Question FAKE Samsung 970 NVME??

[fzabkar]

Haha, there's nobody on the planet as specialized in "research" like the Chinese. They've gone so far as "researched" early Intel 386 cpu's and recreated them (and modified), successfully.

True, but research in data recovery involves a lot more than reverse engineering. Reverse engineering is necessary to find out how the device works, or why it doesn't work. That's only the beginning. The hard part is to come up with techniques and solutions to repair or undo the damage.

Just to give you an idea as to what tools are available ...

PC3000 - Russia​

MRT Lab - China​

DFL (Dolphin) - China​

UFS Explorer - Ukraine​

Cellebrite (mobile phones) - Israel​

VNR (Rusolut) - Russia​

Flash Extractor - Russia​

​

WD Marvel - Ukraine​

Sediv - Russia​

SHTR (Samsung) - Kazakhstan​

HDDSuperClone - USA​

​

BTW, these guys are arguably the best known and most widely respected [data recovery] researchers in the US:

https://thedigilab.com/about-us.html

 

[Karadjgne]

Ya know, I shouldn't be surprised by all that. Every spy movie made about the Cold War was about the US designing/building something, and the CIA having to chase down the Russian spy who had stolen the plans so that Russia could build it bigger and better. Even Armageddon was about that.

So while most are designing the 'next best thing' Russia's best have spent all their energy in reverse engineering the wheel. In Illinois.

 

[fzabkar]

Ya know, I shouldn't be surprised by all that. Every spy movie made about the Cold War was about the US designing/building something, and the CIA having to chase down the Russian spy who had stolen the plans so that Russia could build it bigger and better. Even Armageddon was about that.

So while most are designing the 'next best thing' Russia's best have spent all their energy in reverse engineering the wheel. In Illinois.

Political comment removed by moderator

Ask yourself, why does WD refer data recovery requests to its "partners", all of whom would undoubtedly be using PC3000?

 

[USAFRet]

This is starting to deviate into territory best avoided.

 

[fzabkar]

This is starting to deviate into territory best avoided.

I have no political agenda. I'm just telling like it is. Fact -- there are no firmware-level data recovery tools produced in the USA.

 

[Karadjgne]

Well that link for digi, it's based in Illinois. No polital agenda here either, just wondering why all the energy spent trying to reverse engineer stuff when ppl like Samsung could simply hand over the design and several prototypes (and a paycheck) and ask dudes to find any issues before they get released and require firmware updates to 'fix', if they are truly fixed. That'd be a bonus both ways, dudes spend less time figuring what's what and get a solid basis for figuring their data recovery etc, Samsung gets a heads up and can release a solid product, and everybody wins.

 

[Darkbreeze]

As already stated, I think people are seeing wolves where there are only sheepdogs. Even if there IS a wolf skittering out around the edges somewhere, it doesn't make everything that's on four legs and isn't a sheep, a wolf.

 

[fzabkar]

Well that link for digi, it's based in Illinois. No polital agenda here either, just wondering why all the energy spent trying to reverse engineer stuff when ppl like Samsung could simply hand over the design and several prototypes (and a paycheck) and ask dudes to find any issues before they get released and require firmware updates to 'fix', if they are truly fixed. That'd be a bonus both ways, dudes spend less time figuring what's what and get a solid basis for figuring their data recovery etc, Samsung gets a heads up and can release a solid product, and everybody wins.

The DigiLab guys are Russian expats, all educated in Russia. Modern data recovery seems to have had its roots in Russia. HDD Guru is a professional, English language, DR forum run by a Russian guy. The data recovery world is pretty much dominated by Russian developers/researchers/hackers etc. It's crazy, but that's the way it is.

As for your suggestion re Samsung, I'm sorry, but it just made me laugh. Samsung won't even release datasheets for their storage ICs, let alone their treasured, tightly guarded IP. And it's not just Samsung that does this. Marvel is another chip maker that comes to mind. In fact, I challenge you to pick any HDD or SSD PCB at random and then locate datasheets for the major ICs.

This is the closest I've come to locating a current part number decoder for Samsung's NAND flash, let alone a datasheet:

https://github.com/iTXTech/FlashDetector/blob/master/FlashDetector/src/iTXTech/FlashDetector/Decoder/Samsung.php

 

[Pextaxmx]

there are lots of Samsung 128 layer NAND failure reports on the internet, and new cases are still coming in. (simply google "870 EVO failure") At this point I don't think anyone can argue anymore that those reports are just bunch of anecdotal cases.
However though, considering the fact that Samsung has the biggest market share, the failure rate can't be as big as, say, the "Seagate 3TB" crisis back in 2013 - in which case the issue must have been surfaced in a much bigger/louder way.

I am trying to say, even if we assume someone's 18 970 EVO plus drives were all 128 layer version in question, 18 samples are still too little to disprove anything.
I wasn't following 840 EVO planar TLC issues but the same pattern must be applicable - problem existed but the use cases that were actually affected by the problem must have been a small fraction of the entire volume which was manageable by Samsung.

 

[fzabkar]

This part of that 980 Pro reddit thread is "curious":

Many users in China complain about loss of data due to bad blocks in Samsung 980 Pro. It's now reported by mainstream media in China. This usually happens when the SSD has been used for 6-12 months. Samsung in China allegedly issued a statement but it was quickly pulled off.

That "mainstream media" URL is now dead. Here it is at the Wayback Machine:

https://web.archive.org/web/20220921232444/www.cnbeta.com/articles/tech/1313507.htm
https://web.archive.org/web/20220921232444im_/https://img.lancdn.com/landian/2022/09/95381-3.png

 

[fzabkar]

This 2TB 980 Pro was used in Chia mining and wrote 3.54PB before "failing":

https://www.reddit.com/r/chia/comments/somv5o/samsung_2tb_980_pro_is_finally_reporting_failed/

Media and Data Integrity Errors = 0​

Available Spare = 100%​

 

[Karadjgne]

Which in unto itself is interesting. China having a wave of failures, but then again Chia was rampant in China and 2.65PB isn't exactly a small number at the levels of usage Chia hits. It could very well have been overly high temps as well, which isn't good for NVMe. Or even an unreported power surge, or an uber craptastic psu with lousy voltage regulation. A lot of supposition, but not exactly wild thoughts. Too many possibilities. Having 0 errors and 100% available spare just means the controller failed, not the nand.