Examining a vintage RAM chip, I find a counterfeit with a different die inside

For those of us that were, seeing "64-bit" and "TTL" together immediately told us it was a chip that stored 64 bits worth of data.

Three years ago I posted a link on HN about TTL (Lransister-Lransister Logic) being 50 years old, entitled "50 years of TTL". Similarly, many people unsurprisingly expected it to be a story about a weird Time To Live field.

Maybe they were printed in the same Fab and a whole batch of these got mislabeled? Maybe someone intentionally relabeled them years ago back when they were worth more?

One case was a fake flash key, another a scammer trying to "buy" my laptop with fake bank transfer receipts, and the last one a genuine SD card, but destined to the chinese market.

In all cases Ebay didn't care much; in the first, I even lost my money.

Nowadays, when something similar happens, I just report the item as not arrived; although this is not 100% ethical, for clients it's sadly the safest way.

This is in reality part of a much larger issue, where the portals profit from being more lenient, so that they have a larger amount of sellers (and sales) of cheap crap, rather than being more punishing, and losing sales.

I definitely advise to report items as not arrived, rather than risking through reporting frauds.

Turns out she opened too many cases and they "suspended" her account, which is an eBay euphemism for being banned for life. That means that her account that she'd been using since 2000 was dead and gone forever, and she was no longer welcome on the site.

I agree with other posters. The eBay of old is effectively dead. You are better off buying things from Amazon or Ali Express 99% of the time. The days of great deals on items are pretty much gone. You can still find some rare and unusual stuff that is worth buying, but the dominance of eBay now means that most things are very close to market rate unless (or even if!) they are counterfeit or some other scam. More like a dollar store than an online auction house.

[1] Potentially NSFW: https://forums.somethingawful.com/showthread.php?threadid=10..., SFW summary: https://www.engadget.com/2004/05/14/scamming-the-scammer/

Either that or a plain old mistake. The new guy loaded the wrong data into the printer.

Also, pushing fakes in small quantities may not be such a bad idea; after all - rip someone off for $1 and chances are he'll curse you and let it go; rip someone off for $1M and chances are he'll curse you and get law enforcement involved.

Maybe not in time. The companies buy lots of stuff that never gets used. It seems totally plausible for them to use up their end-of-year-budget stocking up on parts they don't necessarily need or expect to use through non-conventional channels.

Plus, unless you dig into it like this guy did, he can just say what he did and give a refund.

There are surprisingly many things in computer which you don't actually need and don't know if those are working or not.

I'm just looking at the pdf they link to in this article: https://www.semiconductors.org/clientuploads/Anti-Counterfei...

In the summary they say, the following on how to 'win the battle'

"The key to winning the battle against counterfeit semiconductors is elegantly simple: Exclusively buy semiconductor products either directly from the Original Component Manufacturer (OCM) or directly from the OCM’s Authorized Distributors/Resellers."

I wonder if they have hardware to test the functioning of chips, like testing them to ensure they meet the parameters on the datasheets (like current consumption etc.).

They're also the kind of organisation that does "whole life" buys, so they'll take the whole lot they'll ever need, sample some for QC purposes, and warehouse them until actually required for manufacturing.

Other people - scam artists - see what's being searched for, and take a different IC in the same package and remove the existing markings, and put new fake markings on. They then sell this as the other IC.

By the time the ICs are tested the money has gone. (Which should be a hint - if they're not offering regular invoicing and 30 days payment they might not be trustworthy).

tl;dr not enough people design _for production_, and it makes things much harder.

Here's the take home clinch for the story, "physical viscerality"...

I wonder about the future of the spy genre in our digital, post-historical era. The essence of espionage is information—specifically, information about the capabilities and intentions of friends and adversaries. The mechanics of this trade has always been the bread and butter of the spy thriller. The whole first half of Le Carré’s masterpiece Smiley’s People, for example, turns on the physical transportation of a single incriminating frame of negative film across national frontiers by couriers, several of whom end up dead. That was the analog world of the Cold War. Our digital world, in which the contents of the Library of Congress can be encrypted and transmitted across the globe with the touch of a key, is far less dramatic and does not lend itself as easily to romance. Kim Philby spent the better part of two decades transcribing the crown jewels of British and American intelligence secrets by hand and turning them over to the KGB. As current headlines attest, a single computer hack or anonymous leak today can yield a far bigger cache of secrets. Do spies still use dead drops? Brush passes? Microdots? Invisible ink? Do they still meet their contacts in smoky cafes and secluded parks? Many of these gritty noir devices may have been retired and replaced by banks of computer screens in the sub-basements of Northern Virginia office parks. All of which makes the spy genre poorer and more antiseptic. We are left with stylish but vapid movies about superheroes like Jason Bourne, pursued by cartoonish CIA assassins.

The digital world also changes our perceptions of political crime, of which espionage is a sub-species. Without the physical Watergate break-in and the amateur-hour rifling through DNC file cabinets, there would have been no scandal. In contrast, the political scandals of the last election cycle seem mired in the geek-squad arcana of passwords, servers, and hard drives. They lack the visceral physicality of Watergate, which is why they are unlikely to amount to much, despite the wishes of political partisans. https://www.city-journal.org/html/alan-fursts-world-spies-15...

The underlying thing that makes the shift register more efficient is that you can build a dynamic shift register stage with two inverters and two pass transistor pairs, while the flip flop for a counter stage is probably six gates.

Indeed, Ken was in group #1 (collector) but a very unusual kind who then proceeded to tear down the chip.

1. this chip supposed to be old era memory chip (64bit of data -> 8 bytes) it wouldn't work on your PC

2. It was DTMF chip not memory chip, so it couldn't work at all as a memory chip

I guess the idea was that it costed $1, the seller shipped it, the chip would never work, but cost of shipping it back would be greater, so you would just eat the cost and move on.

As with other kinds of fakes, they try to emulate the functionality, and often work, some even surprisingly better than expected, but they are always worse than originals. So either you won't get performance that's mentioned in the spec, or it break sooner or has some other issues that get apparent once you start using it.

If you need to check external memory (GPU, IoT, etc) then there is some tools online.

Note that memtest86 doesn't test against Rowhammer like attacks probabilities.

You can also use some overclocking features in advanced motherboard to push the memory beyond factory settings and see how far it gets before it starts to misbehave.

On the other hand I've seen memory module with sufficiently unfortunate pattern of broken bits, that it could not be detected by memtest86. Even such tools need some part of memory that they run from and thus does not get tested (also some memory is reserved for BIOS). The problem manifested itself by completely unusable Windows and occassionally flipped bits in files downloaded from network on Linux (I assume that on Linux the faulty bit ended up in DMA bounce buffer, while on windows in something significantly more critical). To confirm my hunch, that it was memory related I netcat'ed few GB worth of CHARGEN output and analyzed the pattern of bit errors, which were perfectly aligned on some boundary. Finding which module was faulty then involved rearranging modules, running memtest86 and repeating the CHARGEN experiment.

Edit: actually diagnosing this took about month and half. memtest86 was passing and there was unrelated known hardware bug in used network card which I though might be related to observed behavior. Experiment with chargen was motivated by finding out whether the bit errors were somehow aligned to TCP segment boundaries, instead I found out they were aligned on some quite large power of two, which clearly pointed to memory issue.