Question Can discarded/rejected microprocessor chips be repaired?

[Xcom1Cheetah]

I was watching an economy program on youtube and in which they discussed how Chinese companies buy rejected/discarded microchips at a negligible cost and repair them and use them in non-branded mobile phones.

In my little reading on the microchip fabrication process, I doubt that such a thing is possible. When you are talking about sub 20 nm chips, they require advanced laser technology and I think there fixing is not possible without more advanced technology then their fabrication.

But i thought to confirm if i am correct or wrong

Thank You

 

[NTMBK]

I was watching an economy program on youtube and in which they discussed how Chinese companies buy rejected/discarded microchips at a negligible cost and repair them and use them in non-branded mobile phones.

In my little reading on the microchip fabrication process, I doubt that such a thing is possible. When you are talking about sub 20 nm chips, they require advanced laser technology and I think there fixing is not possible without more advanced technology then their fabrication.

But i thought to confirm if i am correct or wrong

Thank You

It depends why the chip was rejected. It might be functional, just unable up run at a fast enough clock speed to meet the product spec, or run at too high power consumption. I can imagine that those sub-standard rejects would be good fodder for bargain bin phones.

 

[VirtualLarry]

It depends why the chip was rejected. It might be functional, just unable up run at a fast enough clock speed to meet the product spec, or run at too high power consumption. I can imagine that those sub-standard rejects would be good fodder for bargain bin phones.

Yeah, functional chips that "fail binning" could still be re-sold to budget phone makers.

I suspect that that article may have be talking about NAND chips, though, which are highly redundant arrays of cells, rather than micro-processor logic chips, which are more likely to have functional defects that prevent proper operation.

NAND chips are also binned for functionality, and defect cell percentage. The cheapest of which, for example look at the list of Spectek NAND grades available to customers (available floating around on the interwebs), shows different grades, and even a bunch of lower grades, that are essentially sold "as-is" by the NAND mfg, and left for their customer to test functionality of the chips themselves. OCZ used to go that. (Look where that got them.)

Edit: To make an analogy for the NAND defective-chip situation, it's kind of like buying a lot of HDDs, with defects. (Modern IDE/SATA HDDs have a factory-defect map, so none of the defects are user-visible, but it's a coarse analogy. They still work with those factory defects, but possibly at a lower capacity overall. Hence some 750GB HDDs were on the market at one point, rather than 1TB.)

 

[Hitman928]

Counterfeit ICs are actually a big problem in the industry. Counterfeit in this context doesn't mean a fake chip being sold as real, but rather a used chip or rejected chip being sold as new.

As far as repair goes, it depends on what is wrong with it. Sometimes it could be something like a bent pin or something like that which could be repaired. If the chip is fried internally, then no, you're not going to fix that. Even if it were possible, you'd be spending way more trying to do it than you would ever get reselling it.

Usually what happens is they take chips from discarded products that have working chips. For example, a phone with a broken screen that won't take a charge or even turn on may be purchased for basically nothing. The chips inside may still be functional so you take them out and sell them as new or put them in a "new" phone and sell it for $100 or whatever they can charge. The problem is reliability of those chips as they have already experienced significant aging, but the people who do this don't concern themselves with such matters.

You also have the possibility of fabs overproducing ICs beyond what was ordered and then selling them on the side or as mentioned, someone along the way taking rejected chips for whatever reason and selling them. Again, performance and reliability is probably not good but the people selling them don't care.

 

[Eug]

That's why I try not to buy DIMMs from 3rd tier memory companies, and I prefer to buy 1st tier over 2nd tier. My preferred memory brands are 1st tier ones like Samsung, Crucial/Micron, Hynix. Second tier memory brands include brands like Kingston, Transcend, Corsair. Third tier would be ones like A-Tech, Timetec, OWC.

And by Samsung and Crucial I mean DIMMs that are actually branded as such, not third party RAM that happens to use Samsung or Micron chips, for reasons already mentioned in previous posts.

 

[Xcom1Cheetah]

It depends why the chip was rejected. It might be functional, just unable up run at a fast enough clock speed to meet the product spec, or run at too high power consumption. I can imagine that those sub-standard rejects would be good fodder for bargain bin phones.

Yeah, functional chips that "fail binning" could still be re-sold to budget phone makers.

