Teardown, Optimization: Comsol 8Gb USB Flash Stick (2015) | Hacker News Reader

Teardown, Optimization: Comsol 8Gb USB Flash Stick (2015)

goughlui.com

27 pointsuserbinator4 comments

The ingenuity to use parts that failed in binning for other products with less demanding/strict requirements... amazing. How does one even get the connections to buy such parts?

I can't imagine Sandisk (or anyone else) be too happy that failed parts in the end did make it out of the factory, alone for the risk of someone applying a fake Sandisk label on a part that actually did come out of a Sandisk factory.

Bunny reported on rebinning a decade or so ago. He also talked about “ghost lines”, a crew would run a batch but not register it in the paperwork.

How does one even get the connections to buy such parts?

In general, it's called "China". Also happens with US companies like Micron - look up what Spectek does: https://www.micron.com/content/dam/micron/global/public/spec... (in particular, see the lines about "carcass die".)

I think in this case the microSD card had a defective controller or perhaps a bonding defect, so it would not have been usable as such, but the NAND was still good. This was caught post-assembly, and so got recycled into raw NAND. Sandisk gets to profit from it, so I think they're definitely happy about selling these.

No, they ship failed units all day every day. Someone will chime in with more accurate numbers, but the typical yield for making silicon dice is like 70-90%. You will always have some failures in a run, it's unavoidable.

So that those chips aren't entirely wasted, all IC manufacturers bin where possible. If 2GB out of your 8GB flash chip are bad, you don't just eat that cost, you sell it as a 4GB chip.

Everyone does this all the time in every market segment. Silicon manufacturing is simply an imperfect process. The only logical thing to do is try to sell defect chips as a different product that works around the defects.

Reply on news.ycombinator.com