I spoke with Winbond today. This SRAM was discontinued around 2003.
W24257AK-10 is in their system, implying the part was made.
W24512AK-10 is not in their system, whereby W24512AK-12 was the fastest in their system. My W24512AK spec sheet from the 90's only went as low as 15 ns, so there must be a newer spec sheet floating around out there somewhere.
They could not tell me if they ever produced these with yellow lettering, nor do they have the specification sheets. Not a lot can be concluded from this. What I've noticed a lot with just about all high tech companies is that once a part is discontinued, they pretty much abandon and shred all previous knowledge of it.
ISSI claims that these chips were only ever produced as fast as 12 ns and with white lettering. I've been using the IS61C1024-10N chips in my everyday use 486 box and have not had any problems. I'm using the fastest cache settings of 2-1-1-1 at 33 MHz. My uptime is 5 days, 17 hours so far. 3-2-2-2 at 66 MHz also works fine, so even if the cache is a Chinese reprint, it is good enough for the fastest possible 486 system.
Another thought is that the California-based divisions of ISSI and Winbond did not produce, or are not aware of, the 10 ns cache, and perhaps only the Asian divisions of the companies made them for some niche market. I only spoke with the California divisions of these companies. When I inquired about the last production date of the ISSI 15ns chips, I was told 1999, but I have some 1024 ISSI chips with 2002 date codes and the white lettering. Evidence of this was sent to ISSI but I have not received any comment.
EDIT:
Of the 100 '1024' pieces tested, there have been 9 faulty modules.
Of the 22 '257' pieces tested, there has been 1 faulty module.
Of the 36 '512' pieces tested, there have been 0 faulty modules.
EDIT:
Given the 9% failure rate of the '1024' chips, I have a new theory as to their origin. It may be that they were produced by a company other than Winbond or ISSI, perhaps by some discount fab, who figured that putting a brand name of the chip would render them more attractive from a sales perspective.
EDIT:
257, 512, & 1024 chips passed MemTest 4.
The authenticity of this cache is still unknown. Does anyone know how to reliably test DIP cache for 10 ns operation?
Plan your life wisely, you'll be dead before you know it.