28

The reason I'm asking is that on http://alibaba.com you can find prices for the Core i7 as low as $20, minimum quantity 1. This looks like impossibly low for a genuine Intel, but then I also can't believe you can clone this kind of CPU. What's the matter here, and what kind of problems can I expect with these cheap CPUs?

Federico Russo
  • 9,658
  • 17
  • 68
  • 117
  • 2
    Why was this downvoted I wonder? – sharptooth Nov 12 '12 at 09:31
  • Maybe they are remarked? That used to be a problem few years back with AMD if I remember correctly. – AndrejaKo Nov 12 '12 at 09:33
  • You can clone the _packaging_ very easily ... eg for compact flash http://www.bunniestudios.com/blog/?page_id=1022 – pjc50 Nov 12 '12 at 09:34
  • 34
    They might be stolen, not cloned. You know, Alibaba and 40 thieves... – Frantisek Kossuth Nov 12 '12 at 12:14
  • 9
    They probably "fell off a truck" someplace in southeast Asia. – Olin Lathrop Nov 12 '12 at 13:52
  • 10
    On Alibaba it often happens that seller's advertise products for a 10th of the selling price. When you then buy the product the seller has 10 days to refuse the order, giving him the time to tell you that the product was mispriced and asking you to pay full price. There's nothing you can do about this except give the seller a bad rating. I had this happen to me when buying a hunting camera. – Peter Nov 12 '12 at 15:55
  • 2
    Normally they send an inferior processor. I have seen tons of complaints in the feedbacks left. – gstorto Jul 23 '15 at 23:33
  • Chances are they are not i7s at all. They may send you a 386 that's been sitting in the warehouse for 20 years where they rubbed off the label and wrote i7 instead. Or a small brick. Don't expect to get something that actually works. – user253751 Apr 07 '18 at 06:58

6 Answers6

36

There are a lot of different things you can get when you see something like this. For a very new part like this I would assume that since Intel is pretty much the only company with the ability to actually build these CPUs (they use a very small manufacturing process), these are either bricks of lead attached to the correct packaging to look like a CPU, or they're failed CPUs. There's actually a relatively low yield on tiny manufacturing processes like Intel's current generation (22nm is the current size they use). I've been led to believe it's something like a 60% yield (i.e. they produce 100 processors they only get 60 that actually work) and the rest have to be discarded. But I have no real numbers on that, but even if it was a 99.9% yield, that would still mean that 1 in 1000 was bad and had to be disposed of, and Intel produces a lot of processors. And someone is probably interested in cheap, mostly functional CPUs.

What functionality is actually missing in these discarded chips could be very minor. Something like "dividing anything by 3879 never gives the correct answer", but clearly a chip with a flaw like that could never be released without permanently damaging the companies' reputation. So if these $20 i7 chips do function in a core i7 motherboard, I would assume that you would find that each one would have some subset of functionality that misbehaved. Alternatively they could only work if underclocked, or if they were much cooler than the specifications normally allow. Who knows! It's a lottery of functionality.

Another unlikely possibility is that these are some other chips which have the same pinout but do something totally different. See Sparkfun's adventure in counterfeit ATMegas (note that the post I linked has several updates where they learn it's an ON semiconductor part from the 1980s). This is extremely unlikely though, as Intel varies its pinouts frequently, so other manufacturers wouldn't be producing parts which would fit in this generations' sockets.

Something which sometimes happens in China, is that the employees will come into the facilities and run the factory when the managers aren't there, and sell the output as genuine product, even though it hasn't been tested. In general the test equipment and the equipment which marks the packaging with "Genuine __ part!" don't work without the manager's password or something similar. This is normally an issue in places that produce SD cards and similar though, where the process is relatively simple and short. These are called "ghost shift" components because they're produced by a shift of workers who aren't supposed to be there. Intel's chips are probably a bit too complex for something like that, and I think they're mostly produced in America anyway.

So long story short: I don't think there's another entity with the ability to produce 22nm parts as complex as Intel's CPUs right now, so I imagine these are either defective core i7s, or completely fake.

Edit: Or, as Olin noted below, the least interesting answer electronically: they're just stolen chips (I like my schemes to be more elaborate!)

