I just got 4 of these and all appear to be DOA. I'll deal with the seller/ebay, etc (and probably double-check them in the process).

Tried at 8 (de-turbo) 20 and 25 MHz on my no-name variant of a PCChips M-205 (with Citygate D60 chipset AKA Hedaka D60, CG logo might be fake on mine, but PCChips used CG chipsets earlier on, and there's no 205 mark on my board, even, so might be vanilla markings).

I installed a socket for the oscillator and swapped in a 50 MHz one for the test. The original 20 MHz Harris CPU posts and seems to work fine (no disk drive hooked up currently) at 25 MHz with 60 ns CMOS DRAM installed, though. (I let it run warm for a while, too, and it still posted again; granted these CMOS Harris 286s seem to run fairly cool at high clock speeds ... not like an NMOS Intel 286-12 I nearly burned myself on)

Anyway, since Tiido and some others seem to have gotten working examples, maybe I just got a bad batch. OTOH, if they're remarked, they're also not ideal for what I had planned: trying to push the limits of the D60 chipset (and maybe some others I can get my hands on) beyond 25 MHz, or at least no better than randomly selected original 20 MHz parts would.

Also, most (or all?) real Harris 286 20s aren't engraved, but use white print on a satin/smooth surface finish, not the rough/stippled texture on these. The print is also somewhat larger and covers more of the surface.

Like these 20 MHz ones (apparently from the late 90s)

I'm tempted to try some of these other ones.

Harris part number and looks like the print style and finish of real ones, but missing the Harris logo and name. And from 1991, so long before the 1999 Intersil spinoff company was created, but maybe some were marked like that.

I'm now convinced the chips in the OP are fake. I ended up grabbing a legit 20MHz part

Those CPUs don't even get warm if left in for several minutes at 25 MHz (and thankfully didn't damage the board at least), so they might be totally fake plastic packages: PCChips fake cache chip style, or something else weird is going on. (maybe remarked chips other than 80286s that use PLCC-68 packaging, like 186s, 188s, 68000s, etc, that might not even complete a circuit when installed in the correct orientation)

I somewhat doubt they're remarked Intersil-era parts, and I also doubt even those would run too cool to notice above ambient temps, at least at 5v and 20-25 MHz. (at 8-12 MHz, my Harris 20 MHz chip is almost cold, though)

At least that original CPU seems to run at 25 MHz (and the board itself does), and now I know the board is almost certainly 1990 vintage, going by the date code. (ie if N9050 means 1990, week 50, that would be December of 1990, or August of 1990 if it corresponds to the US Fiscal calendar)

The BIOS posts with a 1990 date, too. (AMI BIOS, Triple D Development Ltd)

Also

maxtherabbit wrote on 2020-02-23, 15:15:

I'm now convinced the chips in the OP are fake. I ended up grabbing a legit 20MHz part

I actually just went for the 5x offer discount there, mostly in hopes of getting at least one that does better than 25 MHz. (or maybe just 25 MHz if my current one turns out to not be stable)

I'm also going to test the waters with that other 25 MHz listing without the Harris branding.

I know there's actual Intersil 25 MHz parts on ebay too, but they're a good deal more expensive and the benefit of domestic US shipping time isn't a deal-maker for me either. (I might try one at some point if my board seems like it wants to do 30 MHz; I don't think I've seen a working 286 system mentioned above 28 MHz before)

Also note, this board (and I think all the PC-Chips style boards, both this chipset and the later Toshiba one used in the M216/219) doesn't appear to have any wait state adjustment, so it's probably fixed at the IBM style 1 wait state configuration (also probably why it came with 80 ns RAM), so higher clock speeds is the only direction to go for added performance. (I know zero wait state 20 MHz boards usually beat 25 MHz 1-wait-state ones in memory-bound tasks too, but for games and benchmarks that are slow-instruction-heavy and register-register activity heavy, the clock speed should be a win)

