While I might understand why you use the word "fake" in this entire video, the evidence suggests you have either a genuine silicon based co-processor or a compatible. I read a few years ago that the Chinese and other developing nations were taking apart our electronic scrap and repurposing the essential bits of technology. Some were going as far as grinding the surface of components and laser etching their own part numbers on them. It would be far easier (and more of a logical conclusion to draw) that this "fake" part is something that has been reworked rather than completely engineered to be a knockoff.

the 1-800 number for the intel program is still, in fact, the intel phone support.

To fix the Environment Test at 4:36, set the DOS environment variable before running. SET 87=YES

Am I hallucinating, or are there 4 seconds of an ancient pre-Linus-branded Linus Tech Tips at the end of this video?

The sounds of the IBM PC starting up as well as the drive chirping... got me right in the feels. Great video.

14:24 Yeah that was really unexpected

Intel 8087 coprocessors were fabricated in two variants: one with ceramic side-brazed DIP (CerDIP) and one in hermetic DIP (PDIP). All models of the 8087 had a 40-pin DIP package and operated on 5 volts, consuming around 2.4 watts. Unlike later Intel coprocessors, the 8087 had to run at the same clock speed as the main processor. Suffixes on the part number identified the clock speed: 8087 clock speeds Model number Frequency Intel 8087 5 MHz Intel 8087-1 10 MHz Intel 8087-2 8 MHz Intel 8087-3 4 MHz Intel 8087-6 6 MHz They were designed to operate in the following temperature ranges: C, D, QC and QD prefixes: 0 °C to +70 °C (commercial use). LC, LD, TC and TD prefixes: −40 °C to +85 °C (industrial use). MC and MD prefixes: −55 °C to +125 °C (military use). I'm willing to bet these "reproduction 8087"s are actually rebranded chips made for industrial or commercial use.

in the mid 1980s I bought a NEC V20 chip for $20 to “upgrade” the 8088 in my XT clone. They mailed me an Intel 80287 chip by mistake which I couldn’t use in my motherboard so I saved it for when I could afford to build a 286 clone. When I finally did build a 286 clone I happily inserted the 287 chip and … nothing was faster. I later installed AutoCAD 2.6 which did run faster but then got a board that had sockets for 386 and 486. After using a 386-50 for a year I finally bought an 80486 DX-2 66 MHz chip (for $500!) which at the time was blisteringly fast. But as time went on computers got faster and cheaper.

I don't think the "fake" one is fake at all, but a refurb. Notice that it is ceramic, which I don't think they would bother to do if it was just straight up fake. (The "real" one is also ceramic, but of a different type. The "fake" one is production type ceramic DIP with two ceramic plates bonded with epoxy. Whereas the "real" one is a solid block of ceramic with the leads integrated and a pocket in the middle where the die is.) I bet that if you were to rub the top with a q-tip and acetone, the text would come off, together with a layer of black varnish. Also of note, what does the "fake" one look like on the bottom?

I've bought regular retail CPUs from chinese sellers before (mainly because they tend to be cheaper, even if no questionable stuff happens) and in some cases received engineering or qualification samples instead. Sometimes (especially with really old CPUs) they tried to relabel them (something like selling a DX4-80 rebadged as a DX4-100 that barely works at that speed), other times they didn't even bother. One very interesting example is what was sold as an i7-2600, but was actually a qualification sample (!) that was in such early stages it didn't even know what it is yet (model number when in a system comes up as 0000) and the specifications of it do not match any released Sandy Bridge i5 or i7. It's a Quadcore with HT (which would indicate an i7), but it only has 6MB Cache (like an i5) rather than 8MB of an i7. The supported feature set is also all over the place. Overall it seems like it's still in "we'll decide later what this is going to be" mode, which must mean it was fairly far away from the release date of the final CPUs (ie January 9th, 2011). It's core stepping seems to indicate that as well (C0, all release i5-2xxx and i7-2xxx were D2). What puzzles me is that it appears to work perfectly fine so far, but some software completely brainfarts because it can't determine what the CPU is lol. This includes Windows 10s taskmanager, which for some reason always shows double the clockspeed it actually runs at (around 3.2GHz in idle (1.6), 6.2GHz under load (3.1) and 7.2GHz max. boost speed (3.6). No idea why it does that, but it's not the only piece of software that acts very strange in combination with this CPU, but otherwise works normal.

