ADATA NVMaybe SSDid it again

[PeachGr]

Summary

Acording to Nada from techtesters (youtube channel), ADATA sent her a review sample that performed differently than a sample that she bought from the shelves herself as a consumer.
The shelf unit was a bit older, but run the same firmware, had the same cooling, same memory, but different stickers, proving that the reviewr sample was made in Taiwan and the retail made in China

The overall performance looked similar and the differences were within the margin of error, until she got to the consistency where the performance was way off.
As it appears on the graph, the reviewer sample was about 3.5 faster than the retail one

 

As a reminder, ADATA was the manufacturer who started the convertation on the subject of manufacturers changing components, after their initial review
I will have the LTT video on the sources

 

It is of course, just one reviewer and just two pieces of hardware, but it is hard to follow a product like that and buy many of them over months or years to chech if the manufacturer is lying or keeping the components uncchanged

 

Quotes

Quote

 On the coinsistency test, the original sample performed extremly well, comming second [...] but the retail model performed really poorly, sitting nearly at the bottom of the graph, alongside cheeper ssds
I wanted to make sure that it is correct, we ran the test several times with both drives after a clean format, also on another system to exlude any of the other parts, and we kept getting the exact same resault

 

My thoughts

Truth is that consistency is the most irrelevant for most people, because not many of us use huge loads of big files, transefering for hours and hours, but this is maybe only the beggining, and the other parts will soon change, making the nvme slower on things that matter for other use cases
Nvme ssdz are the parts that i, and i believe not many of us check after we install them on our systems, mostly because it's hard timewise to do so, and alongside with PSUz, we are expected them to just work and don't die, but those stories, makes it hard to trust the products and the reviewers.
I guess that we actually need the Lab to test those things

 

Sources

 

 

[Mojo-Jojo]

Thoughts.

 

First: if the device has the same controller, flash chips and can run the same latest firmware, then that points to something other than mal-intent. It's either a defect or some other outlier.

 

Second: They don't say whether they've haven't contacted ADATA to confirm these results. This makes the video untrustworthy to me. They bring out a statement about alleged shady behaviour without verifying their test results with the manufacturer first. Until other testers corroborate these findings, I think it should be treated as an outlier. The only hard data they currently have to accuse ADATA of shady behaviour, is past behaviour.

Edited July 25, 2022 by Mojo-Jojo

[LAwLz]

6 minutes ago, PeachGr said:

As a reminder, ADATA was the manufacturer who started the convertation on the subject of manufacturers changing components, after their initial review

Small correction, ADATA was not the manufacturer who started this.

This has always been an issue. ADATA just happened to be the one who got caught recently. I remember when Corsair were caught slightly changing their SSDs in ~2011. Although I think they slightly changed the SKU name back then.

 

 

Anyway, I am not entirely sure what the "consistency test" in PC Mark 10 tests, but before bringing out the pitchforks it might be worth looking into if this actually matters. I wasn't able to find that much info about how the "consistency test" is done, but that is a very important detail in determining how good/bad this drive is.

It might be that ADATA has secretly downgraded the NAND on the drive without saying anything, and that results in very poor performance in certain tests. Those tests might be important, or maybe it is something barely anyone will notice. Maybe this is just a bug that gets fixed in some future firmware release. Probably not, but who knows.

 

Facts before pitchforks.

[PeachGr]

10 minutes ago, Mojo-Jojo said:

They don't say whether they contacted ADATA to confirm these results

They actually will 7:33 on the video

 

2 minutes ago, LAwLz said:

consistency test

it is the transfer of big files for many hours. It's common for server workloads, and not relevant to gaming or other personal usecases. But still, it's something different than the addvertised specs, so i don't know what else will be different over the next months or years

[Mojo-Jojo]

2 minutes ago, PeachGr said:

They actually will 7:33 on the video

Right. But they haven't. They decided to release a video, accuse ADATA of shady practices and knowingly cause a commotion, without having all the facts.

If they turn out to be right then they're right. I just don't believe this is the proper way to handle it.

[Dabombinable]

56 minutes ago, LAwLz said:

Small correction, ADATA was not the manufacturer who started this.

