What is Intel planning to do?
I was originally going to talk today about how Intel could reduce or even eliminate overclocking through the manipulation of memory synchronization using an integrated memory controller, perhaps speculate that this would be the means through which something like this could be implemented.
But then it dawned on me that the real question isn’t the “what” but the “why.” After all, overclocking didn’t start yesterday, nor has Intel suddenly developed a dislike for it.
Yes, one can recite the conventional wisdom on the matter: unscrupulous people remark CPUs, overclockers pay Intel less for their CPUs than they would otherwise, more CPUs fail when overclocking is part of the picture. More recently, I’ve contended that Intel is creating a premium/luxury brand of CPUs and needs to create bigger distinctions between those and the mass CPUs. And if we find Bloomfields can overclock, but Lynnfields and Havendales won’t, that may be all that’s going on here.
So why do I get this gut feeling the other shoe is going to drop?
One policy Intel has followed since introducing C2D has been to leave more performance off the table than they have in the fast. They could have added at least another speed bin to C2D, they could add a speed bin or two to Penryn, all indications are they could do the same with Nehalems. Yes, you can say that Intel is restricted to certain power envelopes, but if you have a separate line with a separate socket, one can have a separate power envelope, too, and it looks like that is just what is going to happen. This will certainly help Intel release faster processors than they have lately. Once they can do that, then they can monetize that difference, and one way to do that is to restrict overclocking.
How might they go about doing that? The idea is to do the equivalent of putting a frog in a pot of water and bringing it to a boil slowly so the frog doesn’t notice much of a change and jump out. Here’s an educated guess:
Bloomfields will probably not have any real O/C restrictions. Instead, if Intel will put in anti-OC measures at all, it will be done with the Lynnfields and Havendales, just to introduce the concept. If people don’t like that idea, well, Intel will leave you the option of buying a low-end Bloomfield platform instead. If the grumbles are held down to a reasonable level, they’ll get implemented in the luxury brand in a generation or two. Then, people will get tossed a bone with a few more selections in the luxury line, and likely a lower price for the entry-level model. It’s possible these models will allow a little overclocking in that you’ll be able to raise the multiplier up a notch. It’s likely that there will be little overlap between the clock frequencies of the premium and mainstream lines.
Of course, this is something Intel would like to do, which doesn’t mean they’ll be able to do it. It would be harder to implement this if AMD started making competitive processors. It would be harder if people started screaming early, loud and often at the first sign of any anti-OC measures.
After reading this, you may well say, “This is pretty vague and unfocused.” You’re right, but that’s only because the crystal ball isn’t exactly putting out a crystal-clear image. Something is showing up, and it’s ominous-looking, but the picture is about as garbled as that talk about CPUs and memory having synchronized voltages.
But something is brewing.