6502.org

NXP would not be using the Rockwell logo, even if they were producing 65C02s. When Rockwell was broken up it was their Conextant division who continued for a short time producing 65C02s. My recollection is Conextant 65C02 production halted not too long after the breakup. What was sold to NXP was Conextant's broadband media product line. An extensive search of NXP's site failed to turn up any signs of the 65C02 or anything remotely related to it. I believe the 65C02's Rockwell lineage died c. 2000.
It is plausible the seller doesn't know he has counterfeit parts, although the onus is always on the seller to ensure what is claimed in the auction is true. That said, he listed them as new parts, which they clearly are not. That is false advertising and deserves a negative review.
I didn't mean to suggest they are a drop-in. However, the effort needed to adapt the WDC product to an older system is small.

(Side note: it was "Conexant" without a "t" after the "x.") Actually, Conexant was spun off as a completely independent company before the final breakup of Rockwell. They almost immediately changed the logo on the CPUs and stopped using the Rockwell name, losing some brand recognition that I would have thought to be valuable. That kinda makes me agree with BDD that Rockwell-labled chips may not ever been manufactured after about 2000.
Well, it depends on whether you live close enough to a Mouser office (or a retailer that carries these parts, if there are any). My strategy these days is to order several CPUs, each from a different vendor on AliExpress, in the hope that I get the one that I need. So far, it's worked every time and my success rate is far, far above the 1 in 30 or so I need to come out ahead on cost. (Shipping from Mouser costs about the same as 20-25 CPUs from AliExpress, shipped.)

Yep! I always seem to be slipping a "t" in there. Been doing it for years.

NXP make IC's for all sorts of companies and put whatever those said companies want as far as logos and part numbers on them. I was told that more than 50% of their manufacturing is for 3rd parties.

You could be right, but given the verifiable facts of the situation we just don't know.

I have in my enhanced Apple IIe a Rockwell branded 65C02 with a date code 0f 0815. According to what your saying it's a fake. However, it's not. It is a 65C02 and it's working just fine. What I can't get my head around is why some one would go to all the trouble of removing the existing printing on a working 65C02, then re-printing it. That makes absolutely no sense whatsoever. The Chinese are not that stupid. No one is.

However, some company wishing to supply a demand and with the rights to the Rockwell logo contracting NXP to make IC's for them fits exactly with the NXP business model and is quite a reasonable situation.

All in all, we just don't know. Are there fakes? Well, yes I suppose there are. Can we say that every Rockwell branded 65C02 with a date code of later than 0001 is a fake? No, I don't think we know enough to make that determination.

I can think of one reason for re-marking an older part and that is to attempt to sell it as new or as a new pull.

He's not saying it's not a 65C02 (or at least I hope he's not). He's just saying that it's been re-marked.

I suspect this is common. Remember this post? There I show two identically marked "Rockwell R65C02" chips, probably both from the same vendor, but one is CMOS and one is NMOS.
No, you've got it backwards. Given a few thousand "6502" parts pulled off of a large pile of e-waste, it's cheaper to simply re-mark them all and sell them than to test each one (or even look at the labels on them) to see if it should be re-marked and then re-mark only those. It will probably be mostly NMOS chips, with some CMOS ones thrown in, and even some 40-pin chips that aren't 6502s at all, but when you're selling them to vendors for a dollar or less each, who cares? Mark them as 1 MHz parts and most will work well enough for all the hobbyists out there.
That's certainly possible, but doesn't seem likely for chips being sold at retail for under $2 each. Who even has rights to use the Rockwell logo?

Yes, that's the exactly the picture I have in mind. A container load of boards comes in, gets farmed out to a cottage industry of chip-pullers, then chips come back in to be remarked and sold.

Edit: see also this comment elsewhere.

Correct. Even though what Bill has appear to be 65C02s, I still classify them as fake due to the bogus date codes. It may also be the speed grade markings are bogus as well, which is something that can probably determined through empirical testing.

I might have to think about how to design a test rig to estimate a 6502's speed grade. This would be separate from the Fake Finder, which only identifies the model and is not well suited for running at high speed.

I wonder if measuring signal timings could be useful for that. Even at the same clock rate they were apparently different for different speed grades, at least for some manufacturers. The Apple IIe, for example, used a 2MHz part run at 1 MHz specifically because the address bus setup was faster.

That is the approach I'm thinking of: run the CPU on an exercise sequence of some kind, and measure the clock edge to signal transition delays for several likely-critical lines. I think there's a reasonable way to do so without requiring external test equipment, just an on-board indicator. For this test it might not even be necessary to install RAM and ROM, if a good and small set of opcodes and data can be devised for tickling critical timing cases. Obviously the data lines would only be relevant on writes, during Phi2 high - but all the address and control signals should become valid during Phi1.

Having established a speed grade candidate, it should then be possible to run a functionality test at a slightly higher clock speed to verify it. That may want a more conventional SBC design with a variable clock generator; maybe it can be built on the same PCB.

I may need to build such a board in two closely related variants, one for NMOS CPUs (which need TTL levels and have only relatively slow grades) and one for CMOS (which should be tested with CMOS levels and may have relatively fast speed grades). The Fake Finder can be used to initially distinguish the two.

Anyway, once I figure it out, I'll start another thread to describe it.

Here's one of many different ways you might implement a relatively simple variable clock source by taking advantage of the 20-bit NCO with 20-bit increment register in some of the more recent PIC microcontrollers.

I have done that. Mine are all the P4 variety and they will all run at 4Mhz but all seem to quit at just under 5MHz. They don't seem to over clock as nicely as the NMOS ones I have, which vary but some can run at double their rated speed.