This has always been an issue. ADATA just happened to be the one who got caught recently. I remember when Corsair were caught slightly changing their SSDs in ~2011. Although I think they slightly changed the SKU name back then.

 

 

Anyway, I am not entirely sure what the "consistency test" in PC Mark 10 tests, but before bringing out the pitchforks it might be worth looking into if this actually matters. I wasn't able to find that much info about how the "consistency test" is done, but that is a very important detail in determining how good/bad this drive is.

It might be that ADATA has secretly downgraded the NAND on the drive without saying anything, and that results in very poor performance in certain tests. Those tests might be important, or maybe it is something barely anyone will notice. Maybe this is just a bug that gets fixed in some future firmware release. Probably not, but who knows.

 

Facts before pitchforks.

Kingston did the same thing several years ago as well. There were many different revisions each performing differently + worse than review units.

[Quackers101]

everyone has been caught doing sh*t, so wouldn't think that ADATA would be any different as with their previous case.

All of them are lying and sadly some change the specs, which sucks a lot when one might pay a premium for it.

 

Not sure some big cases happened 1-2 years ago with the LTT video too, to some bad defects being used or returned.

[leadeater]

This is why I still only buy Samsung, because when they do it this is how they do it

 

Quote

On a hardware level, the original 970 Evo Plus employs Samsung's Phoenix controller (S4LR020) and 92-layer 3D TLC NAND flash with the K9DUGY8J5B-DCK0 identifier. The new 970 Evo Plus, on the other hand, utilizes the Elpis controller (S4LV003). 

 

For those not familiar with Samsung SSDs, the Elpis controller is the one that powers Samsung's 980 Pro SSD. Although the Elpis controller is designed for the PCIe 4.0 x4 interface, it's also backwards compatible with the PCIe 3.0 x4 standard, which is why Samsung can recycle it for the 970 Evo Plus. Basically, the new 970 Evo Plus is like a 980 Pro, but without the PCIe 4.0 speeds

https://www.tomshardware.com/news/samsung-is-swapping-ssd-parts-too

[bcredeur97]

this wouldn't be as a big of a deal if it wasn't such a huge difference. my gosh.

 

it's a totally different product at that point

[wanderingfool2]

9 hours ago, LAwLz said:

Anyway, I am not entirely sure what the "consistency test" in PC Mark 10 tests, but before bringing out the pitchforks it might be worth looking into if this actually matters. I wasn't able to find that much info about how the "consistency test" is done, but that is a very important detail in determining how good/bad this drive is.

https://support.benchmarks.ul.com/support/solutions/articles/44002171488-drive-performance-consistency-test

Pretty much lays out how consistency test is calculated.

 

The tl;dr is they fill up the entire drive, then they do it again to try making sure all the blocks are written to.  They then slam it with a 50 GB or a minute to prevent cleanup and then proceeds to do a few more tests.

 

I'm guessing that it's the 50GB tests where they overwhelm the controller so it can't do maintenance that is the part that is tanking the score.  With my thought maybe the flash isn't responding as quickly as the test sample.  Either way though, most people wouldn't get into this scenario...except I guess if the drive is super full, or after it has aged and had a ton of writes on it.

[williamcll]

Different memory controllers I assume?

[Senzelian]

8 hours ago, leadeater said:

This is why I still only buy Samsung, because when they do it this is how they do it

And what's with the manufacturers that haven't done it at all? Why do you not buy from them?

[Dabombinable]

20 minutes ago, Senzelian said:

And what's with the manufacturers that haven't done it at all? Why do you not buy from them?

Samsung has extremely high reliability, long warranties and overall good quality. You can't go wrong with their SSD - pretty much a rule for now (there is always the chance that they'll go OCZ levels of bad - MLC with lower endurance than theoretical QLC NAND of the same capacity was absurd).

[leadeater]

1 hour ago, Senzelian said:

And what's with the manufacturers that haven't done it at all? Why do you not buy from them?

Who? Ah.... nobody? I can't think of any other company that hasn't or wouldn't. Intel maybe but I'm very sure they have on their lower end products, but even on higher end datacenter SSDs they've made some changes without changing product model number that I would question wasn't only firmware updates since known issues were not back fixed, see Linus's bad time with the SSD's he purchased.

