Et tu, Samsung? Samsung Too Changes Components for their 970 EVO Plus SSD

[Raevenlord]

A number of manufacturers have been caught red-handed, so to speak, by changing components on their SSD products without as much as a product specifications change. This has happened in the past with ADATA, Patriot, and more recently with Western Digital and its WD Blue SN550; now, it's Samsung that's being on the receiving end of a more attentive look at their recent batches of the 970 EVO SSD - particularly its 1 TB configuration.

According to Computerbase, a YouTube channel in Asia seems to have first noticed the difference. They've tested the older version of Samsung's 970 Evo Plus 1 TB (product number MZVLB1T0HBLR, April 2021 production date, a Phoenix controller [S4LR020] and 96-layer 3D TLC NAND) against the newer (product number MZVL21T0HBLU and is equipped with an Elpis controller [S4LV003] and 3D-NAND with the identifier K9DUGY8J5B-CCK0), which likely features different packaging and density for the same 96-layer 3D TLC NAND.





It seems that the new version of the SSD is faster in workloads up to 115 GB - the size of its revised SLC cache, meant to absorb burst of write activity at faster speeds than the remainder NAND working in TLC mode. The original SLC cache provision stood at 42 GB, which is what the original version shipped with, and whose behavior is confirmed in testing - the original release of the Samsung 970 Evo Plus exhausts its SLC cache (with the appropriate performance drop from overflowing into TLC) at around 40 GB. This performance change is a welcome one, even if it too should be updated on the drives' official specifications. The drive does however deliver lower performance once its SLC cache is exhausted - the original version kept chugging along with data writing speeds of around 1,500 MB/s after the first 40 GB were written, while the new revision drops that performance to 800 MB/s after 115 GB of writes. So users can be faced with either improved performance, or worsened performance, in this same product compared to the original. Check the results in the gallery below, where the original revision is on the left side (BIOS 2B2QEXM7), and the new revision is on the right (3B2QEXM7).



Component changes in electronics are not uncommon themselves; there are a number of products and chips that can employ several alternatives in the supply chain. this has become especially more important a safety mechanism for manufacturers with the still straining electronics supply chains, which are still reeling from Covid-19. However, sometimes changes are enacted not because manufacturers are forced to do it on account of component shortages; sometimes, they do it to save some extra dollars here and there. The problem arises when a products' performance characteristics suffers from the component change, which has been the case, every single time, for these changes in SSD components. The Samsung case is a different one - there are specific performance improvement and degradation according to workload.

Samsung differs in its peers by actually introducing a new serial number - as well as a new BIOS version - for this particular 970 EVO Plus rendition. However, would it be so hard to do this the right way again? Samsung had already done it with their 850 EVO V2 - which clearly demonstrated that two different revisions were available for the same product, with differing specifications listings as well. this is the way to not confuse customers. However, maybe Samsung felt they had no good way of doing this change: the drive isn't clearly better or clearly worse than the original revision it replaces. That said, slightly more information - or a press release from Samsung - could have mitigated these issues. Users should be able to know exactly what their money is purchasing. Check the original video below.

View at TechPowerUp Main Site

 

[GabrielLP14]

It's saying "9700" instead of 970 in the first paragraph, thanks for the head's up with this news

 

[billeman]

I think that it will actually work faster for most users. writing 115GB in one go is not something that is done often. But copying small files is something that frustrates me on NVME drives since it's not that much faster than SATA ssd's. I have checked, have the old 2TB version but actually wouldn't have minded having the new version.

 

[GabrielLP14]

I think that it will actually work faster for most users. writing 115GB in one go is not something that is done often. But copying small files is something that frustrates me on NVME drives since it's not that much faster than SATA ssd's. I have checked, have the old 2TB version but actually wouldn't have minded having the new version.

one think we also need to consider is the time it takes for the pSLC Caching algorithm to recover, if the drive still have some Static portion, it would recover immediatly, however the dynamic portion, could take a while depending on how well the firmware is configured.
Smaller buffers have the advantage of higher sustained speeds in most cases

 

[SuMMoN]

Why you exclude the fact that new revision use the same controller and PCB layout as 980 Pro from the video in this article?

 

[Space Lynx]

I love the ancient Rome Brutus reference in the title lol

 

[Hardware Geek]

I agree they should have changed the product name, but as already stated, this will likely improve performance for most people. It's refreshing to see a change that has potentially positive performance implications considered the recent trend of degrading performance without any mention by the manufacturer that it has changed the specs.

 

[rusTORK]

At least Samsung changed boxes...

 

[ThrashZone]

Why you exclude the fact that new revision use the same controller and PCB layout as 980 Pro from the video in this article?

Hi,
980's have had lots of read/. write issues hitting spec's so this might not be a good thing.

Would of been nice to call it 970 evo minus lol

 

[TechLurker]

Hi,
980's have had lots of read/. write issues hitting spec's so this might not be a good thing.

Would of been nice to call it 970 evo minus lol

So a 970 "Devo"? A slight Devolution of the drive's quality depending on use case?