Now that I know X-Wing (1993 FD version) actually runs on a 286, that's going to be high on my benchmark list. I'm also going to bet it relies more (or entirely) on multiply/divide instructions for its 3D math and not look-up-table enhancements/tricks like Wolfenstein 3D does. (Wolf3D also makes heavy use of arbitrary shifts, which is one of the reasons it can't run on an 8088 unaltered ... not sure about 186/V30 given those include the added instructions)

Wing Commander is probably a good one to use, too, but I know it's finicky about EMS implementations being strictly 100% MS/Intel/Lotus compliant. (I know the Headland scheme is, but not sure about the D60 ... albeit WC can work fine in just 640k if DOS isn't eating too much of it)

X-Wing also runs in plain MCGA/VGA 13h screen mode vs unchained mode that Wolf3D uses, but I'm not sure how wait states impact that.

No problem getting a refund from the seller. They also mentioned checking with their supplier.

I'll hang on to these, but as far as I can tell they're totally dead and/or not even 286s.

I realized the 286 board I was testing them in only has a 1/2 divider for the de-turbo mode, so I was trying those duds at 10 or 12.5 MHz (forget what oscillator I was using at the time, probably 50 MHz), so they might still work at 8 MHz. I'll try that at some point, but given they didn't get warm at all, that doesn't seem to be the likely cause. (maybe they're 286s that got resurfaced and a corner was re-cut in the wrong place, so I've been plugging them in the wrong orientation: I've done that once with a real 286 and it stayed cold and didn't seem harmed, but it might depend on just what wrong-way it's inserted)

The fake date code on there also seems possible to be the date they were resurfaced as Chinese remarked chips tend to have that done for some reason (date of 'remanufacture' rather than a plausible counterfeit date).

These appear to have had the original printed markings scrubbed and the surface etched or retextured (sand blasted maybe) and then engraved with a reasonably convincing Harris logo and markings. If Harris actually used engraving, it'd be more convincing, but the technique avoids the simple tests and signs of other remarking methods and obviously won't be affected by acetone or any solvent short of something that can dissolve epoxy resin.

I'm not sure if acetone will take off the original (real) print, but I don't want to try and ruin one of my good ones.

The original 1990 marked 286-20 that came in my M205 seems to be a pretty good overclocker in itself, which is kind of neat. (though tinkering some more and trying 54 MHz oscillators points to the board tolerating 27 MHz, but the CPU not doing so when even moderately warm, and these plastic casings don't have good enough thermal properties to solve that with cooling, I think ... maybe with a thermoelectric cooler stuck directly to it for sub-ambient cooling, but then there's condensation issues)

Heatsinks and/or fans are probably more useful for keeping hot-running NMOS 286s stable (and industrial 286 boards tend to have little sheet metal heatsinks on them). Back when 286-12s were the fastest grade available, that was probably relevant, though soldered-on NMOS 286-16s (and people unwilling to buy another CPU) would be too. A Siemens 286-16 I tried seemed to go OK at 20 MHz until it warmed up to the too-hot-to-touch range, so a little heatsink or good airflow would probably do a lot for that. (and I've seen at least one very late generation 286 board with an NMOS Siemens 286-16 soldered onto it, definitely one oddball PC_Chips looking 'SARC' marked chipset one, but I think a Headland HT12 one as well)

That other seller on this former listing seems to have had the same problem:
https://www.ebay.com/itm/184105123040?

The photo there looks more promising, and it's possible they had different lots of CPUs in the past (as that other listing also seems to have had), but the current ones seem to be non-functional as 286s at any speed (down to 10 MHz at least, and don't get warm, so I don't think they're 286s at all).

Same implausible lot and date code as the engraved ones (A50004 N1901) but with printed rather than engraved writing. Similar rough surface and smaller logo, too. (the pulse wave also has filled in pulses rather than just lines)
The pins are worn in a way that they look like socket pulls (no obvious signs of de-soldering or re-tinning, just abraded leads). So probably not fake, plastic leaded rectangles ... and also probably not new production failed/rejected parts.