I suspect that that article may have be talking about NAND chips, though, which are highly redundant arrays of cells, rather than micro-processor logic chips, which are more likely to have functional defects that prevent proper operation.

NAND chips are also binned for functionality, and defect cell percentage. The cheapest of which, for example look at the list of Spectek NAND grades available to customers (available floating around on the interwebs), shows different grades, and even a bunch of lower grades, that are essentially sold "as-is" by the NAND mfg, and left for their customer to test functionality of the chips themselves. OCZ used to go that. (Look where that got them.)

Edit: To make an analogy for the NAND defective-chip situation, it's kind of like buying a lot of HDDs, with defects. (Modern IDE/SATA HDDs have a factory-defect map, so none of the defects are user-visible, but it's a coarse analogy. They still work with those factory defects, but possibly at a lower capacity overall. Hence some 750GB HDDs were on the market at one point, rather than 1TB.)

Counterfeit ICs are actually a big problem in the industry. Counterfeit in this context doesn't mean a fake chip being sold as real, but rather a used chip or rejected chip being sold as new.

As far as repair goes, it depends on what is wrong with it. Sometimes it could be something like a bent pin or something like that which could be repaired. If the chip is fried internally, then no, you're not going to fix that. Even if it were possible, you'd be spending way more trying to do it than you would ever get reselling it.

Usually what happens is they take chips from discarded products that have working chips. For example, a phone with a broken screen that won't take a charge or even turn on may be purchased for basically nothing. The chips inside may still be functional so you take them out and sell them as new or put them in a "new" phone and sell it for $100 or whatever they can charge. The problem is reliability of those chips as they have already experienced significant aging, but the people who do this don't concern themselves with such matters.

You also have the possibility of fabs overproducing ICs beyond what was ordered and then selling them on the side or as mentioned, someone along the way taking rejected chips for whatever reason and selling them. Again, performance and reliability is probably not good but the people selling them don't care.

Thank you for your insight and given other perspective.

But if a Chip is rejected by Intel/ARM due to an issue in the ALU or FPU, then can it be fixed?

 

[VirtualLarry]

But if a Chip is rejected by Intel/ARM due to an issue in the ALU or FPU, then can it be fixed?

Not that I am aware of.

But possible a quad-core with one defective core could be sold as a triple-core. Maybe that's what they're talking about?

Intel themselves have never, to my knowledge, done triple-cores, but maybe there are some unscrupulous third-parties that could get a hold of their reject lots, and build PCs for the Chinese-only market with Intel triple-cores. It's possible, but somewhat implausible.

 

[Xcom1Cheetah]

Not that I am aware of.

But possible a quad-core with one defective core could be sold as a triple-core. Maybe that's what they're talking about?

Intel themselves have never, to my knowledge, done triple-cores, but maybe there are some unscrupulous third-parties that could get a hold of their reject lots, and build PCs for the Chinese-only market with Intel triple-cores. It's possible, but somewhat implausible.

Can a core be disabled without advanced technology?

 

[VirtualLarry]

Can a core be disabled without advanced technology?

Yes, in micro-code, or even in BIOS.

 

[Thala]

Yes, in micro-code, or even in BIOS.

Typically its being fused off.

 

[piokos]

But if a Chip is rejected by Intel/ARM due to an issue in the ALU or FPU, then can it be fixed?

I suspect you may be thinking about modifying the structure of an existing processor (so that it doesn't have to be downgraded or trashed).
So we're missing a few atoms and can they be added. Something analogous to soldering broken electronics - but on nanometer scale.

The general answer is: yes! We can manipulate single atoms to some extent (and we keep getting better at it). So in some situations such a "fix" would technically be possible. But it makes no sense economically. Just the process of localizing the fault would usually cost more than replacing the chip.

In reality, you're making a chip with some functionality in mind and you test that functionality. You don't test the actual physical quality of fabrication.

 

[DrMrLordX]

And by Samsung and Crucial I mean DIMMs that are actually branded as such, not third party RAM that happens to use Samsung or Micron chips, for reasons already mentioned in previous posts.

That's pretty limiting. There are many markets where Samsung won't even sell memory under their own brand. G.Skill and Corsair have arguably the best DIMMs on the market, so I'm not sure why you would list Corsair as "second tier" and why you wouldn't even list G.Skill .