SSD Revisions of same product show dissimilar and lacking performance

[rezqme]

 

Not-So-Solid State: SSD Makers Swap Parts Without Telling Us

Buying SSD storage can be an uncalculated risk

 

Author: Sean Webster

Originally posted: Dec 1, 2020

 

Summary

 In a review of ADATA's 1TB M.2 XPG SX8200 Pro SSDs Tom's Hardware writer Sean Webster found disturbing variances in performance within the model's revisions (V1, V2, V3). The deeper dive found that many components and firmware settings were different among the revisions that yielded slower performance (especially in write performance) across the benchmark suite conducted.

 

Quotes

Quote

"For both of the newer SX8200 Pro variants we tested, the revisions performed much slower than the original. The v2 variant was close in terms of sequential performance but fell short in every test thrown at it."

 

My thoughts

The article takes on a great point that drive manufacturers (in this case ADATA) need to be greater transparency when they make hardware changes in their products. The storage universe is a great example of how hardware changes can affect the consumer in not-so friendly ways, remember the Western Digital Red Drives going to SMR write technology and not getting a new label? Many of us may already be aware of how the NAND flash memory market has evolved over the years with SLC>MLC>TLC> and now DRAMless QLC drives getting larger and cheaper but not necessarily "better" but for those new to the scene this can be confusing and prohibitive to adoption or purchase satisfaction. 

 

I enjoyed that the author is well versed enough to discuss the finer intricacies of the situation being a product of the supply chain of current parts available and would read his work again. 

 

Sources

https://www.tomshardware.com/news/adata-and-other-ssd-makers-swapping-parts

[Rune]

Not too surprising from Adata. Really, most companies are pretty scummy these days when they know they can get away with it.

[HannahLewis]

I remember this being a problem with Kingston V300 SSD's back in the day too. 

[TrigrH]

It's super BS, you make a top tier v1, throw it out to all the reviewers. Then downcost the unit and give the consumer an inferior product. 

[GDRRiley]

nothing new. done before by larger companies.

[Blademaster91]

Its not new, companies have done that before, but for a high end SSD, I think it's misleading and IMO the difference should be made obvious.

[Mark Kaine]

this is essentially true of "cheapskate brands" 

 

There's a reason why Samsung and co cost more, and that's reliable quality. 

 

6 minutes ago, Blademaster91 said:

for a high end SSD

I wouldn't call Kingston high end but ok! 

[TrigrH]

1 minute ago, comander said:

It's not going to matter much unless it's a caching drive in a server application... and this device shouldn't be used as such for a multitude of reasons. 

 

If you're defining performance based on SLAs related to application load times... virtual tie. Similarish story for loading a directory in windows. There will be some cases involving large file transfers where it'll be slower but for most people that's rare. 

 

If you are in a situation where performance and responsiveness matter THAT much... you probably want to tier an optane drive (118+ GB) ahead of this and/or some RAM. 

It's not the point though. How is a savvy consumer meant to know what the heck they are buying when companies pull this shit?

 

You never know someones use case, its kinda like SMR / CMR all over again.

 

There should be regulations in-place preventing drastic product changes without a clear change of product name.

[Taf the Ghost]

Most of the SSD companies do this. It's BoM tricks.

[Taf the Ghost]

1 hour ago, Abyssal Radon said:

While this is super shitty of Adata, surely there are other companies behaving similarly? This is why I have a couple of cheap ass SSD's that has nothing but game's on them and really don't need the 'ultimate' performance for loading games. 

Most do. Only one that really isn't is Samsung with their top-line parts.

 

Main thing is that the SX8200 Pro is one of the best consumer SSDs, so it gets looked at a lot more.

[Elisis]

4 hours ago, Rune said:

most companies are pretty scummy these days when they know they can get away with it.

most all companies

[Bombastinator]

5 hours ago, Rune said:

Not too surprising from Adata. Really, most companies are pretty scummy these days when they know they can get away with it.

These days?  They have gotten worse I suppose.  There was a law instituted a few years ago the effect of which is that it more or less requires a CEO of a publicly traded company to behave in the most absolutely venal way possible or s/he can be sued by a shareholder.  

[Dabombinable]

My SX8200 Pro failed. My replacement had the slightly slower controller. Still no where near as bad as the Kingston SSD Now V300 debacle. Inferior controllers and NAND, with performance being flat out worse unless the files were highly compressible.

[RejZoR]

Revisions these days don't mean product improvements, but the level they can cheap out on product and still get away with it. And since all reviews are usually done with Revision 1 of any product, people base their purchases on those reviews that don't necessarily reflect the performance of later revisions. It's pretty scummy tacting, but works in favor of companies and not consumers.

[Letgomyleghoe.]

I don't think it will really matter but that kinda scummy as I bought a sx8200 pro for 100 on best buys website on black friday.

[Kisai]

1 hour ago, RejZoR said:

Revisions these days don't mean product improvements, but the level they can cheap out on product and still get away with it. And since all reviews are usually done with Revision 1 of any product, people base their purchases on those reviews that don't necessarily reflect the performance of later revisions. It's pretty scummy tacting, but works in favor of companies and not consumers.

 

I'm not sure how true this is.

 

Between all the laptops I've done maintenance on, the SSD's of the same brand (Hynix, Samsung, Toshiba) all perform similarly if they're the same capacity and type (eg XG50 != XG60) and these are the ones I see the most often. Dell also has firmware updates for Western Digital, Seagate, Micron, and Sandisk SSD's but I've yet to see these in a stock laptop.

