Wierd Flash Stick Behavior

[msuffidy]

I like to see if people are giving away stuff on the internet and today I got these 5 new 256GB flash sticks. I was testing them for reliability. When I formatted them ntfs it won't mount because I think the (mostly empty) filesystem was corrupted, but when I format it as exfat, it works fine. So far I wrote 14GB and diffed it without problems. So my theory is there is a different block size or buffer flushing for exfat?

[ataradov]

Based on the label, I assume they are supposed to be 256 GB? The drives are likely much smaller and controller lies. NTFS stores multiple copies of the metadata across the drive, FAT only stores stuff at the beginning, so on a formatted drive you won't know that something is wrong until you write over the real physical size.

There is software that does full testing with random data and no FS, that will tell you the real size.

[msuffidy]

Oh that could be the problem. I will keep writing to it and see what happens.

[msuffidy]

Something else now it did which is even more retarded is I think it tries to slow down the write after like 20GB to avoid you figuring it out. Yes it probably is a scammer product. I will try a chipset utility to just make it what it is supposed to be or just make a 16GB partition.

[msuffidy]

Some of them are even more braindead so I wanted a utility. I could not find the chip info without taking it apart. I think this was my first successful non destructive usb teardown.

Weird huh there is an SD card in there.

[msuffidy]

The only information I am finding was in Chinese and I translate it. Also the one I took apart seemed especially non responsive but a long shot shoving the card into the slot maybe did something and it is working ok for a 16GB partition.
Detailed description

1.1 Features

 

◆System Clock

Mode: built-in low-speed clock, built-in high-speed clock.

When the system runs at the lowest frequency, the system is in low power consumption and normal operation.

◆ General features

Handshake communication control for iOS system.

◆ USB controller

Support USB2.0 High Speed ​​Device Controller.

Support USB2.0 High Speed ​​Host Controller.

USB2.0 is compatible with USB1.1 mode, and the power consumption of the whole machine is less than 10mA in USB1.1 mode (conforming to iOS Accessory Protocol 2).

Comply with IF, WHQL certification.

Support USB multi-endpoint.

◆ IIC controller

◆ Power Controller

Built-in 3.3V LDO voltage regulator.

3.3V PG output controller.

With 4KV ESD-HBM Protection.

◆ Working conditions

5V LDO input;

3.3V LDO input

Working temperature: -30℃ to +85℃.

◆ Package form

[gnuarm]

Based on the label, I assume they are supposed to be 256 GB? The drives are likely much smaller and controller lies. NTFS stores multiple copies of the metadata across the drive, FAT only stores stuff at the beginning, so on a formatted drive you won't know that something is wrong until you write over the real physical size.

There is software that does full testing with random data and no FS, that will tell you the real size.

The software I use for this is h2testw.  This program is very popular.  I still use the version I downloaded some years ago.  It is possible that there are sites with a copy of this that contains a virus, so scan it before you use it.

I've tried buying Flash on eBay, finding that the lower price end has lots of fakes.  Actually, below some price point, I expect they are all fake.  Even some of the higher priced devices are fakes as well. 

In this case, I would try writing a bit more than 16 GB, perhaps 18 or 20.  If it works at that amount of data, it's likely good.

[Fraser]

The other issue is reliability. If someone goes to the trouble of creating fake flash drives I do not see them using the better quality components within, including whatever flash memory IC they choose to use. I would not be at all surprised to see production line rejects (scrap) used for the flash memory. Such QC fails are not something I would want to trust my data with. Using junk flash memory is more trouble than it is worth in my opinion. If a USB stick turned out to be a fake I would chuck it in the bin without hesitation. Having such knocking around my workspace is just asking for data loss so why bother risking it ? A flash memory IC may pass a single, or even multi run test, but that does not mean it will not suffer from flipped bits over time.

[msuffidy]

I just got these for free because they were being offered and I agree it is perplexing why anyone would even bother to make misreported flash. I think what happened here is maybe reject 64GB sd cards got used for some sort of ECC 32GB about drive that fails past 32GB. I just feel sorry for the person who gave the 5 of them to me who claimed $200 Canadian was spent getting these on Amazon. Apparently that person had various confusions with how to use them and got something else before it even didn't work.

[Berni]

Yep watch out for any suspiciously cheap/free USB drives that are over 8GB in size. It is pretty common to see them with faked capacity.

They have the controller inside misconfigured for the wrong size on purpose and then formatted in a way that keeps data in the lower memory area. They work fine up to a few GB of data, then they start overwriting data and corrupting it.

So do use one of those flash memory testing tools. Those write junk data all over the drive and read it back for verification. So they notice if anything gets overwritten.

[gnuarm]

The other issue is reliability. If someone goes to the trouble of creating fake flash drives I do not see them using the better quality components within, including whatever flash memory IC they choose to use. I would not be at all surprised to see production line rejects (scrap) used for the flash memory. Such QC fails are not something I would want to trust my data with. Using junk flash memory is more trouble than it is worth in my opinion. If a USB stick turned out to be a fake I would chuck it in the bin without hesitation. Having such knocking around my workspace is just asking for data loss so why bother risking it ? A flash memory IC may pass a single, or even multi run test, but that does not mean it will not suffer from flipped bits over time.

I had to laugh about this.  Flash is made in several grades, based on the density.  Some years ago, they started making flash with more than one bit per cell.  That means multiple levels of voltage have to be detected.  So there is a very clear tradeoff between cost and density.  The highest density had 16 levels or 4 bits per cell, the last time I checked.  Pretty much all consumer flash that isn't sold specifically as high quality (and high dollar) flash, most likely uses this high density, low cost devices.  You will be lucky if it doesn't fail in the first couple of years.