The polished circle marks on the reverse-side corners have manufacturing marks for Malaysia and Indonesia and both those markings and the dimple on the pin-1 edge making look more like they might have originally been 68000s or 68HC000s. Given the faces have probably been resurfaces, I suppose they could also be Toshiba 68HC000s though I think Morotola used plants both in Malaysia and Indonesia.

There aren't enough reverse-side photos of PLLC 68000s to compare. I actually have some ebay purchase PLLC 68000s that look very similar, but I'm not sure those are legit either and had them on hand for a future project, but don't have anything to test them with currently. (though I could rig up a PLLC socket to DIP-64 adapter on a project board, I guess, and test it in a Sega Genesis)

Whatever they are, they don't seem to be remarked 286s (I think dead 286s would still get warm, too) unlike some earlier buyers received.

That first seller (same as the thread starter's ) mentioned contacting their supplier about the issue, so it may be a common problem on that end.

I've got some pics of these, and some of some working 286s (the A50004 J9724 may be a really good remark, but they all work at 20 MHz and work or at least POST at up to 25 MHz, so seems like graded remarking if anything. The logo size, font, printing color and surface texture seem identical to the original: the pulse waveform in the logo looks more rounded at the crests than my 1990/1991 vintage 286s, though). They also don't seem to tolerate 25+ MHz as well as the 3 1990 and 1991 Harris 20 MHz chips I have, but that could just be coincidence or related to a greater demand for genuine 25 MHz parts in the late 90s, at least proportionally to 20 MHz demand. (ie if those are real 1997 parts, they're less conservatively rated at 20 MHz than older 20 MHz parts)

If someone has more experience working with PLLC 68000s, you might recognize them. (I'd think Tiido might be able to)

Just a minute, it seems I forgot to take pictures of the reverse side. (that odd melted/glued 'B' marking was also on some of the other batch of bad 25 MHz 286s)

Edit:

All 4 are marked Malaysia (maybe one of my 68HC000s is marked Indonesia) the 1990, 1991, and 1997 date coded 286s are all marked Korea and have white printed serial numbers (or some sort of code) on the bottom.

The two chips on the far left are the real 286s (one the 'probably not remarked' '97 one from the previous photos, the other is a 1990 one pulled from my PCChips M205 motherboard, not the N9146 one from the previous photos), the 4 on the right are the 25 MHz remarked ebay ones.

And on second look, those 'B' markings almost look like ball point pen markings, or have a similar color to dried black (slightly reddish) ball point pen ink and are smeared/dipped like roller ball markings, but there's way too much ink there to be a normal ball point pen. (looks almost like the marks you get from a leaking/failing ball point pen where the roller ball has gone loose)

Poking around a bit more, the Malaysia marking would be consistent with Motorola manufacturing. (though I'm still not finding much in the way of exact reference photos of the back sides of PLLC 68000s or 68HC000s)

here's some pics of the back sides:

My working Harris 286s all appear to have similar Korea marks and all of these fake/dead (or non-286) parts seem to have Malaysia markings.

I wonder if whatever supplier these Ebay sellers are using happened to do a big lot of 80C286 remarking in 2019 (the N19 in the date code might be a real remarking date or just a mistake attempt at N91, which would be a good, fake old stock Harris marking to use). And the demand for these re-marked parts might have led to remarking of readily available, non-286 parts, and PLLC 68000s seem like they'd be up there in probability. (the people who did get working 286s might have remarked 80C286 12s, 16s, and 20s that happen to overclock well, or at least run at 20 MHz as Tiido's example did)

I haven't done an acetone test right out (tried with 91% isopropanol and acetone-doped ethanol, but that's not so aggressive) but by the limited tests and scratch testing I did, it really looks like solid plastic that's been resurfaced by solvent and/or abrasion. The engraved examples won't have anything added to dissolve away and these white printed ones might be no different from the original print as far as solvent sensitivity goes, so that wouldn't be a good indicator anyway.

