The Troubled Beginnings of the Pentium

EDITORIAL

Back in 1993, Intel released the successor to the i486, the Pentium. Originally to be called the i586, Intel was facing the problem of rivals using the same naming scheme as them and were unsuccessful in trademarking 586 due to it being a string of numbers. To combat this, Intel approached Legicon Branding, a branding firm behind the likes of Apple’s PowerBook and Adobe’s InDesign, who came up with the name Pentium. Penta being five (with the Pentium being the fifth iteration of x86 CPU), and -ium being the suffix found with many elements on the periodic table. With Intel deeming the processor to be a key element in a PC, the Pentium brand was born.

On 22 March 1993, Intel released the original Pentium processors based on the P5 architecture for Socket 4 and running at 60 MHz and 66 MHz (the 50 MHz model was only available as an engineering sample and never made retail). It was originally meant to launch at 66 MHz exclusively, but thanks to the rush to launch resulting in a combination of the same 800 nm production process as used in the older 486 and the extremely high core voltage of 5.15v (at this stage the 486 has dropped down to 3.3v) it was very difficult to get the chips to run reliably. By reducing the bus speed to 60 MHz and the voltage to 5.0v, the power consumption dropped from 16 watts to 14.6 watts, and a lot of chips that failed validation at 66 MHz could still be sold at a 10% reduction in performance.

That was the end of the line for the original Pentium, and Intel soon moved over to the P54C architecture which launched on 7 March 1994. This new architecture, while still based on P5, including a new socket (Socket 5, and later Socket 7), a change in production process to 600 nm, and dropped the voltage to between 3.135v and 3.60v. This resulted in a massive drop in power consumption (and therefore heat output), and allowed Intel to release the P54C at 90 and 100 MHz. The Pentium 90 had a TDP of only 9 watts, or just over half that of the much slower Pentium 66.

That was not the end of it, however. Thomas Nicely, a professor of mathematics at Lynchburg College, had written code which made heavy use of the FPU (Floating Point Unit) of the Pentium processor. He noticed inconsistencies in the calculations on 13 June 1994, just after adding a Pentium to his collection of computers. Unable to pinpoint the source of the errors, it wasn’t until 19 October 1994 that he finally confirmed it to be the Pentium processor.

The issue was reported to Intel five days later, who admitted to knowing about the problem since May that year. Nicely forwarded the issue to several of his contacts, and it was soon reported in Electronic Engineering TimesCNN. The errata was dubbed the Pentium FDIV Bug, with FDIV being an assembly language mnemonic for Floating-Point Division.

The issue affected certain Pentium CPUs from both the P5 and P54C families clocked at 100 MHz or below, all of which were officially recalled on 20 December 1994 at a cost of almost US$ 500 million to Intel. The affected SKUs have all been removed from the Intel processor specification finder web page. The FDIV bug occurs during floating point operations, where “a programmable logic array with 2,048 cells, of which 1,066 cells should have been populated with one of five values: –2, –1, 0, +1, +2. On the buggy chips, five cells that should have contained the value +2 were missing, instead returning 0.”

This was still not the end, however, with the next popping up in 1997 and affecting everything up to and including the P55C-based Pentium MMX, Pentium Pro and Pentium OverDrive processors. Called the F00F bug, which is shorthand for the hexadecimal address of an offending instruction F0 0F C7 C8, the bug causes the processor to stop functioning altogether until the computer has been hard reset.

As you can see, the Pentium may have been a commercial success, but it was far from a smooth launch. These were not the last errata to affect Intel processors, but they were the last of their kind and stepping stones to what we have today.

FORUM DISCUSSION