Samsung is swapping parts in their 970 Evo Plus SSDs and sabotaging performance

mongeese

Posts: 611   +122
Staff member
A hot potato: Samsung has been caught changing the controller model used inside their 970 Evo Plus SSDs without updating the public spec sheet or publicizing the performance ramifications. It’s a sad thing to see, given the position of high esteem their drives have been held in for years.

The ‘new’ 970 Evo Plus was first tested by a Chinese YouTube channel (whose video on the topic does have English subtitles, if you want to watch). It’s unclear if the new model is exclusive to China or Asia, or if it’s being distributed globally.

Its only distinguishing feature is a controller branded Elpis, instead of Phoenix. The drives look otherwise close to identical, and they’re labeled the exact same -- 970 Evo Plus 1TB -- with only a small difference in part number denoting the drives as ‘technically’ different: MZVLB1T0HBLR becomes MZVL21T0HBLU on the sticker atop the drive.

As far as we know, the part number isn’t on the package, so you can’t distinguish them that way. The package is different though. The original’s box is oriented horizontally, while the new box sits vertically. And the label on the controller is blurred out in the picture of the drive on the front of the new box, which is something of a confession of guilt.

Photo credit: 潮玩客

In terms of performance, the new drive won’t be much worse for most users. But it has different strengths and weaknesses.

In a basic CrystalDiskMark run, the new drive was slightly faster across the board and even took a significant lead in some shallow write tests. In other synthetic benchmarks they traded blows, although the differences between them were minor. Their read speeds were all pretty much the same.

The real differences between the two crop up during sustained write loads. In a synthetic 200 GB copy test, the original drive sat at 1.8 GB/s up until 40 GB, then it dropped to 1.5 GB/s. The new drive began at 2.6 GB/s before dropping to 0.8 GB/s at 115 GB.

In a Windows File Explorer test, copying Blu-ray movies, both drives started out at about 2.5 GB/s. About 15% of the way through, the original drive dropped to 1.5 GB/s. The new drive worked its way down to 2.2 GB/s steadily before dropping suddenly to 0.8 GB/s at 80% of the way through. In the end, it took the two drives the same length of time to copy the 155 GB file.

Credit: 潮玩客

Judging by the benchmark results, Samsung has tripled the new 970’s SLC cache to make up for the slower controller. It’s not the worst solution, but it will become less effective as the drive fills up.

A second potential issue with the new drive is thermal output. During the benchmarks, which admittedly are rather extreme, the Elpis controller brushed up against 100° C. The Phoenix controller was more comfortable in the low eighties.

Although the new 970 Evo SSD isn’t bad, it isn’t okay for Samsung to leverage the reviews and reputation of the original 970 to sell the new one. They’re fundamentally different products that suit different use cases, and that should have been denoted by a change of name.

Believed to be a consequence of chip shortages, Western Digital was also caught recently changing the NAND flash in one of its budget SSDs with a lower-grade version that underperforms compared to the original -- an unfortunate misleading behavior that's slowly turning into a trend.

Permalink to story.

 

BSim500

Posts: 906   +2,131
"And the label on the controller is blurred out in the picture of the drive on the front of the new box, which is something of a confession of guilt
the original drive sat at 1.8 GB/s up until 40 GB, then it dropped to 1.5 GB/s. The new drive began at 2.6 GB/s before dropping to 0.8 GB/s at 115 GB.
And this is why I LOL at the current e-peen p*ssing contest of those arguing how worthless motherboards are that lack PCI-E 39.0's x4 theoretical 256 Exabyte per femto-second in CrystalDiskMark v17 and are peasants "stuck" on PCI-E 3/4 as if it actually means something in the real world with race to the bottom cr*p like this from even the Premium Brand market leader going on...
 

fps4ever

Posts: 984   +1,477
And this is why I LOL at the current e-peen p*ssing contest of those arguing how worthless motherboards are that lack PCI-E 39.0's x4 theoretical 256 Exabyte per femto-second in CrystalDiskMark v17 and are peasants "stuck" on PCI-E 3/4 as if it actually means something in the real world with race to the bottom cr*p like this from even the Premium Brand market leader going on...

I have a cheap Dell 256GB NvMe drive in my pc that does 1700/1100 down/up that seems really snappy compared to any Sata SSD's I've owned. For the average home/gamer even the cheap ones work fine. I have to say I do not do large file transfers of 100GB+ very often so your mileage may vary....
 