I agree with the comments below that the "fake" chip is actually a genuine Intel manufactured part, but the date code discrepancy does suggest that it has been remarked (god knows why!). Having said that I would have thought that it was difficult to remove markings from ceramic parts in order to remark them. The marking is also quite crisp, unlike some other parts which have been remarked in China (plastic MC68000 in PLCC) which had poor quality Motorola batwing logos. I also bought some MC68000 plastic DILs from China which were remarked from -8 (8MHz) to -16 (16MHz), because the new marking came off when I used IPA and a cotton bud, to reveal the original markings. When reported to the auction web site, they were not interested in following up. The big surprise was that at least in a basic system test the parts did work at 16MHz (but this was not a rigorous test) Nevertheless I was surprised that it worked at all at double the original marked rating. Googling brought to me to Ken Shirriff's Twitter feed in which he shows a 100% identical part to yours (perhaps he also bought it from the same seller as you?) which he de-lidded in order to see the die. He is an expert in this field and if the chip had been fake (i.e. non Intel) it would have been clear when looking at the die. I was also intrigued by your comment of the splayed legs. This is commonly found in many brand new DIL ICs, and would imply that the part is NOS. No way would Chinese fakers go to the effort to splay legs to make the parts look "new". Anyway the date code discrepancy is a puzzle, as a faker "should" have had no problem in choosing a genuine sounding date code. But it could be like my Moto parts, and was simply given a free speed upgrade! Seems a waste of time and effort given the already low selling price. BTW I've got no problem with the Chinese rescuing old parts from PCBs and re-selling them, but only if the parts are NOT remarked and are sold as re-cycled with a truthful listing.

Oh, the nostalgia your ST-225 brought back to me. That was my first hard drive in the summer of 1987. That was a banner summer for me -- I earned enough money in my summer job to buy this hard drive, a Microsoft bus mouse, AND upgrade my XT clone to 640 KB RAM. When I got back to college in the fall, I was stylin'! Thanks for the trip down memory lane!

9:12 sounds like a dirt bike

It's a bit harsh to call it a fake if it performs the same as a genuine Intel chip. It's possibly a reproduction for the enthusiast vintage PC users. Old hardware will inevitably fail and there's nothing wrong with someone reproducing these chips for those who still use vintage computers. The only scenario where "fake" would be appropriate is if these chips fail after a short period of time.

Yes, quite the unexpected edition of ncix tech tips, seeing as ncix doesn't even exist anymore.

I always see these and always wondered about them. I can now rest assured now that I know you have tested them. Thank you. You're easy going style and no frills video production makes you my favorite electronics and HiFi youtubeer.

I wonder if the chip failed the "Environment Test" due to have the NEC CPU instead of an Intel. While it would not be worth another whole video, It might be worth testing sometime, even if only for kicks and grins.

PC-9801-33 is a mathco meant for use with v20 and v30 CPUs on early NEC PC-9801 personal computers.

We gotta turn that ending into a meme

9:10 - I'd forgotten just how wonderful an XT sounds while booting. First, the crisp snap of the power switch (and accompanying flicker of the overhead lights in the room), then the slow growl of a vintage Seagate spinning up... (Wasn't Seagate the first manufacturer to use a green LED?) Then, the drive check on the 5 1/4"... 10:31 - Thank you for more Seagate noises. For some reason, I've always found those sounds soothing. I wonder if the slight degradation in overall performance has something to do with the co-pro being introduced into the bus in the first place. The CPU has to be able to communicate with the co-pro, I'm not sure how they did it but I imagine in that era it was handled either with a fixed IRQ line (interrupt request) or by forcing the CPU to "poll" it every so many cycles. It's all just theory, as my education started in the 486-era...

Could you do a comparison, using something math-intensive (i.e Lotus 1-2-3)?

The 8087 FPU was 'released' in 1980, but the date imprinted on it refers to the copyright date of the design. It may not be enough to indicate a 'fake' chip. It would be normal to see a date earlier than the purchase date. The Raspberry Pi Foundation does the same. Their first unit is ©November 2011, yet it was not revealed until the 29th February 2012, and released around Easter. Design times are faster today, yet there's still a discrepancy.

Isn’t it possible that the IBM also slowed down a bit when the chip was installed, but the slowdown wasn’t reflected in the results due to rounding? The numbers were all rounded to a single decimal place. The numbers on the Tandy may have just been closer to being rounded down. Like 1.26 vs 1.24 would become 1.3 vs 1.2. But 1.02 vs 0.98 would become 1.0 vs 1.0 after rounding. I think it’s natural there would be some minor penalty in a system for adding another element. It has to interact with the rest of the system somehow even when it’s not doing anything. Must be one of the reasons it was considered optional unless you really needed it.

Hearing the words "reliable" and "seagate" in one context let my brain fail the Exception Test.

