Broken pin on S5500WB12V socket

[matthias_r]

Hello,

I've a broken pin on LGA1366 socket, so it runs only with one L5640 CPU. I would like to go from 64 to 128 GB ram and need the second socket.

Is there a possibility to replace the broken pin with one of the reserved pins?

Actually the board does not work for me with the broken pin, so it is no big deal to try pin repair before replacing the board.

Which one can i securely pull out.

Thanks!

[matthias_r]

Okay, here in the docs is some info... https://www.intel.de/content/www/de/de/processors/xeon/xeon-5600-vol-1-datasheet.html

Think the broken pin is VCC, AL12, PWR (red) and i'm going to replace it with RSVD AV1(green).

[matthias_r]

Just one question... actually the AL12 should be just an one of hundred power supply pin. Are they internally supplying power for a special region of the chip or is it like a stranded wire where a single broken wire makes mostly no difference?

[lEdge]

I know what you're saying.

Unfortunately the pins are not reconstructive. They are actually wrapped around the placement and soldered down. To ensure guaranteed connection.

You best bet is replacing just the whole socket using a UV soldering iron. It's better than the conventional oven. You might be able to find a shop for that.

[matthias_r]

Well, replacing the socket in a shop excluding warranty is nearly the same amount than a new board.

I've seen smb. replacing a single pin (https://m.youtube.com/watch?v=RUL-Ii61GfQ).

Most likely it will not work heating up a single pin with a conventional soldering iron, but i'll give it
a try tomorrow and order the new board after i messed it up...

Btw. why is this server equipment so special? Actually once you're done with it, the entire thing (maybe excluding the changeable disk frames if you stick with same manufacturer) goes to trash. I had my ATX desktop tower for over 30 years and upgraded any component more than once...

[matthias_r]

Don't get me wrong, I appreciate the stuffing as much as you can in 1U art. But producing unnecessary huge amounts of electronic waste while driving Tesla and living in Solar Roof house when 1/3 rack is 120 €, and full rack with 10 times the traffic in colo is around 200 € doesn't make sense to me...

[matthias_r]

Pulling out the pin did not work... Has anybody tried to rip off one pin and solder it to the broken one?

[matthias_r]

Don't ask me how, but it is showing 2 CPUs, 128 gb and running memtest86 right now...

Is there another test suite especially for CPU?

[matthias_r]

Maybe it is a stupid idea, but why not soldering the socket to the circuit board and having the pin designed as replaceable feather, like the socket CPU connection?

No more wasted motherboards because of broken pin sounds good to me..

[matthias_r]

Unfortunately the hardware field trip has come to an end, and I'm going to waste my sparetime with Software development again.

Thanks for the support!

[SergioS_Intel]

Hello matthias_r,

The interactive technical support for the Intel® Server Board S5500WB you have requested has been discontinued. Due to this product being discontinued, Intel Customer Service no longer supports inquiries for it, but perhaps fellow community members have the knowledge to jump in and help. You may also find the Discontinued Products website https://www.intel.com/content/www/us/en/support/discontinued-products.html helpful to address your request. 

Thank you for understanding.

Best regards,

Sergio S.

Intel Customer Support Technician

For firmware updates and troubleshooting tips, visit :https://intel.com/support/serverbios

[lEdge]

Well whatever. The reason they make sockets is purely for the manufacturing. How hard to put in the CPU is it anyway? However I know it's only safely done 80% of the time.
You could also manufacture your own plate to contact the pins and transfer the transit for the cpu contact.
And it's also done to manage heat done by the cpu.

[matthias_r]

I worked in a hardware shop and the technicians had issues to put the heat sink on the raw die without destroying the corner. Then they decided to put that steel plate on top of the CPU which we have right now.

Guess the issue with the socket pins occur mostly in the aftermarket when shipping discarded mainboards without the socket plastic cover.

Currently I'm setting up another System with a Supermicro X9DRi-F (which also came at first with bend and broken pins) and two E5-2650L v2 in a RackMatic case (https://www.amazon.de/Cablematic-Rack19-2x3-5-F350mm-RackMatic/dp/B00P04LM9Y). After that my half depth 19" wallmounted rack will be running with 64 virtual cores and up to 640 Gb ram.

That's not just the limit of the rack, but also of my budget, and is more than enough for a VM-Host running a development system, so I do not have to work on the live system all the time...

As you can see, I'm far far away from manufacturing my own plate...

[lEdge]

Okay. Well as long as it doesn't compete with Tesla Dojo powerhouse.

[matthias_r]

Not yet, we're running easy x86 instruction set. Do intel run instruction set statistics on common software and optimize the pipeline for most used instructions or do they get all the same love from the cpu?

Something like... oh, it's add/sub/mul/div/inc/dec and we have 64 native cores and then it reaches a ffs scan and then it goes... oh, there only 16 native cores with that instruction?

[lEdge]

Also you know that the Tesla bunny also has an alternative allowance.