First post, by Am386DX-40
- Rank
- Member
Hey guys, another tricky (and interesting) one.
I've always read that there's absolutely no difference between the IIT 3C87 and the IIT 4C87DLC, and that the 4C87 is just a relabelled 3C87 that changed the name to go with the Cyrix/TI 486DLC line of processors.
Fair enough.
But today, I've found something really curious. I have 2 4C87DLC-40, which seem identical to each other, until I tested them with IEEETEST and then this happened:
Testing the 1st one:
Testing the 2nd one:
Clearly they are not the same internally, so we can throw the "they are all the same" out of the window.
On closer inspection, I notice that the one on the right is deep black, while the one on the left is more brownish-redish (this is the one that has rounding errors in the test). The manufacturing date differs by a few weeks, nothing major. I've been looking at photos everywhere and it seems that the black ones are all "1002FA" and the brownish ones are "1002FB". Maybe that indicates a different factory, maybe the brown ones are in fact relabelled 3C87 while the black ones have a new core with bug corrections and IEEE-754 compliance.
So I'll ask you guys (if you have the time and want to do it) to run the test on your machines and report back which IIT math co you have, and what results you get.
Also, according to "the bible of math coprocessors" (https://dougx.net/gaming/coproc.html) you can check on the accuracy tests that he uses a 3C87 (dated 1992) and he gets another result, different than the ones I posted. So there are "at least" 3 different versions floating around. Intriguing.....