Needed: Reviews on 2TB SSD's

[WarthogARJ]

There is a general lack of Reviews on 2TB SSD's (and I'm obviously only referring to variants where that exists).
I'd bet that there's at most a 30% coverage (or worse) of 2TB SSD's.
I'll welcome disagreements with any data to back them up.
But my survey shows that.

And I'd argue that with the continued drop in SSD prices, and the general acceptance of SSD's as both boot, internal data, abnd external drives, that the "sweet spot" for SSD's, is not at 1TB, but somewhere in between, 1TB and 2TB.

As in, if people had the information, and understood the potential benefits of a 2TB, then they'd buy one.
Or two.

People are pretty fixated on performance, even if they don't actually NEED it, and then at buying time, jump at what seems a bargain, without thinking if it's really the best deal for them.
If more people understood that a SSD will last longer, and run faster if it's given more overproportioned space, and not filled to capacity, then the idea of buying a 2TB and then giving it extra OP will be more common.
And you could then get the performance of a more expensive model, say Intel or Samsung, where you can only afford 500GB or maybe 1TB, and instead potentially pay LESS, and get a 2TB drive.

So I'm talking about increasing OP to 20-25% from the initial 9% that's usually provided, and then only filling what's left to say 80% at most (which is the max you should be filling one in any case).

So it means a nominal 1TB goes from an effective 740GB size (using the usual 9% OP, and 80% max fill) to 600GB (25% OP and max 80% full).
And using same idea, a nominal 2TB drive goes from 1440GB to 1150GB.

And if you REALLY needed 1TB useable space, a nominal 1TB doesn't give you that, anyways.

But if there are good reviews that consider this, you can find a 2TB drive that is both fast as well as cost effective compared to a high end drive.

If you've got a laptop, you're better off with an internal 2TB SSD than a 500GB, or even 1TB internal, and a 2TB external HDD because you need the extra capacity.

About the only issue is that Reviewers want to be first off the block with a new model release, and the 2TB sizes aren't always released at the same time as the 500 and 1TB ones.
But then they should do the 2TB when it comes out, and add it to the Review.

And if the excuse is they can only review what the Manufacturers give them, then I think if a Review site cannot get the 2TB, variant then it shows you're not being taken very seriously by the Manufacturer. So maybe need to up your game.
Or at worst, buy one, and then re-sell on eBay.
Costs you maybe $50 in the end?
Until at least you can convince Manufacturer's you're worthy of doing the bigger drives.

Feedback?

 

[EarthDog]

Outside of your overprovisioning requirements (some drives have built in overhead, no?) seemingly unusually high (who cuts off 25% MORE of their drive?!) i agree there aren't that many reviews out there.

Overprovisioning and only filling the drive XX% is almost doing the same thing twice (one is for spare nand, the other to leave space for background activities, trim/gc). Some drives (most?/all?) have provisions them already (some adjustable). Why make more? Side note, I havent overprovisioned an ssd of any kind for several years. There isnt a need, today, really. Keep some free space on it and let the factory provisioning do its thing.

SSD Over-provisioning (OP) - Kingston Technology

A percentage of space on an SSD is reserved for Over-Provisioning (OP) during firmware programming. Over-provisioning can improve the performance and increases the endurance of the SSD, alleviating NAND Flash wear over its useful life.

www.kingston.com

An ssd doesn't need any additional op to 'last longer'...or better said, last through its useful life. TBW values on most drives allow users to write 10s of GB /day for YEARS. And who's writing 10s of GB daily to their drives? It jsnt necessary to neuter capacity for additional op. 2010 ssds? Sure. Not today or the first few gens, however.

The SSD Endurance Experiment: They're all dead - The Tech Report

I never thought this whole tech journalism gig would turn me into a mass murderer. Yet here I am, with the blood of six SSDs on my hands, and that’s...

techreport.com

.

Think about this... a 1TB samsung 970 evo is 600tbw... that is 600,000GB. If you can, somehow, write 10GB / day, out of the box it should last 60,000 days. 100GB / day... 6,000 days. Something with a lower TBW on a cheap QLC drive? How about 240tbw on the intel 660p... 240,000 at 10GB / day is 2,400 days. Writes arent an issue on thess drives for an overwhelming majority of users, they wojt be with a lot higher tbw... and if they are, you aren't going cheap qlc in the first place.