Kit Scuzz
  • 3,279
  • 21
  • 19
  • 60% yield seems extremely low to me, that's not even 2 sigma. Does anyone have accurate figures on this? – Federico Russo Nov 12 '12 at 12:22
  • `I don't think there's another entity with the ability to produce 22nm parts as complex as Intel's CPUs right now` I'd say AMD comes close... :) – TC1 Nov 12 '12 at 12:47
  • Isn't the standard procedure for minor hardware bugs to fix them in the microcode by distributing updates to OS and firmware vendors? – AndrejaKo Nov 12 '12 at 12:53
  • 1
    @FedericoRusso: I've seen wafers where the yield was <10%, at which point they had just discarded the entire wafer. At a certain point, you're going to assume a common root cause and not trust any die from that wafer. I don't know at which point Intel throws out entire wafers, but it could certainly explain part of that 40% loss. – MSalters Nov 12 '12 at 13:33
  • 2
    *these are either defective core i7s, or completely fake* or stolen. – Olin Lathrop Nov 12 '12 at 13:56
  • 1
    There are a few other companies that can compete with Intel's manufacturing process, including Samsung and the independent "for hire" TSMC (though probably not with the same capacity as Intel), but to actually clone a Core i7 would require some serious industrial espionage. Also, I would guess that the design of the Core i7 and its manufacturing process are very closely related, to the point that it would be practically impossible to make the same chip with another foundry's process. – Jonathan Amend Nov 12 '12 at 15:42
  • 1
    @TC1 and Jonathan Amend: AMD may come close, but they're not a fabrication company anymore (they design and then get other companies to fabricate for them). I don't know if there are companies making non-flash 22nm chips, as AMD's most recent chips are still 32nm. See the [wikipedia entry](http://en.wikipedia.org/wiki/22nm) – Kit Scuzz Nov 12 '12 at 17:22
  • 1
    @FedericoRusso : I'm pretty sure the real numbers are closely guarded secrets until several years after the fact, but in general from what I understand, yields can start out as low as 50% for a new process and get better with time. See [this article](http://www.tomshardware.com/news/intel-ivy-bridge-cpu-processor-supply,15421.html) for Intel's possible i7 issues, and [this one](http://www.tomshardware.com/reviews/intel,751-6.html) for an old article about low production yields. – Kit Scuzz Nov 12 '12 at 17:28
  • @AndrejaKo : It depends on a number of factors. For example, not performing to within the temperature specs or up to the required clock speed can't be fixed by microcode updates. Really, the types of problems which can be solved by microcode updates are pretty limited, because they will affect performance negatively, and require that the underlying hardware still works near-perfectly. – Kit Scuzz Nov 12 '12 at 17:35
  • 2
    I checked here before I posted: http://en.wikipedia.org/wiki/List_of_semiconductor_fabrication_plants (also I didn't mention AMD). – Jonathan Amend Nov 12 '12 at 18:09
  • @JonathanAmend : Ah! I stand corrected, though I'm still curious about any of them producing chips of a similar complexity. I know that a lot of fabrication companies will get to a small nm process and simply produce flash wafers, and I'd be curious if TSMC or CNSE (or any of the 20nm fabs) are actually producing processors. – Kit Scuzz Nov 13 '12 at 02:21
  • @KitScuzz, IBM would be the first candidate and AFAIK, the latest POWER chips are still on 32nm. – AProgrammer Nov 13 '12 at 13:14
  • The SparkFun adventure you linked is epic and is well worth a read. Thanks! – Roman Starkov Nov 18 '12 at 13:44
12

They are probably Engineering Samples (ES). Intel produces them in large batches for testing, after which I imagine they are supposed to be discarded, but evidently they become available to the general public. These chips are pre-production, and they are usually functional, though some features like SpeedStep and thermal throttling might be disabled depending on the stepping (version). You can also find these items on eBay. Of course the items on Alibabi could also be non-functional ornaments, stolen, or re-badged, but I highly doubt they are actual "clones".

Jonathan Amend
  • 221
  • 1
  • 4
  • Back in the late nineties, engineering samples were sought after items because they were often multiplier unlocked and would overclock well. – Morten Jensen Oct 23 '17 at 11:50
7

It is extremely difficult. As far as I know impossible for the current organizations in the world beside Intel.

First of all let me say that there are a few quite different chips that goes under the name Intel Core i7. They may be older versions of the Intel Core i7 that those companies want to get rid from the inventory, but the price still look way too good. It may just be a scam or stolen chips.

