486sx

A big messup by Intel, when they decided they'd say they couldn't put a FPU on a processor, but did anyways and just kept it disabled in the 486SX series, but enabled in the 486DX series to keep cost down (same processor design basically). At least that's how I understand it.

A poor move at the time, but in retrospect, it panned out. The original 25 and 33 mhz SX's had a present but disabled FPU, and they cost Intel more to produce than the 486DX. But it had the upside of making 'wimpy' processors acceptable to the public -- without which the Celeron and even the DragonBall would not have seen the popularity they have.

The other positive side effect of 'underpowered' processors becoming 'OK' was that it opened the door for later Processor manufacturers, notably AMD and Cyrix, to make their own FPU-less processors for low power applications and embedded tools. Now-a-days (end of the Year 2000) you can pick up an AMD Elan 400, which is a low-low-low power 486/66 SX for less then 15 USD each (in lots of one thousand).

Unlike the SX convention with Intel's 386SX, which had a 32-bit internal path but 16-bit external path, the 486SX was 32-bits all around, which helped balance its lack of strong math performance.

The 486 SX is one example of a typical Intel marketing ploy: arificially disable or dumb down a CPU for the sole purpose entering another price point.

It started with the 386 SX. This was just a 386 DX with its external bus stepped down to 16 bits (unlike 32 bits for the 386 DX). The 486 SX is exactly the same as the 486 DX, but with the FPU disabled. Several years later, the first Celerons arrived, which were simply Pentium IIs with no L2 cache.

Today, it still goes on. The newest Celerons, with 128 Kb of on-die L2, are the exact same silicon as the current Pentium !!!s, but with half of the L2 cache disabled.

Actually, what I heard was that the 486sx was a 486dx chip with a bad FPU and that, during testing the chip people said, hey, instead of trashing the chip, lets add some chip kung-fu to the chip so that we can disable the fpu portion and sell it as a cheaper chip, thus lowering the apparant failure cost of the 486 models.

I've heard this as being the reason for lower speed pentiums as well though, so who knows if this is true.