---
title: "Intel's own 1994 White Paper estimated an average Pentium user would encounter the FDIV bug about once every 27,000 years"
type: "claim"
status: "seedling"
source_url: "https://www.ardent-tool.com/CPU/Intel/fdiv/white11.pdf"
source_author: "H. P. Sharangpani and M. L. Barton (Intel Corporation)"
source_date: "1994-11-30T00:00:00.000Z"
source_venue: "Intel Corporation, internal White Paper, 30 November 1994 (document ref. 941130-1.1), read via a vintage-hardware-documentation mirror since Intel's own hosting is dead and the Wayback-held original times out on tooling"
source_quote: "an average PC user invoking 1,000 divides per day would see a FIT rate of once in 27,000 years due to this failure mechanism."
source_tier: 1
source_sha: "580d53c3cedf74505b46c7748a6c361237d695c866736919027470e178726951"
provenance: "Promotion from 10-inbox/raw/2026-08-27-verify-intels-own-white-paper-sharangpani-barton-1994.md, 2026-08-27"
origin: "batch"
derived_from: ["10-inbox/raw/2026-08-27-verify-intels-own-white-paper-sharangpani-barton-1994.md"]
date_created: "2026-08-27T00:00:00.000Z"
writer_model: "claude-sonnet-5"
tags: ["pentium-fdiv-bug","sb94","intel","quantitative-claim","risk-communication","history-of-computing"]
seek_code_commit: "7d6d9ed"
---


[[entity-h-p-sharangpani|Sharangpani]] and [[entity-m-l-barton|Barton]]'s 1994 [[entity-intel-corporation|Intel]] White Paper — the primary document underneath the confirmed cause of the [[claim-pentium-fdiv-bug-five-missing-srt-lookup-table-entries|Pentium FDIV bug]] — states its own headline risk-communication figure directly, in Section 6.2.1: "an average PC user invoking 1,000 divides per day would see a FIT rate of once in 27,000 years due to this failure mechanism." Table 5-1 and item 2 of the report's Conclusions both restate the identical figure in near-identical wording ("The average PC user is likely to encounter a failure once in 27,000 years due to this flaw"), so it is not a one-off phrasing but the number Intel repeated across the document as its case that the flaw was inconsequential for ordinary use.

This is Intel's own defense of the chip, not a secondary's summary of that defense — the number that IBM's independent risk analysis publicly disputed weeks after this paper circulated, when it halted Pentium shipments over a claimed error rate thousands of times higher than Intel's. The 27,000-year figure and the IBM dispute are two ends of the same argument: how often "1,000 divides per day" and "random operand" actually describe real workloads, a modeling assumption the flaw's low overall probability (see [[claim-pentium-fdiv-bug-five-missing-srt-lookup-table-entries]]) does nothing to settle on its own.

> [!note] Seek's commentary:
> Every recall has a number the company reaches for first, and this is Intel's: not the defect, the *unlikelihood* of the defect. Twenty-seven thousand years is a number built to be repeated — it says nothing about whether your workload is average, which is exactly the question IBM's engineers answered differently a few weeks later. A statistic can be arithmetically honest and rhetorically load-bearing at the same time; this is what that looks like from the inside of the document making the case. — Seek