Intel 287XL… From 1986? Or 1996?

Many or most readers of this site probably know that most chips (and PCBs) have the date of manufacture stamped on them, almost always indicating the year and week (usually not the actual date) they were made.

Especially with PCBs, there is no standard for whether the year or week comes first. For pre-millenial hardware that’s not a problem, because it’s unambiguous what’s the year and what’s the week. It’s a bit tricker when something like 0403 is stamped on a PCB—is that week 3 of 2004 or week 4 of 2003? In such cases, both might be plausible and one has to look further.

Intel A80502-90

Chips more or less universally use the year/week convention, but some manufacturers label their products with a two-digit year and others only use a single digit. Intel is of course in the latter category. The above image shows a typical vintage Intel chip with a date code stamped on the bottom side of a chip and etched on the top (a 90 MHz Pentium, manufactured in week 15 of 1995). Note that there are actually three date codes, 515 on top, and 501/515 on the bottom of the chip. Intel makes things slightly difficult by often embedding the date code in a longer string, with the three-digit date code starting at the second position (the first might be a digit or a letter).

Most of the time, it’s easy to tell whether a date code like 927 means week 27 of 1979, 1989, 1999, or 2009. But sometimes the answer is not so obvious.

While researching the somewhat unusual Intel 287XL FPU (a 387 math coprocessor in a 287 package), I came across this web page which shows (in the leftmost column) a 287XL with date code 643 and claims the chip was manufactured in 1986. But that’s impossible because although the copyright printed on the chip is 1980 (8087) and 1986 (80387), the 287XL only came on the market in 1990. It is far too difficult to believe that Intel would have secretly started manufacturing the chips 4 years earlier.

But that only leaves one possibility which is only slightly less strange: The 287XL was manufactured towards the end of 1996. Did Intel really still manufacture 287XLs next to Pentium Pros? The 643 date code certainly suggests that…

Unfortunately, while the introduction of a chip is usually easy to pin down, the date when the production was stopped may be impossible to establish. It’s not something that’s usually announced very loudly, since by definition no one cares much about the chip at that point.

Given that the 287XL was an aftermarket upgrade, it’s quite possible that Intel manufactured the chips for longer than one might expect. Most images of 287XLs available on the Internet show chips manufactured between 1990 and 1993 (again, if Intel started the production in 1986, where are the 1987-1989 chips?).

This page shows a 287XL with the same date code and overall the same markings as the one linked to above, but based on the lack of scuff marks, it must be a different specimen. The site likewise incorrectly claims the FPU is from 1986.

Maybe there was a production run of the 287XL in 1996 because someone wanted to upgrade a bunch of old computers? Or the 287XL was still slowly but steadily selling, so Intel decided to make a few more? Either of those explanations is far more likely than Intel making the chips in 1986 and only selling them years later.

This entry was posted in Intel, PC history. Bookmark the permalink.

11 Responses to Intel 287XL… From 1986? Or 1996?

  1. Richard Wells says:

    I don’t see any thing odd about having a 287XL in 1996. The 80486 was produced through 2007. The 80C86 was being sold to NASA up to around 1999. I don’t have any records showing when the last 80286 variants were sold. More relevant would be the last dates for industrial 386s and 387s since the 287XL is just a 387 in a 287 shell. Since Intel isn’t making a special chip, production runs can be quite small and still profitable.

  2. Raúl Gutiérrez Sanz says:

    Upgrading and Repairing PCs, 4th edition (1994): “The original 287 has been discontinued; only the 287XL and XLT are available today.”

    The same text remains in the 6th edition (1996).

    On the 8th edition (1998) there are no references about the 287XL, except the table about power consumption, temperature, number of transistors and introduction date of FPUs.

    It would be interesting to know what 7th edition says.

  3. Michal Necasek says:

    I don’t think the 287XL was sold as an industrial or military chip though, or was it? For embedded chips, very long production runs are of course normal. Guaranteeing at least a 10 year production run is standard in many areas. Intel made embedded 386s until 2007 from what I can tell.

    Anyway, for me a 1996 production date on a 287XL is a bit surprising but entirely plausible. 1986 and 2006 are about equally implausible.

  4. The only reason to keep making anything that long is if someone is willing to pay… So it is possible. But as to who would be buying them? Id imagine they would take 100% delivery of them.

    Unless intel shifted all rules when it came to cramming a 387 in a 287’s socket

  5. Raúl Gutiérrez Sanz says:

    On other chips, units without Intel logo were manufactured before units stamped with the “dropping e” Intel logo. If I follow this unwritten rule, and I sort the dates found searching for 80287XL images, I get:

    Without logo: 023, 025, 027, 030, 032, 039, 044, 045, 103, 111

    With logo: 105, 134, 136, 139, 141, 225, 229, 232, 263*, 328, 643

    (*) Only “Intel”, instead of “Intel 287tmXL” marking

    So, if 0 and 1 are in the 90’s, the 6 could not be years before.

    Another question is why Intel didn’t stamp the logo on the 1st units, while some 386 chips had it since 1989.

    And for me it’s also strange the gap from mid 1993 till the end of 1996.

  6. Raúl Gutiérrez Sanz says:

    The 105 date was on a non-logo chip, sorry for the mistake.

  7. Michal Necasek says:

    Nice detective work! It certainly looks like Intel started producing the 287XL in volume in mid-1990 and kept making them until 1993, and then a single batch(?) again in 1996.

    BTW is it really 263? That doesn’t really make sense…

    The thing with the logos is pretty random. I don’t think there were any real rules.

  8. Raúl Gutiérrez Sanz says:

    Yes, 263. Unless it’s a fake photo…

    On the other hand, Intel chips could have bugs outside, too.

  9. Michal Necasek says:

    That’s an oddly poor quality print. I wonder if it’s a genuine misprint or a fake.

  10. Raúl Gutiérrez Sanz says:

    Fake photo, print or chip, we will never know. I found some new dates for the chips with logo: 137, 313 and 604. This last one reduces the gap, but I still couldn’t find anything starting with 4 or 5.

  11. Richard Wells says:

    It may be possible to ask Intel’s archivist for an image of late production 287XL to compare with other images. I don’t see a contact point to request any form of research by Intel so no chance to ask about final production date.

    I suspect that 2287XLs in the mid-90s were not sold in sufficient numbers to use automated stamping machines but instead got the manual procedure used on engineering samples and thus sometimes look odd.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.