Sausagemeat

Posts: 1,597   +1,422
I mean they changed a component on the board and the drive still performs more or less the same. Maybe they had supply issues? Are we really expecting them to release a whole new SKU because one component is slightly off? I bet most people who purchased the drive don’t even know how to find out which one they have or even care.

We are demanding impossible standards here. Samsung can’t even really go public an announce it was changed because the tech community will just hurl abuse at them and it’s not worth the fallout.
 

Reehahs

Posts: 1,297   +970
I mean they changed a component on the board and the drive still performs more or less the same. Maybe they had supply issues? Are we really expecting them to release a whole new SKU because one component is slightly off? I bet most people who purchased the drive don’t even know how to find out which one they have or even care.

We are demanding impossible standards here. Samsung can’t even really go public an announce it was changed because the tech community will just hurl abuse at them and it’s not worth the fallout.

As the author pointed out that this change on the sly is becoming a trend. In the case of WD, it was worse for the consumer. If they change things, they ought to make it clear what they changed and how it will have an impact on performance (if it is a key parameter for the sale/marketing).

Couple of things that review sites can do is to remove the drives from their recommended list citing this practise and update the figures where possible. All of the new products that they receive from manufacturers doing this should have a clear disclaimer at the top that the version may change without notice to have worse performance without disclosure so buy with caution.
 

Theinsanegamer

Posts: 3,523   +5,933
I mean they changed a component on the board and the drive still performs more or less the same. Maybe they had supply issues? Are we really expecting them to release a whole new SKU because one component is slightly off? I bet most people who purchased the drive don’t even know how to find out which one they have or even care.

We are demanding impossible standards here. Samsung can’t even really go public an announce it was changed because the tech community will just hurl abuse at them and it’s not worth the fallout.
So losing 12% of your max speed, nearly 50% of sustained performance for large file transfers, and running significantly hotter is "more or less the same" to you?

I'm pretty sure if your employer took 12% of your pay and said you pay was "more or less the same" you'd be pissed. Same if a spec change made your new GPU 12% slower, or your new car 12% less efficient then the review models you based your purchase on.
 

Sausagemeat

Posts: 1,597   +1,422
As the author pointed out that this change on the sly is becoming a trend. In the case of WD, it was worse for the consumer. If they change things, they ought to make it clear what they changed and how it will have an impact on performance (if it is a key parameter for the sale/marketing).

Couple of things that review sites can do is to remove the drives from their recommended list citing this practise and update the figures where possible. All of the new products that they receive from manufacturers doing this should have a clear disclaimer at the top that the version may change without notice to have worse performance without disclosure so buy with caution.
This is something that has been happening for a long time if you ask me. When you buy something like a laptop, or a console or a digital camera or even a phone or a graphics card, the components inside it can change. Some models have different screens to others for example. Even on graphics cards we get some model of cards with different memory or caps etc. I had 2 R9 280X in crossfire a few years back and I remember taking the coolers off to clean them and they had different labelled components on their PCBs and ever so slightly different steel brackets.

I am not someone who needs fast PCIe drives so I don’t care. But I am curious as to who would care about this tiny change? Which users does this hurt? Please do correct me if I’m wrong but there isn’t anyone or any actual sensible application I can think of.
 

Sausagemeat

Posts: 1,597   +1,422
So losing 12% of your max speed, nearly 50% of sustained performance for large file transfers, and running significantly hotter is "more or less the same" to you?

I'm pretty sure if your employer took 12% of your pay and said you pay was "more or less the same" you'd be pissed. Same if a spec change made your new GPU 12% slower, or your new car 12% less efficient then the review models you based your purchase on.
If you read the entire article it shows that whilst some things were slower, others were faster.

And I’m not necessarily defending Samsung, I’m just saying this is a small change that seemingly people needed to test to find out. Why are we holding SSDs to higher standards than practically all other electronics, which have run to run variations on the production lines.
 

seeprime

Posts: 681   +893
I mean they changed a component on the board and the drive still performs more or less the same. Maybe they had supply issues? Are we really expecting them to release a whole new SKU because one component is slightly off? I bet most people who purchased the drive don’t even know how to find out which one they have or even care.

We are demanding impossible standards here. Samsung can’t even really go public an announce it was changed because the tech community will just hurl abuse at them and it’s not worth the fallout.
Yes. Samsung should have renamed the revised part V.2, or something similar. Customers should not tolerate quiet changes that change how an item performs. We need to keep the big boys as honest as possible, for our own sake.
 