I also think that 512-1tb is still the sweetspot for most m.2/nvme based ssds. For warm storage, its still pretty pricey.

Edit: note, the only difference (I can think of before coffee) between 1/2TB drives outside of capacity is speed and perhaps a denser IC. Controller is the same, form factor, etc.

 

[gen_angry]

I've got a WD 2TB SATA SSD that I just use as a single drive. Have had the thing filled to the brim a few times without any slowdown issues. Thing is a beast.

I think the reason it doesn't get posted about is because most still just get a 500-1TB NVMe for their OS/common programs and re-use an old spinner for bulk storage/games. Most current games (even COD, as ginormous as it is) are designed around spinners anwyays due to current generation of consoles so there is very little to no benefit from NVMe.

I think this will shift more when the next gen comes out since newer games will actually start leveraging the speed.

 

[TheLostSwede]

 

[Palladium]

My SX8200 Pro 2TB was a steal for $220 back in last Dec.

 

[WarthogARJ]

Outside of your overprovisioning requirements (some drives have built in overhead, no?) seemingly unusually high (who cuts off 25% MORE of their drive?!) i agree there aren't that many reviews out there.

Overprovisioning and only filling the drive XX% is almost doing the same thing twice (one is for spare nand, the other to leave space for background activities, trim/gc). Some drives (most?/all?) have provisions them already (some adjustable). Why make more? Side note, I havent overprovisioned an ssd of any kind for several years. There isnt a need, today, really. Keep some free space on it and let the factory provisioning do its thing.

SSD Over-provisioning (OP) - Kingston Technology

A percentage of space on an SSD is reserved for Over-Provisioning (OP) during firmware programming. Over-provisioning can improve the performance and increases the endurance of the SSD, alleviating NAND Flash wear over its useful life.

www.kingston.com

An ssd doesn't need any additional op to 'last longer'...or better said, last through its useful life. TBW values on most drives allow users to write 10s of GB /day for YEARS. And who's writing 10s of GB daily to their drives? It jsnt necessary to neuter capacity for additional op. 2010 ssds? Sure. Not today or the first few gens, however.

The SSD Endurance Experiment: They're all dead - The Tech Report

I never thought this whole tech journalism gig would turn me into a mass murderer. Yet here I am, with the blood of six SSDs on my hands, and that’s...

techreport.com

.

Think about this... a 1TB samsung 970 evo is 600tbw... that is 600,000GB. If you can, somehow, write 10GB / day, out of the box it should last 60,000 days. 100GB / day... 6,000 days. Something with a lower TBW on a cheap QLC drive? How about 240tbw on the intel 660p... 240,000 at 10GB / day is 2,400 days. Writes arent an issue on thess drives for an overwhelming majority of users, they wojt be with a lot higher tbw... and if they are, you aren't going cheap qlc in the first place.

I also think that 512-1tb is still the sweetspot for most m.2/nvme based ssds. For warm storage, its still pretty pricey.

Edit: note, the only difference (I can think of before coffee) between 1/2TB drives outside of capacity is speed and perhaps a denser IC. Controller is the same, form factor, etc.

Thanks for the feedback.

So let's not gewt distracted, there are two points here.
Maybe I should have posted two separate points.

The FIRST is that for various reasons I'd like o see ore revies of 2 TB drives.
For many reasons,

The SECOND point, is that for those people who actually NEED a high peformance SSD, there can be advantages from deliberately buying the next size up.
Quite possibly 90% of those who read this Forum don't need anything faster than a SATA drive.
Even a SATA HDD.

And maybe reading about bigger, faster and flashier compter parts is just geeking out.
Is maybe like reading about a fast car you'll never buy, or if you do, never actually drive it to max the speedo.
Fine.