 

The only thing that is really true is that Samsung never updates their firmware on the drives they sell OEM's. So if you have a Samsung PM981, no firmware update will ever exist. Toshiba has had the most firmware updates for their drives.

 

Machines with bigger SSD's are faster, and stay faster longer, and usually people working out of the last 10% of the drive complain about slowness the most frequently.

 

Adata, AFAIK, mixes other vendors parts, like think about those store brands you see in stores like best buy. If you peel the labels on their parts you'll probably find different brands on the part with the same part name. But that's just one of the reasons why I've avoided this brand except for throw-away parts.

 

 

 

[williamcll]

Looks like @VEXICUSSSD List may need a revision too then.

[Sarra]

6 hours ago, Mark Kaine said:

this is essentially true of "cheapskate brands" 

 

There's a reason why Samsung and co cost more, and that's reliable quality. 

 

I wouldn't call Kingston high end but ok! 

This makes sense when applied to SSD's, but not HDD's. I've got a dieing Seagate 3TB drive with very little time on it, and I'm pretty disappoint. It's out of warranty, since it's been sitting on a shelf for most of it's life, waiting for me to build a PC to put it in, and now... Clicking. Constant, repetitive clicking.

 

Anyway... Back in the day, I went for the Marvell controllers, since they had better reviews. Now, I just go for... Well, I'm not sure. I just picked my Corsair MP600 because it was in stock. And... Fast. I don't know, I wouldn't even consider an Adata anything. The reviews and such I looked at when going for an NVMe drive were all not very good, and that turned me away from it, even with the lower price.

 

The 'cost paring' isn't unique to SSD's, though. Auto manufacturers do this. Mercedes Benz sold a car with brake by wire; each caliper was two separate braking systems, with redundant hydraulic pumps, redundant wiring, extra sensors, etc., with full manual hydraulic backup (brake by hydraulic, ie, conventional brakes). By the time Toyota makes brake by wire in a freaking Prius, or whatever, it'll be the cheapest, easiest to mass produce, and least redundant system possible, until a few people die, then they'll add some token redundancy and call it good.

 

At least, with an SSD, all you're losing is some data, not your life...

[Dabombinable]

16 minutes ago, Sarra said:

This makes sense when applied to SSD's, but not HDD's. I've got a dieing Seagate 3TB drive with very little time on it, and I'm pretty disappoint. It's out of warranty, since it's been sitting on a shelf for most of it's life, waiting for me to build a PC to put it in, and now... Clicking. Constant, repetitive clicking.

 

Anyway... Back in the day, I went for the Marvell controllers, since they had better reviews. Now, I just go for... Well, I'm not sure. I just picked my Corsair MP600 because it was in stock. And... Fast. I don't know, I wouldn't even consider an Adata anything. The reviews and such I looked at when going for an NVMe drive were all not very good, and that turned me away from it, even with the lower price.

 

The 'cost paring' isn't unique to SSD's, though. Auto manufacturers do this. Mercedes Benz sold a car with brake by wire; each caliper was two separate braking systems, with redundant hydraulic pumps, redundant wiring, extra sensors, etc., with full manual hydraulic backup (brake by hydraulic, ie, conventional brakes). By the time Toyota makes brake by wire in a freaking Prius, or whatever, it'll be the cheapest, easiest to mass produce, and least redundant system possible, until a few people die, then they'll add some token redundancy and call it good.

 

At least, with an SSD, all you're losing is some data, not your life...

Word of advice with NVME SSD (especially Adata ones): Never BCLK overclock - the controller will die. Adata was lying when they were saying that the SX8200 Pro was good for overclockers / overclocking.

[LAwLz]

I remember this happening when NAND flash moved from the 34nm to the 25nm manufacturing process.

A lot of SSD companies updated their models to use 25nm flash, which was slightly slower than the 34nm versions. Some companies like Corsair decided to change their model names slightly, but some did not.

 

https://www.corsair.com/ca/pl/blog/force25nm

[Moonzy]

-kingston v300 flashback-

 

pitchfork time?? no?

[Dabombinable]

1 hour ago, Moonzy said:

-kingston v300 flashback-

 

pitchfork time?? no?

The difference in benchmarks is very minor...and its only the controller not the NAND. The NAND Kingston switched to was vastly inferior to what reviewers got.

[leadeater]

Such is why I only buy Samsung SSDs. Pitty as I thought the SX8200 Pro was actually extremely good and very very good value.

[StDragon]

5 hours ago, Sarra said:

The 'cost paring' isn't unique to SSD's, though. Auto manufacturers do this. Mercedes Benz sold a car with brake by wire; each caliper was two separate braking systems, with redundant hydraulic pumps, redundant wiring, extra sensors, etc., with full manual hydraulic backup (brake by hydraulic, ie, conventional brakes). By the time Toyota makes brake by wire in a freaking Prius, or whatever, it'll be the cheapest, easiest to mass produce, and least redundant system possible, until a few people die, then they'll add some token redundancy and call it good.

 

At least, with an SSD, all you're losing is some data, not your life...

General Motors is the worst about it. Cents on the dollar of material saved that literally costed human lives; all 124 of them. This caused 30 MILLION recalls of which also they settled out $900,000,000. And to add insult to injury, they knew of the problem a full decade before the recall!!

 

So yeah, caveat emptor (buyer beware). And now you've been made aware of how shitty a company GM is.

 

...but at least ADATA didn't cause a fatality because a game loaded a little slower. /s

 

 

[DoctorNick]

14 hours ago, f1ght43v3r said:

I remember this being a problem with Kingston V300 SSD's back in the day too. 

I do too. Also I only remember V300, because of it's many problems.