 

Samsung was basically believed to be the only one not doing it since they made the controllers, DRAM and NAND themselves and most of these parts were common between consumer and datacenter products. It took silicon shortages for Samsung to do it and they up spec'd the controller rather than down spec'ing it which is the customer first way to do it.

 

But if you know of a brand that hasn't done it, and you for sure know, I'd be quite interested. Companies such as Sabrent excluded since they only entered SSD market in 2018 and aren't exactly on my "existed long enough to trust you" books, regardless of if they have or have not which I have no idea. I'd say not since I doubt they've had enough supply pressure, but who knows. 

[leadeater]

1 hour ago, williamcll said:

Different memory controllers I assume?

More likely NAND, fewer larger NAND packages. Worst case change in NAND type.

[PeachGr]

About who to trust, Samsung has the fab to produce the nand, I think the controller as well, and I don't know about the PCB, they are also the falinal producer of the overall hardware. I don't believe they have the absolute 100% of the product, but they sure have a huge percentage of it, so I guess they are to be trusted. It's a guess, but they wouldn't slaughter their own products 

[Tan3l6]

Can't get the damn PCMark 10 to run either in Steam, nor  separate installation. Otherwise I could confirm or deny the case with my S70 Blade.

[LAwLz]

Since we are on the topic of "which brand to trust", I really wish this forum and people in general would stop generalizing brands so much.

Just because ADATA potentially downgraded one of their SSD SKUs does not mean all SSDs from the company are bad. Another example would be that just because Nvidia's highest end 3090 Ti uses a lot of power does not mean the 3060 does too. Or just because AMD's Ryzen 9 processor might be a better buy than Intel's i9 does not mean the Ryzen 5 is a better buy than the i5.

 

People need to stop forming opinions of an entire company based on a single product. Judge products based on their own merits, not some particular SKU.

 

I am kind of a Samsung storage fanboy. Up until recently I had 4 storage devices in my PC, all from Samsung. I recently swapped out two of my Samsung SSDs for a bigger SSD and as it turns out, the replacement drive was also from Samsung. I didn't choose the Samsung 970 Evo Plus as my replacement drive because it was from Samsung. I chose it because it was the best drive for me, based on availability, performance (verified by third parties), price and features/spec.

 

Samsung makes a ton of great SSDs. They have a few not-so-great ones as well but even those have their strengths. But it is very important to note that their drives are good because the product itself is good, not because the drive happened to have a Samsung sticker on it. Same goes for all products. 

[leadeater]

30 minutes ago, PeachGr said:

About who to trust, Samsung has the fab to produce the nand, I think the controller as well, and I don't know about the PCB, they are also the falinal producer of the overall hardware. I don't believe they have the absolute 100% of the product, but they sure have a huge percentage of it, so I guess they are to be trusted. It's a guess, but they wouldn't slaughter their own products 

Everything that matters Samsung makes, PCB can come from anywhere as a change in supplier doesn't matter. Resistors and caps won't matter here either.

 

There's only two major parts that matter performance wise, the controller and NAND. DRAM kind of but not a lot, Samsung makes and supplies themselves all 3.

 

Which to be spin the other way means if they start screwing around it's actually worse of them because they basically have the most ability to not do it.

[leadeater]

4 minutes ago, LAwLz said:

Samsung makes a ton of great SSDs. They have a few not-so-great ones as well but even those have their strengths. But it is very important to note that their drives are good because the product itself is good, not because the drive happened to have a Samsung sticker on it. Same goes for all products.

Far as I'm concerned because they are made by Samsung then they are good. Every SSD I look at I compare back to something from them, I wish comments also weren't assumed and painted as "blind trust" by you.

 

You can have strong faith in a brand and express it without having the carry the label of blind trust, thanks.

 

Until Samsung give me a reason not to, they'll be the only ones I purchase from. On the slim chance something is better then I'll buy that, but I do not factor price unless it's a huge margin of difference because I simply am not willing to switch just because something is 10% cheaper. 

[Quackers101]

13 minutes ago, LAwLz said:

Since we are on the topic of "which brand to trust", I really wish this forum and people in general would stop generalizing brands so much.