I'm not saying this is appropriate for every user, but for some, who are heavy users, it could be very useful.
And my main point is that I think Reviews should be done with the aim of:
- Doing a thorough sample of the Market, both good and bad
- Doing it in a consisent and rigorous way, to allow others to use the data to make their own opinions
- Looking at the extremes of performance to some extent, if only to benchmark the drives they review

In terms of the Over Provisioning, I did an analysis of 125 SSD's where I had OP data, and got a pretty good correlation of about 9% OP.
In fact, the acual fit was:
Formated Size = B + M x Raw NAND
Where B = -7.3 GB
M = 0.91
With R^2 = 0.996
It was of drives from 2012-2020
Included 2D and 3D with MLC, TLC and QLC (but not many)
All consumer, no enterprise.
I removed three outliers (> 3 x s.dev)

So except for a few cases, most consumer drives don't have a lot of OP.

And it's very clear that overprovisioning up to 25% has strong effects on behaviour.
It reduces after that, but does'nt stop.

Sure, if you personally haven't needed to adjust your own OP, then it does'nt mean it's not a good idea for others.
In any event, if you've never done it, how do you know it won't help you?
Have you never had to wait for your drive?

In part it depends on the value you put on your time.
If you load a drive to saturation it gets REALLY slow: by close to several orders of magnitude in extreme cases.
And you have to wait an extra few hours.
And maybe more times than you'd really like, if given a choice.
If you could alleviate that by spending an extra $50 or even $100, to get the 2TB variant, well, some people would pay that.
I would.

Looking at the time one spends on a computer, and the difference in cost for a substantially better performance by buying the next bigger size for the cases where the SSD does affect your work, the cost is peanuts.
But a lot of people don't use a computer very heavily, so this is not for them.
Maybe more RGB's are a better investment....:-}

In terms of TBW, I don't agree that it's not something to consider for some users.
These values of TBW that people are happy to throw around, how does anyone know how accurate they are?
We don't.
Not even the NAND manufacturer's really know.
It's more of an exercise that an Actuary does for life insurance, to calculate risk premiums, than something that's based on materials engineering reliability statistics.

Consider the initial values for Program/Erase cycles for SLC of 100,000 cycles.
Hmmmmm.....a REAL thumb suck.
Then notice how it's rapidly reduced as the NAND got more complex: both in terms of architecture, as well as in requirements upon it, in terms of the voltage levels it needs to retain.
It's gone from 100,000 erase cycles down to some uncertain value between 300 - 3,000 for some types of NAND.
Hmmm.....that's a BIG change.

Well, not surprising, when you go from a 1D structure, with up to a 100nm scale down to a very complex 3D structure with < 20nm features.
You're getting very small.
The 1D designs were at their limit as they were.
Everyone knew that.
It was obvious that reliability and endurance were starting to become quite uncertain.

So going to 3D designs helped, but it made it much more difficult to really predict reliability and endurance.

The point of the TBW's from the Manufacturers is not to say your specific SSD will be fine.
Samsung, Micron, Intel don't care about one user.
If your's fails before waranty end, it's "oops, sorry, here's a new drive".

But if a large number fail, OK, it is an issue.
That's why I say it's more an exercise in actuarial statistics than actually making an accurate estimate in reliability.

And again, for the users who value the RGB led's, this is not likely to be an issue.
And nothing bad implied here, it's like having fuzzy dice hanging from your your car rear view mirror.

But there's not much data now on SSD endurance, at least not with the newer NAND.
Not with all the variabilities you can have, like distribution of defects in the raw silicon, how the coatings are deposited etc etc.
Especially not with users who run their NAND flash quite hot.
All these things are thin coatings of various ceramics, or metals, with lots of fancy junctions, oxides and geometries/dimensions.
It's pretty clear that you're going to get big variations in lifespan if you get any variation in the manufacturing process.
Or in how it's used.

Have you seen the work that shows the detrimental effects of running your NAND flash hot?
Anything over 40 C has a measureable effect (I'm not talking about the controller and thermal throttling, but the actual NAND flash).
It's an Arrhenius relationship.
The Manufacturer figures don't take this into account, so if you do write a lot, you might well have an issue.
Or perhaps you have a batch of NAND with higher defect concentration.

As for possibly increasing OP and well as not overfilling a drive, sure, maybe it is redundant to provide extra OP in the configuration, as well as to deliberately try not to fill your drive up too much.
But maybe not.

