Fixing a broken pin on a H11DSi Supermicro motherboard.

[erock]

Hi All,

I am in the process of building 4 dual socket machines using the supermicro H11DSi motherboards and Epyc 7f52 processors. These machine are truly amazing especially considering that the price of used 7f52’s from reputable dealers on eBay is around $330. I have finished 3 of these machines with 256GB of 3200 ecc ram and their performance is comparable to or better than my TR 3970x machine for the types of multithreaded and MPI parallel calculation I do. However, on the last machine I decided to buy a processor from a cheaper seller with less volume and lower ratings. This cpu did not fit in the socket properly and damaged some of the pins in the cpu 1 slot.

I was able to bend back most of the pins but one broke. I can get the machine to post for 1 cpu but the second socket does not work. How would one go about getting a pin repaired? Also, is it more economic to buy a new board at around $350?

 

[BlueFox]

With a broken pin like that (one being required for operation), there's not really any easy fix. Socket replacement isn't something you could do yourself either. You can see how much Supermicro will charge for it, but otherwise you're looking at a new motherboard.

 

[rtech]

You should post pictures so others might recognize this defect and avoid putting CPU in socket.

 

[RolloZ170]

This cpu did not fit in the socket properly and damaged some of the pins in the cpu 1 slot.

impossible.
do you used the SP3 Carrier ?

 

[erock]

impossible.
do you used the SP3 Carrier ?

Yes, it was an SP3 carrier as far as I can tell.

I successfully installed around 12 Epyc CPUs so this was a surprise to me as well. Here were the key observations (maybe you can find a different explanation):

1. I successfully installed a processor into the cpu 1 slot, installed some RAM sticks and got the machine to post.
2. I then inserted a processor into the cpu2 rail frame until I heard a click and made sure that carrier was properly inserted into the frame.
3. When I lowered the spring loaded force frame I noticed that it did not fit as easily as the other 12 I installed. I also notice a bit more wiggle room of the cpu in the carrier.
4. I ignored my intuition (big mistake) and proceeded to screw the force frame into the socket and installed more RAM sticks.
5. I then tried to get the machine to post which did not occur.
6. I then removed CPU 2 and got POST to work.
7. My curiosity then got the better of me and I uninstalled the cpu from socket 2 and installed it into socket 1. I observed that the frame again did not fit as well but proceeded to install since the tray looked identical to the other. NO POST!
8. I then reinstalled the good cpu back into socket 1, NO POST!!!
9. I then inspected the pins in each socket and noticed the pins along the same edge were damaged along with the black plastic frame.
10. After bending the pins back into position and breaking one pin I was able to get the good cpu to post with all RAM slots registering.

I already returned the bad cpu to the seller and got a full refund so I can’t post picks. I will post some pics of the damaged mono later today.

 

[RolloZ170]

Yes, it was an SP3 carrier as far as I can tell.

I successfully installed around 12 Epyc CPUs so this was a surprise to me as well. Here were the key observations (maybe you can find a different explanation):

1. I successfully installed a processed into the cpu 1 slot, installed some RAM sticks and got the machine to post.
2. I then inserted a processor into the cpu2 rail frame until I heard a click and made sure that carrier was properly inserted into the frame.
3. When I lowered the spring loaded force frame I noticed that it did not fit as easily as the other 12 I installed. I also notice a bit more wiggle room of the cpu in the carrier.
4. I ignored my intuition (big mistake) and proceeded to screw the force frame into the socket and installed more RAM sticks.
5. I then tried to get the machine to post which did not occur.
6. I then removed CPU 2 and got POST to work.
7. My curiosity then got the better of me and I uninstalled the cpu from socket 2 and installed it into socket 1. I observed that the frame again did not fit as well but proceeded to install since the tray looked identical to the other. NO POST!
8. I then reinstalled the good cpu back into socket 1, NO POST!!!
9. I then inspected the pins in each socket and noticed the pins along the same edge were damaged along with the black plastic frame.
10. After bending the pins back into position and breaking one pin I was able to get the good cpu to post with all RAM slots registering.

