking down a memory leak due to bad power???

Anyway, keep an eye on your dual-processor x346's. Bon appetite.

David Marcus
David@Dimarc.com

 

[Dimarc67]

IBM has since corrected the baffle design. We just received a new x346 (884011U), and the air baffle nows clears the VRM component correctly.

The air baffles on our original seven x346 servers are labeled with part number 90P4626, FRU 26K4767, "2004-09-21". The baffle on the new server is labeled with part number 25R5233, FRU 25R5234, "2004-12-17". (Note--the part number molded into the plastic of both types of baffle is the old part number.)

David Marcus
David@Dimarc.com