After watching your video, I quickly rushed to grab my box of a single 287XL FPU bought on eBay from China; my heart sank when I saw the copyright mark: '80 '86. Markings and legs looks new even if the gold plate looks scratched.

That USA phone number for support in the Intel Math Coprocessor diagnostics still reaches Intel Support line all these years later lol

Hearing that computer wind up, chug chug chug, and then beep, brought back so much nostalgic memories. Thanks!

For a moment there, I got my hopes up of adding a math-co to my old Tandy. :-\ It didn't even occur to me to get one (it's probably the last upgrade I'll be able to give it). Thanks for the idea and the warning. Some day…

An 8087 dated 1978 is obviously fake since it wasn't available until 1980! I spotted that immediately.

I admire how thorough you are in these videos. Great job,

That old Seagate gave me chills. And I wasn't even alive then.

I'll save you 14 minutes and 27 seconds of your life - yes...

El-Fish also uses the math co-processor to speed up the animation pre-renderings you need to do before adding fish to a tank :) It sped the process up a huge amount.

6:40 The 8088 was actually introduced in 1979, not 1978. The 8086 came out in 1978, the 8088 being a cost reduced version of the 8086 with an 8bit external data bus (as opposed to 16 bits on the 8086); both are 16bit internally. also, @ 5:20, there's a performance hit from just having the 8087 installed because the V20/8086 still has to address it whether it's actively being used or not. It's a latency issue, probably the Tandy has a longer delay between the CPU and the CoPro socket. The Exception test errors may be an issue with the V20, then fixed in the V30. you could try swapping them between the 2 machines. a test with FPU aware software would be a good idea, though. the fake 8087 may pass the diagnostic tests, but it could still either be slower, or just fail hard on any real work. if it's good a modern repro, though, it could be significantly faster than the original.

Maybe "unlicensed" is a more suitable term than "fake"; the cores of these two chips are probably identical. It's interesting that you assert it doesn't even matter whether they perform identical or not when the result comes out. I wonder would you say the say thing if the "fake" chip proved inferior to the real one?

Oh man, I forgot how satisfying listening to old computers boot is.

Wow, blast from the past. If I recall correctly, software that used the 8087 were scientific academic programs often written in Fortran. I do recall installing a 8087 for a guy that was running AutoCad. Man, that was some time ago. All of a sudden, I feel old....

Is that Linus at the end? 😀

It can be a real intel co-processor that they recycled. Removed scratches and re-laserd the intel description on it. If that is the case than the processor is real, just the outer layer is not

I think you’re encountering an incompatibility with the V20. It would be interesting to see if it fails that test with a real 8088 installed.

Yeah I wondering why there was a few seconds of black at the end.

I don't think you should mistake the copyright date on the package for a date code that could related to the manufacturing date. The copyright is on the mask and/or design, and you'd want that to date back to when the design was realized to claim priority. Like the copyright on a book; it doesn't change on subsequent printings. The copyright date is there to document and protect the intellectual property, and in theory, could be used to prosecute those that would clone the design by reverse engineering it. Also, I think at the time, there was not an implicit copyright on works upon their creation; you hand to affix a copyright notice and register it. So not surprising that early versions of the 8087 would have a copyright (for the mask/design) in advance of when they were actually first available for purchase.

The "Fake" chip may be genuine Intel design they used to outsource fabrication to other manufactures. AMD and Cyrix for example manufactured the 8087. Intel used to licence third parties to make chips. The chip is probably NOS from some third party.

Unfair to call it fake if it works without any issue, or at least with the same faults as the original.

From these era's I also know one name very prominently concerning math coprocessors: "Weitek 80C87 Arithmetic Coprocessor" or also known as the "Weitek Abacus". These were somehow "famous" for being very powerful. And they were indeed, compared to those "standard mpu's" from intel. On 286,386 and as well 486SX boards there was an empty socket or even a plain soldering space to have such an optional mathematic coprocessor installed. From the DX series onward, such a processor was included in the main CPU.

the environment-test in MCPDiag may fail due to the NEC-V20 (main) processor (since that is not "genuine Intel")

OMG that hard drive spin up at 9:10 took me right back.

14:24 - Holy crap, that's a very young Linus.

the PC-9801 stamp means it came from a Japanese PC that was recycled.

There were Soviet-made 8087 clones, the KM1810VM87 thar are quite rare as collectibles, so I don't think that if one finds a tub of them in a warehouse will rebrands them as the more common Intel chip.