Just because ADATA potentially downgraded one of their SSD SKUs does not mean all SSDs from the company are bad.

 

People need to stop forming opinions of an entire company based on a single product. Judge products based on their own merits, not some particular SKU.

Would say, just dont know them, and just dont trust them to begin with. Also they had more issues than being a "one off", maybe they have good products, also it would be hard to judge these products when one can get a totally different product or change by time what they sell. like when you get a long warranty with the big like samsung or others.

[LAwLz]

20 minutes ago, leadeater said:

Far as I'm concerned because they are made by Samsung then they are good. Every SSD I look at I compare back to something from them, I wish comments also weren't assumed and painted as "blind trust" by you.

 

You can have strong faith in a brand and express it without having the carry the label of blind trust, thanks.

 

Until Samsung give me a reason not to, they'll be the only ones I purchase from. On the slim chance something is better then I'll buy that, but I do not factor price unless it's a huge margin of difference because I simply am not willing to switch just because something is 10% cheaper. 

I think that Samsung and SSDs is an exception to the rule, but even then Samsung has better and worse products. Just because the 980 Pro is a fantastic drive does not mean the 860 QVO is as good.

I totally understand why you trust Samsung so much regarding SSDs, but I would still argue that we should judge products by their merits and not the sticker.

[Senzelian]

1 hour ago, leadeater said:

Who? Ah.... nobody? I can't think of any other company that hasn't or wouldn't. Intel maybe but I'm very sure they have on their lower end products, but even on higher end datacenter SSDs they've made some changes without changing product model number that I would question wasn't only firmware updates since known issues were not back fixed, see Linus's bad time with the SSD's he purchased.

 

Samsung was basically believed to be the only one not doing it since they made the controllers, DRAM and NAND themselves and most of these parts were common between consumer and datacenter products. It took silicon shortages for Samsung to do it and they up spec'd the controller rather than down spec'ing it which is the customer first way to do it.

 

But if you know of a brand that hasn't done it, and you for sure know, I'd be quite interested. Companies such as Sabrent excluded since they only entered SSD market in 2018 and aren't exactly on my "existed long enough to trust you" books, regardless of if they have or have not which I have no idea. I'd say not since I doubt they've had enough supply pressure, but who knows. 

 

I'm not gonna check each individually now, but there's gotta be a few that haven't been involved in such shady practices.

People tend to always jump to the most obvious choices like Samsung, but there are plenty of manufacturers out there:

Angelbird

Corsair

HGST

HP

Intel

Kioxia / Toshiba

Lexar

Micron / Crucial

Mushkin

Patriot

PNY

Sabrent

Silicon Power

SK Hynix

Supermicro

Transcend

 

And these are just the most common ones. There are at least double the amount. 

[leadeater]

6 minutes ago, Senzelian said:

 

I'm not gonna check each individually now, but there's gotta be a few that haven't been involved in such shady practices.

People tend to always jump to the most obvious choices like Samsung, but there are plenty of manufacturers out there:

Angelbird

Corsair

HGST

HP

Intel

Kioxia / Toshiba

Lexar

Micron / Crucial

Mushkin

Patriot

PNY

Sabrent

Silicon Power

SK Hynix

Supermicro

Transcend

 

And these are just the most common ones. There are at least double the amount. 

I didn't want or need a list of SSD makers, I asked for a list that hasn't made undisclosed changes to parts for a product. Doesn't have to be good or bad. I would not be willing to say any of them have not done it ever.

 

Edit:

Also parts changes are both fairly normal and not automatically bad.

[Quackers101]

7 minutes ago, Senzelian said:

I'm not gonna check each individually now, but there's gotta be a few that haven't been involved in such shady practices.

People tend to always jump to the most obvious choices like Samsung, but there are plenty of manufacturers out there:

Angelbird

Corsair

HGST

HP

Intel

Kioxia / Toshiba

Lexar

Micron / Crucial

Mushkin

Patriot

PNY

Sabrent

Silicon Power

SK Hynix

Supermicro

Transcend

between brands and the creators, like samsung and others.

like maybe above half of those above there had some similar issues or done some similar stuff with previous reportings. if its speed or other, think LTT did some on corsair and the Dram-less ones?