I already returned the bad cpu to the seller and got a full refund so I can’t post picks. I will post some pics of the damaged mono later today.

ok, bad carrier, they loose hooks when switching from cpu to cpu.
look here what can happen if the Carrier has broken hooks(sorry german, and threadripper but it is same carrier)

 

[erock]

Here are pics of my disaster (turns out I took photos before sending the cpu back). Note that the cpu tray seems visually OK but the edges of both sockets were damaged in the same location after the bad cpu was installed.The force frame fit fine with the other good cpu that did post.

Note also I had to reshape the damaged black plastic frame, remove alot of deformed black material along the edges, and reshape the bent pins to get the cpu 1 slot to POST with the good cpu.... $300 bucks down the drain since I don't think the second cpu slot works now (but I need to do more work on this to confirm).

Let me know if you see any clues in the photos I should have picked up on.

The video above sums up very well my reaction to this disaster.

 

[alex_stief]

Looks like part of the pin is still there. Someone who is really good with a soldering iron could try to attach an extension to that.
On the other hand, that is a risky operation. And success is far from guaranteed. I would probably cut my losses and get another board.

 

[RolloZ170]

Let me know if you see any clues in the photos I should have picked up on.

match the triangle of the EPYC with the Carrier please.

 

[Kizune]

Your CPU inserted in the carrier in the wrong orientation. The golden triangle on the CPU should be oriented towards the handle. No wonder your socket is broken - with mounting screws you forced the CPU into it the wrong way.

 

[rgysi]

I wonder who rips all these CPU's from the carrier and then reassemble them wrong or sell the carrier seperately?
I had a similiar issue, but luckily I send back the CPU because it was supposed to be new and came without a carrier frame.

Just a simple google picture search for sp3 carrier frame and you see people selling carrier frames that have been ripped of CPUs and CPU installed wrongly in carrier frames!

 

[erock]

Your CPU inserted in the carrier in the wrong orientation. The golden triangle on the CPU should be oriented towards the handle. No wonder your socket is broken - with mounting screws you forced the CPU into it the wrong way.

OMG! Yes, this makes sense especially given the shape of the edges of the cpu. I should have checked this, but the cpu was described as new so my guard was down. Like I said before I was taking a risk with a not so great seller on eBay. This was clearly used, then cleaned and put back in the tray the wrong way. I should have paid a little more from a good seller and inspected the product before installation. Lesson learned!

 

[erock]

Update: I invested in a 10x hand lens and was able to do a better job bending the pins back into place. I installed two 7302 CPUs with correctly oriented trays and the motherboard is working fine with 2 cpu’s. I ran a rasdamon/stress-ng memory test with no errors reported. So the broken pin appears to not be critical for my current setup.

 

[Johnnyooo]

Update: I invested in a 10x hand lens and was able to do a better job bending the pins back into place. I installed two 7302 CPUs with correctly oriented trays and the motherboard is working fine with 2 cpu’s. I ran a rasdamon/stress-ng memory test with no errors reported. So the broken pin appears to not be critical for my current setup.

Man. I wish I could have pulled this off in my situation. I'm truly impressed. I only had a couple pins bent, but they were under a tiny pile of crushed plastic and I never managed to get the plastic lifted off so I couldn't get the pins straightened out. I tried to get an xacto knife underneath, but only managed to get more pins bent. Lesson? Always check orientation of the CPU in the carrier, and always buy from someone with integrity who's willing to fess up to a mistake and help you out (I did). But this mobo was fried by a (at least to me) simple mistake. It actually took remarkably little effort/jiggling to get the cpu holder screws engaged and 1>2>3 screwed down to 1.4 nm, crushing those little notches. I'll never make that mistake again.

 

[erock]

Man. I wish I could have pulled this off in my situation. I'm truly impressed. I only had a couple pins bent, but they were under a tiny pile of crushed plastic and I never managed to get the plastic lifted off so I couldn't get the pins straightened out. I tried to get an xacto knife underneath, but only managed to get more pins bent. Lesson? Always check orientation of the CPU in the carrier, and always buy from someone with integrity who's willing to fess up to a mistake and help you out (I did). But this mobo was fried by a (at least to me) simple mistake. It actually took remarkably little effort/jiggling to get the cpu holder screws engaged and 1>2>3 screwed down to 1.4 nm, crushing those little notches. I'll never make that mistake again.
View attachment 45553

I feel your pain! I got lucky. I bet I would have issues if I installed one of the parts with more cores. My machine is still running with no issues so I think I will stick with the dual 7302's. I learned my lesson as well.