I always wonder what the FPU (as we know it today) actually was in it's core. Without one you (I never had one until I got a 68882 on my "accelerator" board for my Amiga) if you needed fast floating point you did precalculated lookup tables (yes used a bit more of the precious main RAM) and that was pretty much it. If you did not need it to be really fast use the microcode for FP or some optimized software. But did the early FPU's (like the 8087) use precalculated lookups (i.e a small processor but mostly memory with tables). Or was the 8087 actually doing "real" trigonometry calculations back then? Anyway as I mentioned the Amiga. I loved the use of the "library" system because of it many programs (or at least some) that was using the system/OS calls for FP operations was just by "magic" faster just changing the name of the "68882.library" to "math.library" (can't remember the exact names today) as it had the same functions but the original was CPU/Software FP functions and the other was exactly the same but used CPU/FPU/Software instead... But then again if someone really needed fast FP it still used lookup tables mostly and if it was optimised it bypassed the library/OS call and was compiled with direct FPU calls. But it was a neat function of AmigaOS and would be really beneficial today if it was still around...

It's great to see fake chips because it means they can reproduce those chips so we are not gonna run out of them any time soon and if they got the 8087 they can surely do the CPU and more

Going old school! FYI... The 486 SX did NOT have an FPU in it, only a DX model would. The 486 SX had the FPU disabled "burned out". Motherboards from 486 on down had a dedicated FPU socket.

Fake chips are very common on Ali Express. I have 74hc595s with a TI logo that's a picture of the USA, not Texas. On others, Texas is just blobby or misshapen. The writing on fakes rubs off if my fingers are sweaty or oily, like if I just scratch my face then insert a chip, the writing comes off. I got a fake 6805 with just an incorrect MOS logo and 6805 printed on it. They work fine, and I wouldn't have even known they're fakes if I didn't know what to look for. I even got some fake 555s and logic ICs in a Radioshack kit I bought locally. The writing on the 74HC32s came off so much I had to label them with masking tape. So I've gotten into the habit of calling out vendors who sell this stuff. I doubt a lot of them even realize they're selling fakes.

Informative and interesting as always 🙂

I love seeing all these old machines still running.

NEC V20 is technically an 80186, so pairing it with 8087 doesn't really make much sense. It's like putting 287 to 386 build. It can be done, sure, but the reason is quite unclear. That's why those tests failed. Also, that is not a fake. You talked about the suffixes, but not about the prefixes. C means ceramic with metal lid, D means double layer ceramic, then there are versions like MD (military), ID (industrial), LD (not sure to be true). And by the way, what you said about the speed isn't exactly true aswell. The speed is not dictated by CPU, but rather by speed of the bus. We did quite a lot of fiddling with those chips and found some interesting outcomes.

My 10 MHz 8087 did not work - I eventually got a refund, the guy wanted me to smash the chip. LOL it's in my collection on the wall, never to be used in a PC.

Now what if you were to try running the diagnostic with an actual Intel 8088?

I found a ton of fake chips in my parts I got off of AliExpress. Never noticed a single thing wrong with them, functionally. But I don't think TI's logo is a picture of the USA with just an I inlaid on it. For hobbyist purposes, it's not a huge deal. You'd be hard pressed to notice fakes if you don't know what to look for.

Could it be a take-out from a recycled machine? I can't imagine why they'd make a copy of a chip that's so obsolete as I can't think of something new that would include it.

I wouldn't doubt that this is a genuine 8087 that was recovered from scrap, then "rebadged" in China. You see this often with cheap Chinese ICs. They'll just scrape/putty-over/lazer chip markings and reapply them in order to sell the chip as "new." You can find a ton of great examples here www.muffwiggler.com/forum/viewtopic.php?p=1186022

I was waiting for you to say the 8087 was mostly used with lotus 123 ...love this stuff

I recognized the sound of that ST-225 right away!

Please don't call functional counterfeits "fakes". It just muddies the waters. Call them counterfeits. There are fake chips that don't work (different chip with the markings sanded off and new markings put in). There are also counterfeits that are a general purpose microcontroller running an emulator, which may result in compatibility, performance and stability issues. This chip is probably a die clone, they get either the original die photolithography prints, possibly from an older pre-release version or take a chip, and sand or etch it off layer by layer, taking photographs through a microscope, literally photocopying the chip, which is pretty cool IMHO. There are also chips that are "clean room reverse engineered", they take the same inputs and output the same outputs but the means to do so are re-implemented from scratch, any similarity in design is either for logical reasons (most efficient or familiar way) or coincidence.

It's good to see China still making vintage models of microprocessors despite their current EOL status in modern computers these days. At least it's better that nothing.

Would be good to see same round of tests with 287 and 387 chips.