Sausagemeat

Posts: 1,597   +1,422
Yes. Samsung should have renamed the revised part V.2, or something similar. Customers should not tolerate quiet changes that change how an item performs. We need to keep the big boys as honest as possible, for our own sake.
I would appreciate transparency. But for fairness to the manufacturer, we can’t always assume they are making the changes to save money and increase profits. If manufacturers are actively deceiving reviewers and customers with faster early runs then that needs to be called out. But at the same time if a manufacturer runs out of a certain component and has to substitute with something else then it’s not really fair to tar them with the same brush.
 

brucek

Posts: 1,205   +1,752
We are demanding impossible standards here.
Nothing impossible or even unreasonable about it. If Samsung can change out one of the two major components that make up the SSD, which involved work at the supply chain, manufacturing, validation, and possible firmware/driver levels, it can also handle the much more minor changes of altering the packaging to include a somewhat different SKU and send out a press release.
 

Sausagemeat

Posts: 1,597   +1,422
Nothing impossible or even unreasonable about it. If Samsung can change out one of the two major components that make up the SSD, which involved work at the supply chain, manufacturing, validation, and possible firmware/driver levels, it can also handle the much more minor changes of altering the packaging to include a somewhat different SKU and send out a press release.
Well I do think they should tell people at least. But our standards don’t actually name the specific components. Technically it looks like this SSD still meets its own performance standards as advertised, it doesn’t look like you would have any legal grounds for mis-selling. It just might be difficult to guarantee specific components for production runs.
 

brucek

Posts: 1,205   +1,752
I guess we won't know if and until there's a lawsuit, and it may vary by region, but if there was a fraud case and I'm on the jury, I'm going with yes this is fraud.

The controller is a substantial and material part of an SSD. The replacement part does not have the same spec sheet as the original part, nor the same performance, nor the same thermal characteristics (which might be a very big deal if bought for a laptop btw.) The unit being sold is no longer a Samsung 970 Evo Plus as the market has come to know it. (I would not feel the same way if the change was truly immaterial, say, the brand of solder on the PCB.)

One question I might ask if I was the investigator or lawyer, what communication or arrangements did Samsung make with their commercial partners for this part? For example, if Apple is a customer, did they just start sending new parts to Apple, or did their own conduct indicate they knew it was a material change and therefore went through a communication, negotiation, and agreement process with Apple before performing the switcheroo? If so, why would the regular consumer not deserve the same protection?

I might also look at what happens in industries where there's a broader understanding among laypeople. If BMW changed out the engine in its car, would they be allowed to do that silently or would the average person feel they had a right to know they were getting a different engine before making the purchase?

All that said, I have no ill will at all for Samsung perhaps needing to make a change, whether for pandemic, general supply, belief its a better product, or even just profit motivation. That's all fine - all they need to do is be straight they are offering a different product.
 

kiwigraeme

Posts: 1,195   +871
Thanks - least this article is more balanced - on a bigger tech site - final conclusion "drastic reduction in performance" - in real life this will only affect a few % of end users - I would posit those users wouldn't be buying a 970pro but a 980 pro ( editing 4k Video , scientific data analysis ). So most users would gain or notice no difference - except when needed to backup the drive.
Yes samsung should have given it slightly different name. But as stated by sausagement this happens all the time by every large manufacturer - though TBF this is part - not a complex GPU .
reminds me off people ringing up sellers and asking batch number , or was it made in Thailand or Japan - I think Nintendo nds had a lottery on screen type. Oleds made in china or poland etc etc .
Personally I always had in the back of mine to buy a sugar tester - so in the supermarket - scan zucchini ( courgettes) , pumpkins to know you are not buying a dud
 
Last edited:

arrowflash

Posts: 521   +599
In this case I'd say that it's clearly not the same product anymore. The thermal characteristics alone are enough, plus even if average performance is the same and might be better in some cases, the new version obviously performs less consistently.

Yes manufacturers often change some product's characteristics during their run without changing SKUs or announcing them. But like @brucek points out, what Samsung did is not the same thing as simply changing the brand of solder, the manufacturer of the PCB or resistors, or some slight revision in PCB layout. It's changing a key core component and key core characteristics (I.e. amount of on-board cache), and while there may be some use cases where it outperforms the old version, I see it as a clearly inferior product. Samsung should have changed the SKU and released it as a 970 Evo Plus v2 or 975 Evo Plus.

