I love the way the rumor mill churns out stories with the most ridiculous premise sometimes. At times its the teasing of any mention of Half-Life 3, others its the idea that AMD is going bankrupt and bowing out of the market. This time though, the mill spat out a rumour so unbelievable that there must be some grain of truth in it. Its so weird and vague and has such far-reaching implications that its impossible to discern how a move like this would affect the PC industry. You literally cannot make this up.
The rumour? That Intel could stop offering socketed desktop chips in 2014.
Now I know you’re scratching your head, or your chin or even your nipple while you’re trying to wrap this around your head. Socketed chips have been around for ages and offers such vast flexibility that its been the only option for desktops and servers for the past four decades. The idea of taking out a chip and replacing it with a faster, leaner and more efficient successor using the same socket was a great time-saver in the past and reduced system complexity and the cost of maintaining systems over time. As time wore on, some designs lasted longer than others – LGA 775 took six years to die, AM2 lasted four and socket AM3+ and LGA1155 are going strong for three and two years respectively. Socketed processors are a godsend to enthusiasts who like to invest in a good board and upgrade the CPU as time goes by.
Now, the rumour was first put out into the internet by Semi Accurate, a site normally not known for its journalistic integrity. It ranks close to Fudzilla in my books for most “WTF?” moments when I’m looking for information on the net and is almost exclusively pro-AMD and anti-establishment, at least as much as they can be without their entire site just looking like a good satire show. They are the resident fear-mongerers and paint every story with their own ass-backwards look at the details, sometimes twisting them to make things look more ridiculous.
Even their logo insinuates that its not always on target with its content. But sadly, like a thousand monkeys on typewriters, they eventually hit on something that makes sense. They’re not dumb enough to not shout this fact to all and sundry, so I’ve been avoiding writing about this until I had seen the reactions of other writers on the net.
The rumour suggests that Haswell will be Intel’s last socketed processor for 2013. Broadwell, a “Tick” in Intel’s architecture strategy with a bump down to the 14nm process, promises huge power improvements and better efficiency. It also, according to Semi-Accurate’s Charlie Demerjian, will eschew a Landing Grid Array, opting to rather move to the commonly-found Ball Grid Array used in laptops and All-In-One solutions like the ITX versions of Intel’s Atom and AMD’s Brazos chips. Both are soldered directly onto the board and can’t be replaced on their own.
Here’s something you may have already guessed: BGA makes Intel and its partners a lot of profit. Go on, find a laptop with a broken board and check out how much a new one is. I’m guessing around R7000 for a mid-range laptop, am I right? That’s a huge profit margin for the spares market and its something that’s been giving Intel and its partners a steady cash flow for years. Because you can’t swap out the CPU, you have to replace both the board and CPU if one or the other dies. Ah, lightbulb! You can see where I’m going with this train of thought.
Laptop users have accepted these compromises because of the necessary requirements to optimise a particular platform or hardware set for maximum efficiency and low heat generation. BGA has been deployed in that segment for years and its well-known that Intel doesn’t actually design desktop chips – they only make laptop chips and server processor, choosing to make up the desktop line by mixing existing designs with a higher power envelope, more or less cache and adding in support for LGA sockets.
Considering that it would be commercial suicide for Intel to make their server market deal with the BGA socket, it follows that they’d continue to offer a socketed option to quad-core buyers in the high-end, LGA2011-ish space, with the low-end market stuck to picking and choosing a board and CPU combo stuck together. Now on the one hand it has some good implications for buyers – worrying about compatibility is now a thing of the past and system complexity is reduced. In addition, Intel no longer has to fight to have buyers see reason behind their socket changes – they can do this as they wish to improve thermal and efficiency numbers as they see fit. Granted, that’s not a lot of good reasons, but its all that makes sense for now and I’m still looking at this announcement with a pinch of salt.
How would this benefit businesses? Let me tell you a secret: back in college my friends and I swapped CPUs from the more powerful desktops in our class into our chassis on the other end of the class so that our PCs, put together from salvaged parts, would run games better. When we were in the computer room screwing around with virtualised Server 2008 installs, I had to hunt down the one computer in the room that I knew was a Core 2 Duo-powered tower and swapped it to where I was sitting. The rest of my friends buggered around with Pentium D chips. I even later swapped the stock heatsink for a much better one and overclocked my chip to 3.4GHz. Later we highlighted to the system admin, when he asked for our input on the computers in the network, that standardising hardware was difficult because you had three generations of CPUs under the same roof. There were spare parts, but they were always for the wrong system.
For a business, having every machine run in exactly the same way with the same hardware as a whole lot of other systems on the network makes administration easier because you have, more or less, a stable operating environment and less chance for some tomfoolery by people like me. Since most corporations simply replace their PCs outright after three or four years and replace them wholesale, this eases load on the network admins figuring out which standard configuration to settle on or which will be better supported.
Are those the only benefits to a Ball Grid Array-based desktop? I’m sure there must be more. Am I missing something? There is better efficiency and lower heat generation in the package, but that’s always been Intel’s staple selling point.
I think this will take a lot of planning and reorganising of Intel’s distribution model. Suppliers will no longer have to stock low-end CPUs and boards separately, but this could lead to more confusion and longer price lists. Perhaps Intel’s board partners will have to offer a good range of motherboards with a large variation of pairings with desktop chips – you could land up with five different variations of a single motherboard SKU. It would force Intel to clean up its portfolio and only offer one or two versions of the Celeron, Pentium and Core i3 and i5 chips and that would be devastating to their bottom line.
If this happens and if Broadwell is indeed BGA-only, not only will the consumer’s choices be vastly limited, but we’ll see the beginning of the end for the Intel ATX desktop as we know it. AMD may even follow suit, seeing as how a BGA platform would work incredibly well for them in the desktop and consumer market.
Still, we don’t have to worry now, do we? Its only happening in 2014 and the world supposedly ends in around 20 days. Yeah, lets just go play Far Cry 3.
Discuss this in the forums: Linky