News It’s been 30 years since Intel’s infamous Pentium FDIV bug reared its ugly head – a math bug caused Intel’s first CPU recall

[Admin]

2024 marks the 30th anniversary since Intel announced a math glitch on some of its Pentium CPUs.

It’s been 30 years since Intel’s infamous Pentium FDIV bug reared its ugly head – a math bug caused Intel’s first CPU recall : Read more

 

[dutty handz]

"was justn't"

God no, "just wasn't"

 

[setx]

However, the algorithm required the Pentium to set aside an array of 2,048 cells (like those in a modern SSD)

Why are you writing nonsense again? Those calls are not like the ones in SSD and couldn't be rewritten to correct values – that's the whole point of recall instead of microcode update.

 

[Ulisses Bebianno de Mello]

Cool story but wrong picture.
The Pentium 200 in the picture wasn't among the affected batch. The bug impacted earlier models only.

From Wikipedia:

"The FDIV bug affects the 60 and 66 MHz Pentium P5 800 in stepping levels prior to D1, and the 75, 90, and 100 MHz Pentium P54C 600 in steppings prior to B5"

 

[ggeeoorrggee]

Fun story about the FDIV bug:

In 1994, I was enlisted in the USAF as a programmer and a group of us were tasked with creating a successor to a very old PDP-11 based system. We ordered 12 rackmount systems from one of the big white box guys back then that did systems for the Pentagon. Even split between 486DX4-100 and Pentium P5 60MHz units.

Our Lt — who incidentally introduced us all to a little gem called Doom which sent my life down a gaming path — found the calculations to test our procs and we verified that we had 6 flawed procs. I was tasked with contacting our sales rep at the pc builder. He told me that due to the size of the recall and how many Pentiums Intel could crank out in those days that it would likely be a few weeks before we could get our replacements.

He then asked what the computers were used for. I should, at this point, mention that the air base on which I was stationed was Andrews AFB which flies one particularly important mission.

“Well, they’re for the ground radio to telephone switching system for Air Force One and priority mission flight communications around the world.” Or something to that effect, I said.

The expression “pregnant pause” was made for this very moment.

“I can be on a plane in 2 hours”, he replied. God as my witness we had a 10 minute debate (with the guy on hold) about doing it and how much trouble we might cause (or get into).

Since the system wasn’t actually on line, we had to let him off easy. We did get our replacement tray of chips in just over a week though. Strangely, the speed measuring software reported the new procs as 63MHz. I remember the speed because we were trying to figure out why the replacements were running the non-FLOP benchmark software faster, so we checked reported proc speed. I have never found any documentation for that speed in Socket 4. Only 60 and 66.

 

[kerric]

Tom was my calculus professor at LC, great teacher and amazing intellect.