I'm not sure you can actually know that it's equivalent having say 250 GB as OP (let's say 25% of a nominal 1TB SSD), or else not using 200G after the drive has been formatted (asuming it has 10% as factory configured OP).

TechPowerUp is one of the few Review Sites that actually says how much the OP is, and Raw Nand size.
Most of the others don't, assuming they understand its importance.

The actual details of how controllers function are quite closely guarded, and differ from controller to controller.
I wouldn't be surprised if the controller does indeed treat either case differently.

Do you know?
Have you seen specific tests, or do you know exactly how all the controllers work?
I don't.
You cannot get a proper data sheet for a conroller.
Or the newer NAND.
It's all restricted by NDA's.

If nothing else, configuring your SSD to have the OP you'd like on Day One means it's likely to stay that way.
And you won't get lazy and keep it too full.


So sure, I'm not saying this is a big issue to most end users.
And it's maybe nothing you personally need consider.

But if someone is going to do a Review, might as well do it right.
TPU reviews are good.
I just wish he'd do more, instead of the real half a$$ed ones you see on some other sites.

And sorry, that was two pretty long posts.

In any case, I'm going to see what data there is about the effects of OP versus the % Capacity of an SSD.
Maybe it is indeed redundant.
But I've not seen anyone look at that.

I think what one would really like to do would be to configure your SSD controller as it suited you.
Adjust pSLC amount, the OP, and how it reacted to things like R/W ratio etc etc.

Temperature has a very obvious effect on NAND flash.
One can in principle anneal out wornout flash in fact: there's a patent for doing that using built in heating coils.
And all of the junction potentials are temperature dependent.
And it's a small enough distance internally that you can get some diffusion effects.

View attachment 162503

Maybe I'm slow, but what are you using this to illustrate?

 

[TheLostSwede]

Maybe I'm slow, but what are you using this to illustrate?

The synthetic performance of my 2TB NVMe SSD? I thought you wanted to see some benchmarks of 2TB SSD, as implied by your thread title, but apparently that wasn't what you were after...

You also seem to have completely missed the fact that going from TLC to 3D TLC actually upped the P/E cycles by quite a bit, maybe not to MLC standards, but enough to not having to worry about drives failing after a year or two. But hey, maybe you write 1TB a day to your SSDs?

Also, as you pointed out, no two SSDs, even from the same batch, are alike. So what's the point of doing long term testing to see how much data can be written to an SSD, as the one you get won't be the same anyhow?

Maybe you should talk to the SSD manufacturers, if you want to get custom pSLC, OP etc. as it's not something that can, or imho, should be done during a review, as it's irrelevant for 99.99% of people. Most users simply don't care, as long as it works.

There was another member here suggesting that we should all be using RAM cache software, or we're not using the caching available to the SSD and thus losing out on performance. I'm sorry to say, but you're pretty much in the same boat as him.
If you're not happy with the reviews here, as you're a Patreon, take it up in the Patreon forum.

Oh and with regards to product sampling, as an ex tech journalist of more than a decade in the business, I can tell you that quite often, you don't have a say in which spec you get sampled, as that's down to the company or PR company that samples you. Sure, you can make requests, but it doesn't mean you get what you ask for. 2TB+ SSDs are still quite pricey and as such, most companies don't send out that large samples, especially when a larger SKU has no performance advantages of a smaller SKU.

 

[EarthDog]

So except for a few cases, most consumer drives don't have a lot of OP.

If they needed more, they'd have more. Again, writes havent been an issue for ssds in years/multiple generations of ssds. You dont need to increase the factory op. It just serves to nerf your free space today.

If nothing else, configuring your SSD to have the OP you'd like on Day One means it's likely to stay that way.
And you won't get lazy and keep it too full.

If there was a need to change op, id agree! I'm not into losing free space for that reason. I see capacity in 'This PC' multiple times a day. Most can manage this without training wheels and losing precious available space.

In any case, I'm going to see what data there is about the effects of OP versus the % Capacity of an SSD.
Maybe it is indeed redundant.
But I've not seen anyone look at that.

If you like. Or you can see that even back in the day with factory op'd drives (with much worse write endurance in the first place) that writes aren't an issue as one of my links showed.

