Intel's LGA 775: There's no conspiracy, folks

THERE ARE a few other details that made the 'bitch rounds', and cost is by far the most prevalent one of them. The cost of a LGA775 socket is in the $3-5 range compared to about $1 for a s478 socket. While that may seem to be a huge step up, s478 was similar in cost at first, and ratcheted down. In fact all new sockets cost a premium at introduction, and as time goes on, volume ramps, and manufacturing improves, costs go down. LGA775 sockets will not stay up in this price range forever, count on it.

The last problem is the more nebulous one, that LGA775 is a true conspiracy in the 'CIA is beaming microwaves into my skull to make my bed-wetting worse' vein ( here). The theory goes that Intel made these fragile, almost self-destructive sockets for no reason other than to shift returns from them to the motherboard makers. Some numbers bandied about are 20% of the defective CPUs returned to Intel are bent pins, and those 20% will no longer be Intel's problem.

This 'grand plan' does not make sense from two perspectives. The first is that while bent pins may be a large percentage of the defectives returned to Intel, how many in absolute numbers are there? Probably not that many. Now, why would Intel take a small number of bad CPUs and turn it into a large number of bad motherboards? Intel is often accused of playing dirty, but pissing off partners simply because you want to manufacture a problem is not a way to long term corporate health. It also raises prices for the end user, and the mobo manufactures sure would not eat the added cost just because.

But there is a more compelling reason, and it is stronger than tin foil underwear worn shiny side out, that is motherboard marketshare. Who makes the vast majority of chipsets for Intel CPUs? Who is one of the biggest mobo makers out there? Hint: The name starts with the non-trademarked letter I. If Intel were to shift the defective parts from CPUs to mobos, and increase the number of them to screw the mobo makers, they would be one of the companies hurt the worst. Which do you think would cost them more?

Even a cursory look at the facts has this last conspiracy not holding any water. That said, there are enough people swearing that it is all a plan so it will probably never go away. Once a story like this grows legs, it runs away on its own accord and dances around the web. Ah well, it is good ammunition when trying to separate out the nutballs from the clued in at conferences.

My lingering doubts all centre around the question 'why now, why fingers'? The density argument makes a lot of sense, but the why now part didn't have a clearly convincing answer. The new socket has enough advantages in mechanical, thermal and electrical areas that it still ranks a good thing in my book. Most of the advantages are not immediately visible, but the key take home word is headroom. Like all new parts, there is a learning curve and a price premium. From the sound of it, both are well on their way to being non-issues. There is no grand plan to sap your precious bodily fluids, communist or not, just something new and different.

A few things came up in the chat with the Intel engineer that didn't really fit into the scheme of the article, but are interesting and somewhat relevant. First is the Xeon socket, s603 or s604. I asked why they didn't just go with that socket if they needed a few more pins, and the answer was quite simple, they could have, but at a cost. S603/604 is much larger than LGA775, complicating board design, and it has 172 fewer pins so the power distribution would be far inferior. The two main goals of the new socket would not be realised by re-purposing the older one, so that idea went nowhere fast.

The next thing is about future cores, and where they fit into this scheme. If you remember, we told you last February that the P4 core was basically living on borrowed time, and it would be replaced by a P-M cored chip, or more likely 2 of them. It seems like a bit of a waste to define a new socket, get it out there, and then dump it.

Like anything else, when you ask Intel about unannounced products, they hem and haw a little. I asked, and they hemmed and hawed. From that dithering, I got the distinct feeling that the dual core Pentium Ms will fit into LGA775 when they start trickling out of the fabs around this time next year. µ