 

[qlum]

Seems Samsung made the best out of their nand downgrade. However, because of the quite different characteristics, I still think it's pretty bad that they gave it the same name.

Sure, most writes will be faster considering roughly 150gb is the break-even point however, for people who regularly write above that it can be quite significant:
118 vs 147 seconds for 200gb
246 vs 384 at 400gb
roughly 30 seconds slower at 200gb
more than 138 seconds at 400gb
Where the maximum time saving the new drive this way would be just be 18 seconds

These are just my numbers plotted in google sheets, I have skimped over details / possibly contains errors,
In case anyone cares: https://docs.google.com/spreadsheets/d/1pGIGu38fhuUbchcOKSaccEU7AMIbXHyoOSGqgJevDyw/edit?usp=sharing

Another thing is temperatures, the new drive runs hotter, possibly consumes more power? not something you want in a laptop which is quite a common usecase.

Either way, I think it's generally bad to keep the sku the same with such potential regressions, even if the vast majority of users will only benefit.
Just call it the 975 EVO Plus or if you want to go the intel route the 970 EVO Plus Plus.
I know keeping things in stock and continuity is an issue but a complete redesign warrants a new name at least.

 

[yotano211]

Seems Samsung made the best out of their nand downgrade. However, because of the quite different characteristics, I still think it's pretty bad that they gave it the same name.

Sure, most writes will be faster considering roughly 150gb is the break-even point however, for people who regularly write above that it can be quite significant:
118 vs 147 seconds for 200gb
246 vs 384 at 400gb
roughly 30 seconds slower at 200gb
more than 138 seconds at 400gb
Where the maximum time saving the new drive this way would be just be 18 seconds

These are just my numbers plotted in google sheets, I have skimped over details / possibly contains errors,
In case anyone cares: https://docs.google.com/spreadsheets/d/1pGIGu38fhuUbchcOKSaccEU7AMIbXHyoOSGqgJevDyw/edit?usp=sharing

Another thing is temperatures, the new drive runs hotter, possibly consumes more power? not something you want in a laptop which is quite a common usecase.

Either way, I think it's generally bad to keep the sku the same with such potential regressions, even if the vast majority of users will only benefit.
Just call it the 975 EVO Plus or if you want to go the intel route the 970 EVO Plus Plus.
I know keeping things in stock and continuity is an issue but a complete redesign warrants a new name at least.

You have any proof of using higher power.

 

[qlum]

You have any proof of using higher power.

Sorry, I do not, I just watched the original video showcasing higher thermals. While certainly not confirmed, I would figure similar heat-dissipation from the surface, thus it would make a lot of sense if the power consumption under load would be higher.

But indeed I do not have any proof.

 

[HisDivineOrder]

And when people keep buying these things up, it'll teach these companies that not only can they charge more for the same products, they can also notably make them worse.

No one will care enough to stop buying. The pandemic/tariffs lessons these companies are being taught are going to have massive implications going forward to how businesses treat consumers.

 

[P4-630]

A number of manufacturers have been caught red-handed, so to speak, by changing components on their SSD products without as much as a product specifications change.

Maybe reasons for W1zzard to review these SSD's with changed components again?

 

[trsttte]

And when people keep buying these things up, it'll teach these companies that not only can they charge more for the same products, they can also notably make them worse.

The problem has nothing to with charging more for the same or worse product. They could charge whatever they want for the new revision as long as they announce it, so people will now it's a different product that might have different specs. Most people will think that it might be better but then, since it's a new product, reviewers can validate if it is or isn't better instead of having to ramdonly find out that under the cooling sticker the product changed!

 

[bug]

Do we have a trend here? "Give them more SLC cache, so they won't notice the flash is actually slower"?

 

[trsttte]

Do we have a trend here? "Give them more SLC cache, so they won't notice the flash is actually slower"?

for regular office work I don't think it's such a bad idea, but clearly market it as such. More intense workloads will suffer and this isn't even a low tier device where we could try to excuse a more varied performance (we shouldn't excuse it either way but anyway)

 

[wiak]

heh hopefully people will sue them for false marketing

 

[Minus Infinity]

Yet again undocumented changes not telling people of how performance will change, so they can decide for themselves if this is worthwhile to them. Have they slashed the price?

So glad I got my 2TB model early in the year.

 

[Crackong]

So it is a 980Pro with PCI-E 4.0 disabled ?

 

[Mussels]

Really should have been a V2, with the drive being slightly better for most use cases and worse in rare cases, most users would not have card or notice... but sales and interest in the new and improved V2 would have been good marketing vs this shite

 

[bug]

heh hopefully people will sue them for false marketing

There's no case here, marketing is smart enough not to advertise the NAND itself

 

[RedBear]

So it is a 980Pro with PCI-E 4.0 disabled ?

They also come with different NANDs, 96L vs 136L.

On the new model itself, I can try to understand why they don't want to make it a V2, but they should have made a short press release to announce the change.

 

[BorisDG]

So it is a 980Pro with PCI-E 4.0 disabled ?

Mostly like defective Elpis chips.