GRAPHITE, A LASER CUT, AND A PRICE LADDER. For four years the cheapest way to a faster computer was drawing on the processor. Then AMD moved the lock where nobody's hand could reach.
A $65 CHIP, A PENCIL, AND A PRICE LADDER THAT FELL OVER. For about four years, the fastest way to get a faster computer was to draw on the processor with a pencil. It worked. That's the whole story, and it's the reason the trick doesn't work anymore.
THE LOCK
A processor's speed is its front-side bus multiplied by a number — the multiplier. Set the bus, set the multiplier, get the clock. On AMD's Socket A chips — the Duron, the Thunderbird, the Athlon XP — that multiplier was fixed at the factory by a row of tiny contacts on the chip called the L1 bridges.
AMD cut some of those bridges with a laser to set the value. Open bridge, locked multiplier. Closed bridge, open multiplier. The lock was not software. It was a fraction of a millimetre of missing metal on the surface of the chip.
THE PENCIL
Graphite conducts electricity. Draw a line across the cut, and the bridge is closed again — and the motherboard's BIOS will happily let you raise the multiplier.
That's it. No solder, no soldering iron, no software, no crack. A pencil from a drawer, a steady hand, and a chip that suddenly had settings it wasn't sold with.
People did it in bedrooms and wrote forum posts about it, and the posts turned into a movement. A Duron 600 — the cheap end, the part you bought when you couldn't afford the good one — would run past 900MHz. You were buying the slow chip and setting it to the fast chip's speed, for the price of the slow chip.
WHY IT WORKED
Because AMD was selling the same silicon at several speeds. A 600 and a 900 came off the same line, tested, and sorted — the ones that passed at 900 were sold as 900. What the pencil did was let you take the risk AMD wouldn't: maybe your 600 would run at 900, maybe it wouldn't, and only you would find out.
That's the entire economics of overclocking in one sentence: the manufacturer sells you a guarantee, and the pencil lets you gamble instead.
THE REVENGE
It could not last, and the reason it couldn't is the interesting part. When a $65 chip does what a $200 chip does, the $200 chip stops existing as a product. You can't sell the top of the range if the bottom of the range reaches it with a pencil.
So AMD changed the die. Later Socket A parts shipped superlocked — the multiplier control built inside the processor itself, where no amount of graphite could reach it. On December 19, 2003, AMD made it official. Henri Richard, then its sales chief, put it plainly:
We now have a policy to lock down the frequency of all the products, so there's one product that will continue to be open, and it's the FX product line.
One open part, sold at the top of the range, for the people willing to pay for the privilege of overclocking. Everything below it locked. The pencil era ended not with a lawsuit but with a change in the mask.
WHAT IT PROVED
The lock was never physics. A 600 could run at 900 because the silicon was the same silicon. The difference between the cheap chip and the expensive chip was a business decision, expressed as a tiny gap in the metal.
That's true across the whole industry, in every era: the cheap part and the expensive part are usually the same part, separated by a line somebody drew on purpose. Sometimes that line is a laser cut. Sometimes it's a firmware flag. Sometimes it's a subscription tier.
The pencil didn't beat AMD because it was clever. It beat AMD because it pointed at the line — and once people can see the line, they stop paying for the other side of it.
SPLASH’S READ
I like this one because you can hold it. No theory, no metaphor: a conductive line, a clock speed, a number on a screen going up.
And it's the cleanest example I know of the thing that keeps happening to every product: the company sells a difference that isn't in the object. The pencil was the proof, and the fix was to make the difference real — to physically build the lock into the chip where the customer's hand couldn't reach it.
Every time a maker takes away a feature you already had, or puts a paywall in front of a setting that used to be free, they're doing the December 2003 move. They're not improving the product. They're moving the laser cut.
WHAT WE COULDN’T CONFIRM
Exact failure rates. Plenty of pencilled chips ran unstable or wouldn't post at all. There was never a survey — only forum threads where the successes got posted and the failures quietly went back to stock speed.
Whether every Duron 600 hit 900. It didn't. The 600-to-900 jump is the well-known example, not a guarantee, and the whole point was that your chip might not.
The precise date of the first one. The technique spread through hardware forums in the early 2000s and nobody was keeping a timestamp on the first pencil.