30 years after: A Retrospective into Intel's infamous Pentium FDIV Bug [Colani Restoration Pt. 4]

6 million flawed Pentium CPU, 10 Lawsuits, 475M US$ in direct cost, a stocks value drop, huge media backlash and all in between, Intel, a company in denial: When Prof. Thomas R. Nicely uncovered the Pentium FDIV bug in 1994, it quickly turned out to become one the biggest PR disasters of all time.

Q: How many Intel Engineers does it take to change a light bulb?
A: 1.9628734221

Jokes like these were omnipresent at the time when Intel Pentium's FDIV bug was discovered.
But it was certainly more serious than that, because if your CPU is performing calculations just wrong, the effects can be to the worse.
Imagine my my dad, working as a design engineer for gas turbines, realizing why his calculations were always off precision ...

This video belongs to my current restoration series of the VOBIS/Highscreen Colani PC.

00:00 Intro
00:28 A look back into the Intel Pentium
03:51 Math Prof. Thomas R. Nicely uncovering the Bug!
04:16 The Bug becomes public, and Intel's reaction to it
06:16 Intel's Denial backfires - It's a PR disaster!
06:50 The "FDIV Replacement Program"
09:37 Does my Pentium have the CPU bug?
11:42 Conclusion

Links:

Lot's of Information (landing page in german, linked documents mostly in english):

About the Pentium and the FDIV Bug:

About Intel and Andrew Grove:

More background Information:

Other general technical backgrounds:

Various Media & News Reports:

The Pentium/FDIV Replacement Program:

Computer Chronicles (by @ComputerChroniclesYT):

John Romero recalling on the FDIV Bug (by @tngtech):

Copyright @ 2024 THE PHINTAGE COLLECTOR, Gianpaolo Del Matto. All rights reserved.

Theme Music composed by Abdallah El-Ghannam.

[scottlarson]

I remember a fellow on Usenet showing how the FDIV bug screwed up his data and sent him on a goose chase trying to figure out which of the tens of thousands of data points he had entered was wrong and throwing the results off. I remember him saying something like, "Thank you, Intel, for wasting a week of my academic career."

[davidflint]

I am Pentium of the Borg. Division is Futile, you will be approximated !

[douro]

SPECTRE is considered a fundamental flaw in speculative execution. It could very well be one of the most important computer science discoveries in recent years.

[johnrickard]

I think CPUs with a buggy instruction, like this one and the MOS 6502s with the shift bug, are like mis-stamped coins, and are more valuable as they are now than the normal, unaffected units.

[SteveMaves]

I have several early Pentium machines, but only one with the FDIV bug. I'm very happy to have one and I will certainly be keeping it.

[jirja]

There was an old joke. Do you know why is Pentium faster than 486? Because 486 is slowly calculating, while Pentium is guessing.

[jndominica]

900 dollars in 90s money for just a CPU? Jesus Christ. I could only dream about a used computer back then.

[OzzFan]

I have a Pentium 60 and 66, both with the FDIV bug. I upgraded the 66 to the Pentium OverDrive 133, but I'm saving the flawed CPUs for posterity.

[Amberlynn_Reid]

Back in 1994 I had one of these CPU and experienced this bug

I was updating my Facebook status and I typed "what a lovely day today" and it bugged out and typed "I'm gay as shit" and it posted.

[johncate]

It's worth more as a collector's item than even if Intel replaced it with a modern CPU for you. Most of them were replaced back in the day, and Intel is known to have destroyed the ones sent back to them. So a flawed FDIV P5 Pentium is a rarity now.

[Pdor_figlio_di_Kmer]

I bought in those years a Pentium 60, because of shortage of that processor the retailer gave me a 66Mhz CPU at the same price, but my joy was tempered by discovering to have one of the bugged ones. I must say though that Intel didn't make too many problems, once they decided to replace all the faulty CPUs. They replaced mine in less than two weeks.

[JeordieEH]

Obviously keep a part of history. This is the most known cpu bug of all time. I wish you could have tried to demonstrate and replicate a bugged calculation. I can't imagine intel has any stock, but it would be amusing to call up and see what they offer.

[Danielau]

I live near Santa Clara (where the Computer Chronicles clip at 0:33 was filmed)... Intel still have some offices in Santa Clara, in a slightly different location. I don't think they have as much office space there as they used to, though.

[TheDiveO]

I wonder why people always mention the FDIV bug when Intel already several years before shipped faulty 80386DX-16 CPUs with a faulty 32bit multiplier instruction bug? So Intel already had a record of shipping defective products. Microsoft knew about this problem and had a test for it in Win386 very early in its DOS IPL. Interestingly, this test was unreliable and did not reliably triggered on my 386DX-16. One faulty 386DX-16 according to Intel's own chip marking: use this only in 16bit applications. Needless to say what we drove to the vendor, presented this to him, and he pillaged another system ready for delivery and swapped CPUs. Yes, swapped. Some of the faulty 386DX were very temperature sensitive and the test did not fail when warmed up. However, I found out that misaligned 32bit operand were reliably triggering the Intel bug at any temperate.

[davidmitchell]

Being a computer tech in the 1990's, I have a fond memory for the first Pentium chips. It may be hard to imagine now but computers were quite expensive in the early 90's, having a family to support meant I couldn't afford a computer. That is until the FDIV bug. When we did a warranty replacement, no one seemed to care about the defective CPUs. I was able to send one back to Intel and got a free 66mhz Pentium chip. The foundation to build my first pc.

[knurlgnar]

I had one of these and never got it replaced. Whether because of the FDIV bug or not it would randomly a few times every month give me a divide by zero error at the DOS prompt using simple commands such as DIR. I never had any other issues with that computer aside from that random recurring error.

[hentosama]

there is an old DOS nes emulator (theone with bleeding hand icon) that has a copu bug test and it locks up your machine if youre on this cpu!

[DerekDavis]

30 years later, it is interesting to see where Intel is today.

14th-gen processors that are little different than 13th-gen, and they still consume way too much power. And so many Efficient Cores (E-Cores), while AMD Ryzen are using all Performance Cores, and at much lower wattage.

Intel is in a tough spot in today's market. AMD is wicked competition. And the consumers benefit!

[channelkerr]

"what about Windows ME!?" That was an interesting look at the bug compared to the wiki page i've read. cheers!

[neozeed]

I can still remember the outrage! I was at a bank, and everyone was up in arms as the potential that excel+pentium could be losing us untold amounts of money was far too dangerous! There was patches to disable the FPU and of course people cried as their speedy and expensive pentiums were now terrible. There was an actual scramble for 486 DX machines lol

Also Intel didnt think to much about the branding and why they avoided the next letter, instead opting for 'Pentium PRO' then Pentium II....