As we mentioned yesterday, Intel has just received a patent for an antioverclocking measure.
Before talking about it, here’s one important detail.
This patent was applied for a long time ago. The application was filed September 29, 1999. Those of you who have been around a while might recall that Intel was quite disturbed at the time by remarkers selling overclocked machines without telling buyers about that.
What was of interest to Intel then may not be of interest to Intel now.
How Does This Works
The general idea is for a set of circuits in the machine to read the rated frequency of the CPU, convert that into a multiple of a fixed clock frequency as generated by an onboard clock signal, then compare it to the multiple of the clock frequency the machine is acually running at. If it’s overclocked, the machine then does a range of not-nice-things.
The fixed frequency Intel cited in the document was 32.968MHz.
So if you had a 2400MHz processor, this set of circuits would read the speed of the CPU, divide that by 32.968, and come up a multiplier of 73.
It would then take the actual frequency of the CPU, and do the same thing, then compare the two. So if you were running the CPU at 2700MHz, this would divide 2700 by 32.968, come up with a multiplier of 82, find out that 82 was more than 73, and say, “Uhh, uhh.”
The patent claims a number of ways this could be done, and a number of ways the machine says, “Uhh, uhh.”
It could be as simple as having a separate clock crystal in the motherboard running at 32.968MHz. However, this could be defeated a number of ways.
What Intel strongly suggested in its patent application was that if they were going to do such a thing, they would build the clock signal into the ICH. An individual (or motherboard company) could replace a clock crystal (and in the earliest days of overclocking, that’s what was done). You don’t replace an ICH.
The patent suggests a number of “Uhh, uhhs.” Some of the milder ones include throttling down the frequency to the proper speed. Some of the nastier ones include throttling down the speed to well below the proper speed (and then keeping it there), keeping a software-readable register on the motherboard to record overclocking attempts (which may be a way for OEMs or resellers to find out if you have been overclocking), or just sending a signal to the power supply to shut down.
Would This Work?
Depends on how serious Intel is about this. If they’re not serious, they would go with a suggested design motherboard manufacturers could ignore or workaround. If they are serious, they build this into the ICH, which make this a whole different story.
Will They Do This?
Well, if they’re hellbent on doing it, we’ll find out next month. The introduction of ICH5 would be an ideal time to implement this.
However, Intel tends to telegraph their moves. Before they take such a measure, they tend to moan and groan about the problem to give themselves a justification for doing it. We haven’t seen nor heard that. If all of a sudden we start hearing it, that’s the warning sign.
One obvious factor deterring Intel is that this would give a boost to AMD, which they don’t want to do.
One less obvious factor that may discourage them from doing this is that this move would give new life to PIV mobos made by companies like SiS and Via, which they don’t want to do, either.
I don’t think Intel wants to help any of these guys out while they have them on the ropes at this time. They have a good chance of knocking Via out of the PIV ring, and they likely feel they have at least a chance to do the same to AMD if they have anything like the problems they had in 2002.
So I don’t think this is imminent.
However, if Via goes and/or AMD falters badly and/or dies (and if they do the first, it’s hard to see how they can avoid the second). Intel will be in a nicely monopolistic position, and could well implement this in ICH6.
Why The Gloom and Doom About AMD?
AMD is not out of the woods yet. In the next fifteen months or so, they’re going to have to 1) have to get desktop Hammer going successfully and 2) migrate successfully to a 90 nanometer process technology.
AMD’s problem is that they’re in a very weakened financial state. They can’t afford any kind of major screwup; they have no safety net.
It’s unclear how badly AMD is going to need desktop Hammer revenues in the end of 2003. It looks like they expect to sell a relative handful of the 1Mb version at very high prices, and a few more of the 256K version at high prices.
They’ll probably be able to do the first, but it’s questionable whether the 256K version of Clawhammer will be very competitive, at least not at a hefty price. Meanwhile, the last of the socket A processors should be in the bargain basement by then.
The real crisis will occur if AMD has problems implementing 90 nanometer technology, and Intel doesn’t. Should that happen, we have 2002 all over again, and AMD can’t survive another year like that.
The coup de grace could be Intel implementing/enabling x86-64 technology, which is still alive if not kicking much. If AMD is yelling “64-bit, 64-bit, 64-bit” and Intel says, “OK, here’s a GHz more of 64-bit, and hyperthreading, and Prescott New Instructions,” and AMD is still stuck at 130nm, then we’re seriously talking about giving up the ghost time.
I’m not saying it’s going to happen; I’m just pointing out that a problem-free 90nm conversion is absolutely critical to AMD’s continued independent existence. They have to do it right, and do it pretty much on time.
If they do, then they should be OK. If not . . . .
P.S. “We have not made an announcement on dropping a nuke on Baghdad.” If the President or the President’s press secretary said this, what do you think this says?
Do you think it says “We’re not going to drop a nuke on Baghdad?” It sure doesn’t. It doesn’t mean “We’re going to drop a nuke on Baghdad,” either. All it says is that we’ve haven’t said such a thing yet; it says nothing about whether we’re going to say something like that in the future.
When Intel was asked about this, that what they basically said. This tells you absolutely nothing you didn’t know already. It says nothing about whether Intel intends to do this or not.
So anyone who says that statement means Intel won’t do this (some have, some haven’t) misread this.
The essence of public relations is to sound like you’re saying something when you aren’t. It’s a good thing to keep in mind.