Going back to the question, creating an Intel Core I7 would require having the Intel chip design and a chip foundry with the capability to use them to produce very cheap products. This is very unlikely.

apaderno
  • 107
  • 4
ElegantSolution
  • 71
  • 2
2

Assuming that you even get a processor at all it is:

A. probably not functional

B. not the right processor

C. has unknown defects that could result in hazardous situations (such as chemical fire, extremely toxic emissions or electrical fire/complications etc)

D. the product (whatever it is you receive after payment) could take out (short circuit) your components making the $20 dollar product actually cost more or equal to an actual i7 processor.

E. May partially function, but like Kit Scuzz mentioned it could have some unknown minuscule fault that would not allow it to be sold as complete processor.

Scenario #2: This could be some online scam where you receive nothing at all and they steal your identity/money.

Usually if the price is too good to be true, something is wrong. There are deals don't get me wrong, but any high end product at 80 percent or higher mark down is going to be a scam. There are times where it can be the odds are not in your favor.

Cambolicious
  • 21
  • 1
1

As the other comments have explained what you'd likely be getting (a scammer who gives you a different product/refuses to sell it to you until you pay full price/a stolen or broken processor), I'll try to explain the original question, whether or not it's impossible to make illegal clones of an i7.

Physically possible? Sure. Likely? Absolutely not. If you've ever read the datasheets for these processors, you'll see thousands of pages of dense technical, yet vague, material. The ACPI v6 specification is a thousand pages. The x99 chipset is over 900 pages. The PCIe 3 spec is hundreds of pages. The Intel 64 and IA-32 Architecture Developer Manual(s) are thousands of pages in total. All together, there are dozens of these documents which necessarily explain the functioning of these processors. That's only the public information. There are far more internal documents which Intel only gives to OEMs, and more still which Intel gives only to extremely high-paying partners (e.g. Google and Microsoft). Still others they only keep to themselves.

Every one of these are required to make a genuine Intel clone. Many of them would be required to make a cheap clone that was even able to start up in real mode, much less support all the documented and undocumented behavior which operating systems and peripheral hardware rely on.

Just a couple examples I could quickly find:

  • Intel Core i7-900 vol 1 - 2: 194 pages
  • Intel C230 chipset PCH vol 1 - 2: 1693 pages
  • Intel VT-d spec (directed I/O): 276 pages
  • Advanced Power Configuration Specification v6: 1056 pages
  • Serial ATA revision 3: 663 pages
  • Intel Software Developer Manuals vol 1 - 9: 5392 pages

Unlike a cheap ATmega32 or 8051 microcontroller with just a handful of instructions, maybe an internal clock, or maybe not, a couple GPIOs, etc., a full-fledged Intel processor with a gazillion extensions to the x86 ISA is neigh impossible to clone without having both access to internal, unpublished Intel documents, and a huge amount of resources. Even creating a barebones x86 processor that is capable of booting a modern operating system takes a lot of work. After all, there are really only three manufacturers who even release such processors: Intel, AMD, and VIA.

guest
  • 11
  • 1
  • Cloning of complex parts often doesn't involve recreation from design documents, but copying a genuine article (using e.g. electron microscopy to examine it) – Ben Voigt Jan 21 '17 at 04:38
  • Good luck recreating an Intel processor from physical reverse engineering. It'd be far easier to recreate it from design documents than to use electron microscopy, voltage/clock glitching, laser fault injection, etc. – guest Jan 21 '17 at 04:44
  • Obviously it isn't possible for such a small process. But for larger transistors, I believe the design is never reverse-engineered. It's more like photocopying, the masking for etching are made using the electron-microscope images, such that no one has to actually know the function of each transistor, just duplicate its size, position, and connections. – Ben Voigt Jan 21 '17 at 04:55
  • Sure. And even for Intel processors, it can be done in combination with laser fault injection to reverse engineer small parts of the system, but not the entire thing from scratch. Smaller processors are much easier, even without fault injection. – guest Jan 21 '17 at 05:00
-1

Actually these mostly are refurbished or used processors which are pulled out from laptops or destops and packaged as new. the processor will work fine but itll be old . no to worry coz they last long. Or there wont be a processor but a dud.

Abhyuday
  • 1