I'm a satisfied owner of a 970 Evo Plus, however if it was this new version I'd probably have purchased a different product.
 

RedBear

Posts: 55   +46
Samsung should have changed the name, or at least made a short press release about the new model, but saying that they're "sabotaging performance" when the actual results are a mixed bag and it's generally better in real world usage for most users (people don't copy 200GB every day) isn't exactly indicative of good journalism either. It's called clickbaiting.
 
As an owner of the original 970 evo plus, I think I have an idea of why they changed the controller and the slc cache..Ive been in contact with samsung trying to get them to do a firmware update on these 970's for ages..The issue is that the slc cache was getting used up and was Not refreshing,thus users were experiencing a severe drop in write speeds to around 800mbs on these over time..it was as if the drive was writing at tlc speeds with little benefit of the superfast slc cache writes..the only way yo get the speed back was to do a secure erase which was an uber pain cuz you loose all data and have to redo everything ...even worse is that yes you can get the cache working again but its gonna go south in about a month of use..The 980 pro had the same issue..Owners bitched about it and samsung fixed it with an update..Not so for 970 plus owners..So I wonder if this redesign is in part motivated by those issues and an attempt to address them on a hardware level instead of a simple firmware update?..changing the controller makes sense and trippling the amount fo slc cache as well... maybe these New drives are more consistant in long term performance than the previous ones, becuz I can tell you, Im NOt pleased with that original controller and the way the slc cache works..its pretty bad and users get a sub par experience if they use the drive over time...some fix was and IS needed.
 

paul1122

Posts: 244   +261
It never ends. I just tore apart my replacement stick vacuum from Shark after reading this article, and found they had used a black screw on the bottom plate, and the replacement had a silver screw. Bait and switch 100%.
 

James Ryan

Posts: 22   +38
They got done for similar crap a few years back with their TV's (LG as well) They were rigging them to die not long after the warranty period expired..
 

whateversa

Posts: 85   +110
Won't the new version have a shorter life span if a component reaches 100 degrees celcius? I certainly do not want any electronics with a shorter life span.
 

whateversa

Posts: 85   +110
They got done for similar crap a few years back with their TV's (LG as well) They were rigging them to die not long after the warranty period expired..

My LG TV is currently 9.5 years old.
My LG surround sound system is around 12 years old.

My media PC's Asus motherboard, Corsair SSD and Seagate drive is also 9 years old (used to be my primary PC)

My Samsung microwave is about 14 years old (never repaired).

My LG dishwasher was 9 years old (and still working) when I replaced it - racks started rusting.

I really expect any electronic device to last as long, or I will ditch the brand. If what you say is true, I will never buy LG or Samsung again.
 

whateversa

Posts: 85   +110
As an owner of the original 970 evo plus, I think I have an idea of why they changed the controller and the slc cache..Ive been in contact with samsung trying to get them to do a firmware update on these 970's for ages..The issue is that the slc cache was getting used up and was Not refreshing,thus users were experiencing a severe drop in write speeds to around 800mbs on these over time..it was as if the drive was writing at tlc speeds with little benefit of the superfast slc cache writes..the only way yo get the speed back was to do a secure erase which was an uber pain cuz you loose all data and have to redo everything ...even worse is that yes you can get the cache working again but its gonna go south in about a month of use..The 980 pro had the same issue..Owners bitched about it and samsung fixed it with an update..Not so for 970 plus owners..So I wonder if this redesign is in part motivated by those issues and an attempt to address them on a hardware level instead of a simple firmware update?..changing the controller makes sense and trippling the amount fo slc cache as well... maybe these New drives are more consistant in long term performance than the previous ones, becuz I can tell you, Im NOt pleased with that original controller and the way the slc cache works..its pretty bad and users get a sub par experience if they use the drive over time...some fix was and IS needed.
I hope TechSpot will read your comment, get feedback from Samsung and update the article/make a new article.
 

whateversa

Posts: 85   +110
So losing 12% of your max speed, nearly 50% of sustained performance for large file transfers, and running significantly hotter is "more or less the same" to you?

I'm pretty sure if your employer took 12% of your pay and said you pay was "more or less the same" you'd be pissed. Same if a spec change made your new GPU 12% slower, or your new car 12% less efficient then the review models you based your purchase on.
No, they said total time to transfer the large file ended up being identical. The high temps on the other hand does bother me - it tells me to expect shorter lifespan.