Hacker News Reader

Comment by news_to_me | Hacker News Reader

When I built my PC, I learned about this beforehand, so when I heard the "crunch" I wasn't worried. Until the thing wouldn't boot.

I took the CPU out, and sure enough, I'd bent a few of the pins in one corner. Fuck. I then spent the next couple of hours with a tiny magnifying glass and tweezers, painstakingly trying to bend the pins to where it would boot again. 20 min of bending, stick the CPU back in, try to boot... nothing. Several times.

Finally it booted, miraculously. I'm never taking that CPU out again.

View on HN

After a decade doing embedded development for bleeding-edge SoCs and working with devkits with sockets for interchangeable chips... I was kinda shocked when I once again built a gaming PC two years ago...

These are antiquated connectors!

Then again, I hear the sockets we use for the BGA chips I work with cost in the thousands of dollars (because custom-desiged), so I can kind of understand that the razor-thin margins the custom PC industry works with stays with cheap old connectors.

What socket was it? I think all modern sockets are LGA - and even before that, they were ZIF for a long time.

AMD processors still use PGA sockets, but they are ZIF.

Really? the last opteron I saw with a zif pga socket was an opteron 940, which was obsolete a decade ago. I haven't worked with the gaming amd sockets, but I bought a tonne of 2000 series opterons, socket F is LGA, and a few c32 and G34 opterons, all LGA.

My current CPU (AMD FX-8350) is a ZIF PGA Socket (AM3+)

And AM4 ("zen" architecture, not threadripper) is a _very_ similar looking and feeling ZIF socket - total pin count might be different, I'm not sure.

The AM4[1] socket (for consumer processors) is PGA-ZIF.

The TR4[2] (for enthusiast Threadripper processors) and the SP3[3] (for server Epyc processors) sockets, however, are LGA.

[1]: https://en.wikipedia.org/wiki/Socket_AM4 [2]: https://en.wikipedia.org/wiki/Socket_TR4 [3]: https://en.wikipedia.org/wiki/Socket_SP3

>Then again, I hear the sockets we use for the BGA chips I work with cost in the thousands of dollars (because custom-desiged), so I can kind of understand that the razor-thin margins the custom PC industry works with stays with cheap old connectors.

BGA means 'ball-grid array' - meaning it's an array of solder balls, meaning that ouside of testing, it will be soldered to the board... no socket required.

Your sockets for testing are super expensive because they are taking a chip that was never meant to be socketed... and socketing it, which is pretty useful when you are testing the things.

https://en.wikipedia.org/wiki/Ball_grid_array

I recently built a Skylake PC and the CPU doesn't have any pins on it - just pads. The motherboard has springy contacts that the CPU rests against.

You still have to clamp it down worryingly hard, but it doesn't go crunch anymore, and I think it's probably quite hard to damage the contacts.

It's been like that on Intel since at least 2006 with the release of LGA 771. AMD CPUS still have pins though.

The Threadripper I put in last week begs to differ

Now when you hear crunch, it's coming from the flexing motherboard that's mounted on just 2 raised screw holes.

Pro tip: Assemble the cpu/heatsink/ram/nve on the table surface then put the whole thing in the case. No flexing :)

This has been my method for a while, but the last 2 PC's I've build have had fan/power connectors in places that I simply can't get my hand too once the board is installed, leading to me either pulling out case fans, or even CPU coolers, just to connect a 4 pin CPU power lead.

Double check your hand will fit in the thing first!

You can still fuck up the connector on the motherboard, and consequently, the CPU. I'm speaking from experience: I had to replace a motherboard and a CPU. I still have the useless i7-870 as a sort of trophy of my ignorance of those pin-less CPUs (it was my first). Edit: actually, I remember now that the initial motherboard had a manufacturing problem on its socket, but as that was my first, I didn't know, and that led me to try and fail blatantly, killing both the motherboard and the CPU ; I only realized when seeing the new motherboard with a non-busted socket.

Use a credit card! I helped a friend build a PC, but when the parts came in he eagerly attempted to put the CPU in the mobo and then proceeded to crush 1/3 of the pins. We used tweezers to "unmash" the pins, and a credit card to fully align them. Worked like a charm.

Always used a mechanical pencil without the lead in it...

I had to use a credit card to fix one of these builds... :-)

Actually, your credit-card suggestion is useful. My last bent-pin disater was fortunately a mostly-obsolete PGA Pentium 4 CPU, tweezers and careful, repeated re-insertions into the socket sorted it out. Given that the CPU could be had for about $5, it wasn't worth the effort, but I was trying to help a friend and I felt stupid for dropping the CPU.

Oh God you've brought back some memories from when I assembled my PC.

I also bent a pin. Now, the pins in my case were on the motherboard (my understanding is that this used to be reversed, and they used to be on the CPU. Frankly, I am quite happy that they're on the mobo, since that's the cheaper part) and IIRC they don't point straight up (they're slanted, and just kinda graze these contact points on the bottom of the CPU).

It turns out that, for my particular motherboard, ASRock's manual's instructions and images describe a different piece of hardware than the one I actually had. Same model number though, IIRC. As best I can tell, at some point, there was some sort of hardware change, but like so much tech, the docs were never updated. It was only after scrutinizing the manual very closely that I discovered the error, but it was already too late and the damage was done.

I contacted ASRock about it, sent them images detailing the differences and obviously laying out the problem. The representative essentially just sent me the same bad instructions back. Finally caved and had it repaired, which they did not cover under warranty (which I consider a violation of the warranty). (I also finally figured out the correct procedure, which I had to figure out in order to send the motherboard back to them for repairs, as they claimed they would refuse it if I shipped it with the pins exposed, which is extremely frustrating when they also can't tell you how to actually accomplish that.)

The next motherboard I buy will not be from them.

Finally got it all back, and now, equipped with the knowledge of how not to screw it up, got it all assembled. Still didn't boot. Got a friend with a working machine to swap parts with until we could isolate the failure; turned out to be our first guess, the PSU. But it was somewhat surprising at the time, as we thought the PSU was good. You can short two of the leads on the main plug from a PSU with a paper clip, and that is essentially the signal to "turn on" and start supplying power. We did this, and it would start right up, but it never started when hooked up to the motherboard itself. Beyond that, IDK. (and since you have to short the two contacts w/ a paper clip, the plug isn't attached to the mobo at that time, so you only really power things like the DVD drive, the fans, and the HDD, which mostly then go into loops b/c they can't find the motherboard.)

Finally, it booted. I'm never taking that CPU out again, either.

Warranties not covering bent pins is standard. You damage it, your fault.

Reply on news.ycombinator.com