Alright China, you won this time.

That chirp of the ST-225....... Brings back so many memories...

Humm, form the fake name, I would think that it would not work, but since it works... I installed one of these ''fake'' 8087 in my IBM PC XT and the chip is extremely hot. Can you tell if yours runs as hot, fake one versus real one?

14:24 Hahaha, I tought the auto play was on! :D

Regarding the Dullnes of the "fake" chip it is in a ceramic package. They always look like that. They are also higher quality (better environmental specs) than plastic ones. For years Military equipments used nothing but ceramic ICs.

Thanks for the video, that brought back memories :) (also, subscribed!)

Brings back some memories. Back in the early 90s my work computer was a 486DX 33 MHz, basically only used it for AutoCAD. Don’t even know if it had a math coprocessor chip in it or not.

The exceptions are probably the errata of the coprocessor. It might be because of the different errata every revision of a chip has and they might be testing for a specific one.

Ugh, I can never get enough of the sound of an ST 225 firing up

Seagate Drives are just the best. Hands down. ALL my old Seagate drives are still around.

Amazing, something that says it will do something does what it should. Hardly fakes. Would prefer to buy 5 of those than buy into the crap that you *NEED* a branded one.

Dos CAD software. Made extensive use of the 8087. I used generic cad extensively it was way to slow without the 8087. Ounce i put one in to my XT it was at least 3 maybe four times as fast. It was faster that way than on the 286 12 I had at work

Don't coprocessors share the computing load with the main processor? I've seen systems with significant speed increase with a math coprocessor.

So, if it's fake, is unable to work [period]; if its just an 'unauthorized copy', actually, can be work. Possibly.

6:22 It's also the line of PCs where Touhou Project got its start

Instantly knew that was an ST-225 by the sounds. The one in my Zenith sounds identical and still works perfectly!

Thanks for answering my question from your last 1000 SX video with an entirely new video. ;)

9:10 Warning: Extreme nostalgia alert.

As others have said I think it might be genuine old stock as well. At worst it would be a clone chip (without branding) but using the Intel branding would make it a counterfeit item which would be risky for importers so I doubt they'd bother. Plus I don't see any chinese company spinning up a fab just to put out clone chips to.... enthusiasts like us who would buy maybe uh... 1 unit in their lifetime? Doesn't make sense. The market would be so limited as to be pointless. Some have commented on a Taiwanese origin which seems to coincide with the timing around when the clones market started up and started gobbling away market share from IBM. Interesting.

I'd say 9 bucks well spent.

Unrelated, but can you give some updated advice on your 2011 "Tour of my radio station" video? I need to transmit from one building to another about 300 ft. away. Are those 5/15W transmitters with an AUX input any good? They seem to market them to churches for some reason.

Would be cool to have an X-Ray machine to figure this out, hell, even to just have some fun and scan all sorts of things. Just don't try to use Big Clive's "X-ray" machine lol.

I LOVE the way those old Seagates sounded.

That NEC could've run some really nice hentai games in its day.

I COMPLETELY AGREE about the ST-225! Solid as a rock!!!

But, if the fake chip passed all tests and has no negative impacts, is it still a fake?

I wonder if the fake 8087 is actually a pull and they sanded it down and redid the writing to make it look new?

Try to use some alcohol or acetone on the fake one, I have seen some are 'painted'. If the surface is real (laser encraved) nothing will happen but if it is fake, black stuff wil come off. Wear handcloves while doing this, because you never know what exactly they used to 'paint' the surface.

Thanks, dude! I got a 8087 for my IIgs pc board - havent tested yet but it is the golden one so hopefully I'm ok here!

great video! thanks for all the tests carried out!

All right! Now I can run my Lotus 1-2-3 programs at top speed!

"Environment" probably refers to the Floating-point Environment. software.intel.com/en-us/cpp-compiler-developer-guide-and-reference-floating-point-environment This specifies how the processor handles rounding, and also includes the status flags, including exceptions. The original PC connected the FPE interrupt pin to the processor's NMI pin (but first you had to flip the NP switch to gate the connection, then leave NMI enabled by writing 0x80 to port 0xA0). Chances are it was enabled, as memory parity errors also triggered NMI. The Tandy may not have made that connection, or requires some nonstandard port command to enable, or maybe even wires the pin to the interrupt controller instead. This is probably why the exception tests failed. Doesn't explain the overall environment failure on the IBM PC though. But there are a huge number of status and control bits, therefore many opportunities to fail.

So you say that the ST-225 is the most realable mfm hard drive... I didn't know that, now it feels good to have two of them in my collection.