The text looks like a good fake, the logos are smaller than the original chips I've seen, and I don't think Harris used engraving on any of their parts. (I'm not sure any 286s used engraved markings ... maybe some of the CLLCs or PGAs, but maybe not even that, though Intel seems to have used a much darker, brownish or gray printing rather than the bright white AMD, Harris, and I think Siemens used)

These latest examples also have more obviously worn contacts (look like socket pulls with modest oxidation) as mentioned above.

The 2 CPUs with white code printed on the back are working Harris 20 MHz 286s from 1990 and 1991. Note the larger pin-1 edge marking dimples on those 2 vs the 4 fake chips. (also the Korea vs Malaysia markings)

OK, I must be going crazy. I got a weird feeling looking at these 4 chips again and decided to double-check the packaging to compare the tracking number with Ebay's info.

Somehow I managed to dig out the old lot of bad chips and think it was a totally different set. I bought 4 in both cases and I could've sworn the original batch was engraved like in the listing's photo, but apparently my memory is broken.

The newer order was confirmed delivered over a week ago, but I must have misplaced it. (I actually didn't remember receiving another package of 25 MHz 286s, and went digging though things when I noticed the delivery confirmation, so maybe something else weird happened)

I did receive some Harris 80C286-16s recently that I didn't remember buying and meant to check on, and that listing I bought the 286-25s from was taken down immediately after my purchase (with notice that the listing had an error in it) so maybe that's where those came from.

I'll have to be more careful about checking the packing label packing slip (if present: though there was no slip with my old order)

Edit: I read over my post from February 23 and don't see any mention of how the examples I got were printed and not engraved. I'm not sure how I missed that detail originally but it's one of the main reasons I made this mix-up now. That seems like the sort of detail I'd have caught when I first opened those. (I also failed to notice the waveform logo being filled in rather than the outline on normal parts and even on the pictures of those engraved ones)

Worse, I did actually post those updates here yesterday before messaging the seller, but did end up messaging the seller before realizing my own mistake, so had to send a retraction on that. Weird luck.

Alright, it turns out the pair of Harris C286-16s were that other 286-25 order. Both went through the Orangeconnex company and got forwarding labels covering the Chinese shipping labels at the same Inwood NY address, so no wonder I was confused.

So I got a substitution without notice and explained such to the seller now.

These two are functioning at least and look like original, unadulterated parts: slightly dirty and oxidized but not worn leads. (look like they were either pulled from sockets a very long time ago or are new old stock that sat exposed to air for a long time)

They both run at 20 MHz too, might go OK at 25, and even post at 27 MHz. Then I pushed my luck and ended up corrupting my hard drive running at 27 MHz for a while. (DOS partition/boot sector got destroyed so the drive wouldn't get recognized by Windows either ... don't have a DOS recovery disk to try and I'm not sure that would've done anything different since scandisk couldn't access it, but it reformatted fine ... I might retire that drive from regular testing use anyway given the age, a few bad sectors, and horrific whine it puts out)

Anyway at least these are real/working chips, and if they're really original (1996 and no date stamp on the other) Harris parts, that also fits my suggestion that some of the remarked 20 and 25 MHz parts that do work were originally 16 MHz ones.

Of course, they also might not operate at 20 MHz in the full temperature range they were officially rated for. (though at 20 MHz they don't really get hot anyway, at 25 it probably gets closer to the 50-60C range depending on airflow and ambient temps, but still not like the NMOS 286s)

I bought three of these supposed 25mhz harris cpus from the same ebay seller (the one using the blue cloth as a background on the picture). Two of the three had bent pins when they arrived and (after straightening the contacts and testing) all three were DOA.

EDIT: and FWIW, after contacting the seller they refunded my money and did not require them to be sent back.