Intel LGA775 sockets bend INQ's brain

AS A BRIEF FOLLOWUP to the LGA775 articles (parts one, two, three and four), I came across a few minor things that I thought were worth reporting.

The first one was the impending apocalypse over the new sockets. As you remember, anything stronger than a warm breeze was reported to lead to bent pins and consequently dead mobos. People said this was all a conspiracy to do one thing or another, and Intel was a bad bad company for doing it.

Fast forward half a year or so, and I can say that none of the mobo vendors I have talked to since then have complained about piles of dead sockets littering the returns department of their offices. Some may complain about how they are still massively outsold by their 478 predecessors, but not many bitch about the sockets shuffling off this mortal coil.

The more interesting part is the heatsink and fan (HSF) mounting, and its effect on the boards. As you know, one of the main benefits of the LGA775 socket was that the socket provided most of the clamping force on the CPU, so the HSF needed only to apply enough force to make the thermal connection.

It worked, and worked well. The first one I used was on the Abit AG8 I wrote about the other day. The heatsink on the P4/3.6 clipped on so easily I was wondering what I screwed up. I just could not believe it went on that easily, the hardest part was lining up the holes right.

Then I got the Intel i925XE reference board, and it went click, shove, grunt, urgh, shove, shove, grunt, swear, swear louder, shove, click. Only two more to go. #*&^$@! Needless to say it was a bitch to get the thing on, I had to really apply pressure, and if it had been in a case, it would have been damn near impossible.

Flash back to the LGA775 articles. Some of the engineers told me there were specs built into the mobo to account for stress. They could, and more to the point, should, flex a bit, and that was normal. Fair enough, I know how these things work, and it was a design choice, not anything to worry about.

That said, the two boards, the AG8 and the Intel i925XE show the extremes the boards can take. Before you look at the pictures below, let me say in no uncertain terms that both boards work perfectly, and I have removed and reapplied the HSF a bunch of times with no problems. There is nothing to worry about here, it just looks cool.

First we have the AG8, and it is sitting on top of a carpenter's level, as are all the pictures, so the surface is straight even though the pictures may be at a slight angle. Notice that the CPU is dead centre, and the board sits almost flush all the way. The left side is raised a bit, and the right a bit more.

A closer look at the left side shows the corner is almost flush, while the right hand side in the picture below that shows it is about 1-2 motherboard thicknesses from level. Nothing to worry about, and no problems. This is about what I would expect.

Then we go to the Intel board, and look at it sitting on the same level. The CPU is in the dead centre once again, but here the board rises off the level in less than the width of the HSF.

When you get to the corners, the board is way way off the level, the HSF is obviously putting a lot of stress on the board. The left side is a bad picture, but it is about three to four mobo thicknesses off the level, but it is hard to tell with the shadows. On the right, it is about the same as the Abit board, a couple of mobos off.

Overall, there is nothing to worry about at all, it simply is what it is. With a sample size of one, there is not enough data to make any conclusions, and even if there was, both boards worked flawlessly. It could be a single bad sample, engineered differences, or nothing at all. The Intel board, to the naked eye looked like it was bending much more, and I just found it interesting.

The last one is the pads on the CPU and the pins that connect to them. If you think back to the S478 days, the pins fit into the hole, and that was that. If it didn't fit in, it was immediately obvious. If you pushed down hard when it didn't want to go, new CPU time. It was a round holes, round pegs, it worked or it did not.

With LGA775, there is a little more play. The pins guide themselves to a flat pad, and there is no locking mechanism. It is a lot more nebulous a mating than the old pin and socket.

A picture by Scott Wasson of The Tech Report first clued me in. Intel didn't really have a comment on it when I asked, but sadly, I don't interpret this as a grand conspiracy, just another in a line of 'that is how it is' engineering choices.

What I am babbling about is the mating of the socket with the chip, and how those 775 little fingers have a bit of play. Look at the picture below, it is a closeup of a P4/3.6.

Due to a combination of miserable camera skills and the limits of the site, I can't put in a detailed enough picture, but you can still somewhat see how the connection points vary on the pads. The bright spots are where the fingers made contact with the pads. On most of the chip, the dot is in the dead center of the pad, in others it is noticeably off center. Others have multiple contact points, I have seen up to four on a single pad.

Again, it works, and I have not heard of this causing any problems in the field. I would put it, and the bending, down to engineering curiosities of the new socket. LGA775 did what it should, and works just fine. I saw these things and thought they were interesting. Six months later, LGA775 continues to surprise me. µ