So, no, these fake devices certainly don't use higher quality flash chips, and in fact,  may well use factory seconds, etc. 

I have some junk SD flash sticks with my company logo on them.  Seems it is so cheap to laser etch the metal wrap, that they will do it for free, even on the fake flash.

[Fraser]

That is why I still archive to multiple low hours hard disk drives and not flash based technology 

That said, I would be peeved if I put an important file on a flash drive and it got corrupted in a short period of time. Thankfully decent quality flash memory seems pretty reliable these days. Longevity of data retention before corruption is a whole topic on its own but I generally trust hard disk drives and DVD-R over flash technologies. I remember people saying CD’s would become unreadable in 10 years due to data corruption through silver layer failure. (It was new technology at the time). I have music CD’s from 1987 that still play fine thanks to them not degrading as feared and the data correction algorithms of a CD player. Flash does seem like a more ‘fragile’ data storage technology in terms of longevity though. That said, EPROM’s seem to retain their data very well and they only had a predicted 10 year data retention life. Would I deliberately place my data on a junk quality flash device though ? Nope. Quality Flash storage is so cheap now that I would not waste the time risking it. I bought a brand new retail WD 510 2.5” 500GB SSD for £31 the other day and flash memory sticks are at low prices as well. We are indeed fortunate that these technologies have reduced in price as SSD’s, in particular, can transform the user experience on computers that previously used hard disk drives. Just buy decent quality 

[Berni]

Longevity of data retention before corruption is a whole topic on its own but I generally trust hard disk drives and DVD-R over flash technologies. I remember people saying CD’s would become unreadable in 10 years due to data corruption through silver layer failure. (It was new technology at the time). I have music CD’s from 1987 that still play fine thanks to them not degrading as feared and the data correction algorithms of a CD player. Flash does seem like a more ‘fragile’ data storage technology in terms of longevity though. That said, EPROM’s seem to retain their data very well and they only had a predicted 10 year data retention life.

Don't put mass produced pressed CD-ROMs in the same bucket as CD-R. The ROM type have the data physically pressed into the plastic that is then aluminium coated. These are very robust and only really get age damaged by oxidation of the aluminium layer (but this is very rare, as it is protected).

The CD-R kind uses a die material inside that gets bleached by the drives laser to write data. This die can fade with time much like store receipts from thermal printers can fade. I did personally have some very old CD-R discs slowly become harder and harder to read by the drive (very slow, lots of retries)  until it would start giving up and throwing errors when reading certain files. Perhaps the newer DVD-RW or DVD-RAM technology is less prone to fade since there the data layer material goes trough some sort of crystalline structure change.

But yeah for long term data archival magnetic storage is the way to go.

[Fraser]

Good point, but I was only giving audio CD’s as an example of a technology that ‘experts’ claimed would fail to read due to degradation of the aluminium coating over the years. All sorts of scare stories were published but it never happened to the extent that it was foretold. I totally agree that writable optical media, being an organic dye, may well be susceptible to age related degradation. My main archive remains on several hard disks with correct data duplication in case a drive fails. I do not bother with a RAID NAS as I have had to repair a few of those with controller failure over the years !

I have driven this thread off topic… sorry to the OP !

I am pleasantly surprised at the lowering of price per GB for flash storage, both flash memory sticks and SSD. The cost incentive to risk buying no-name or less well known brands of flash products on eBay has evaporated for me.

[ataradov]

it is perplexing why anyone would even bother to make misreported flash.

You can go on AliExpress now and find "1TB", or event "2TB" flash drives for like $15. People buy them, so they are making them.

And I assume the batch you've got was people buying and then realizing that they were fake.

I think what happened here is maybe reject 64GB sd cards got used for some sort of ECC 32GB about drive that fails past 32GB.

There are not that many rejects, they just use whatever cards they can get and program them to whatever size they want.

on Amazon.

Amazon is basically overpriced AliExpress at this point. The best thing you can do is buy from reputable brands. Those are still faked, of course, but a complete no-name is guaranteed to be junk.

[Monkeh]

I do not bother with a RAID NAS as I have had to repair a few of those with controller failure over the years !

This is why you use mechanisms which are independent of the hardware - mdraid, LVM, ZFS.

[AndyBeez]

I might suspect that being for 'Windows' , 'Android' and 'Apple', the only file format that is compatible with all is  FAT/ExFAT. So internally, the controller may only support this file type?

[gnuarm]

I might suspect that being for 'Windows' , 'Android' and 'Apple', the only file format that is compatible with all is  FAT/ExFAT. So internally, the controller may only support this file type?

By "controller", do you mean the part in the flash device?  That doesn't know anything about the file system format.  The file system is completely independent of the hardware.

[AndyBeez]

I have never managed to format a SD card as NTFS or even APFS. Internally, those sticks look like something from the dark side of the bargain bin. Remarkably someone sat down and designed them. Even making an injection mold for the case and fancy retail packaging. Who are these people? I would be tempted to send one in to Dave's Mailbag.

Edit - I wonder what those sticks are reporting to the USB bus with and without the flash fitted? In Linux use the lsusb command or cat the file /sys/kernel/debug/usb/devices
Odds are the VID PID values are zeros or 'borrowed' from another legitimate production run.

[msuffidy]

I was doing more tests on these sticks. They seem reliable at a binary level to around 45GB then I think the writes go nowhere until it clobbers itself again at a higher bit. My guess is that it is a 64GB sd card that has extra bits for ECC. It may have been originally intended to be 32GB sticks and then some scammers reprogrammed them to appear to be larger sizes. I will diff everything I write for a while just to see but I am partitioning them as 32GB ntfs sticks, and they seem to work fine that way.

[eutectique]

Will they hold charge in 2 weeks time? In 3 months?