Most of the others don't, assuming they understand its importance

i believe (and judging from the thanks, others) you're placing an inordinate amount of importance on op. That's all I'm trying to say. Move past that point and know that 2tb over 1tb is really only different in capacity and speed (again # of nand or density may go up). The rest, for an overwhelming majority of users, even tpu readers, is minutia.

More 2TB drive reviews can't hurt...i agree... but for different reasons.

 

[WarthogARJ]

The synthetic performance of my 2TB NVMe SSD? I thought you wanted to see some benchmarks of 2TB SSD, as implied by your thread title, but apparently that wasn't what you were after...

You also seem to have completely missed the fact that going from TLC to 3D TLC actually upped the P/E cycles by quite a bit, maybe not to MLC standards, but enough to not having to worry about drives failing after a year or two. But hey, maybe you write 1TB a day to your SSDs?

Also, as you pointed out, no two SSDs, even from the same batch, are alike. So what's the point of doing long term testing to see how much data can be written to an SSD, as the one you get won't be the same anyhow?

Maybe you should talk to the SSD manufacturers, if you want to get custom pSLC, OP etc. as it's not something that can, or imho, should be done during a review, as it's irrelevant for 99.99% of people. Most users simply don't care, as long as it works.

There was another member here suggesting that we should all be using RAM cache software, or we're not using the caching available to the SSD and thus losing out on performance. I'm sorry to say, but you're pretty much in the same boat as him.
If you're not happy with the reviews here, as you're a Patreon, take it up in the Patreon forum.

Oh and with regards to product sampling, as an ex tech journalist of more than a decade in the business, I can tell you that quite often, you don't have a say in which spec you get sampled, as that's down to the company or PR company that samples you. Sure, you can make requests, but it doesn't mean you get what you ask for. 2TB+ SSDs are still quite pricey and as such, most companies don't send out that large samples, especially when a larger SKU has no performance advantages of a smaller SKU.

I see.
And i'm not so interested in an isolated single metric from a 2TB drive, but in a thorough Review.
Like the ones you usually do.

And I'm not expecting to get information about long term, or short term, performace from the tests that Reviewers do for Reviews, but I AM interested in the trends.
I'd like to see tests for steady state seqyential, random, and probably a mixed one too.
And also a test to see the effect of compressed vs uncompressed.

Proper steady state tests are what the Manufacturers's and SNIA recommend, as well as other things, but I don't see many Reviewers following the guidelines very consistently.

And no, I did not miss the change in P/E cycles that occurred on going to 3D architecture.
What I'm saying is that you can see very large changes in the quoted P/E cycles, as a function of architecture, and it would be, for lack of a better word, quite trusting, or even naive to accept them at face value.Especially when as I was saying it's very hard to get comprehensiove information of the newer NAND.
The data sheets are secured by NDA.
All you get are very brief summaries.
Same with the controllers from what I've seen.

And if you are saying you don't do the sort of test I was suggesting, because 99.99% of readers don't care, then i can accept that.
With some disappointment I guess.

But I think that is the wrong outlook: after all, if you look at AnanandTech when it was being run by Anand Shimpi, he did some very thorough and rigourous analysis.
And that's why it became such a popular and respected site.
Not just by gamers or button pushers who aren't interested in any in depth analyusis, but by people who are.
However now he's less involved, it's gone downhill.

And I think you do a good job, and I'd like you to do MORE of it as well.

As far as what the companies provide for tests, I think that if a Review site builds a good reputation for rigorous reviews, then it will be more likely to get the products it asks for.
And in some cases, I think you could buy a few (or maybe buy at cost), or perhaps SHARE with some other Reviewers.
Why not?
I understand that there's some feeling of competition, but if you and 2 or 3 other sites shared some items, or even the data, and co-operated, then you'd ALL look better, and attract viewers from sites not in your circle.

It's not a new idea. It's exactly what the airlines and rental car companies do.

As a reply in general, I replied too quickly when EarthDog was saying that having un-used space from extra OP, or from un-used space for formatted space is the same thing.
It is not the same thing, as I understand how many, perhaps all, SSD controllers work.

And to be clear, I am saying that let's say you have an SSD with a raw capacity of 1,000 GB.
Fresh out of the box.
And you prepare it for testing as follows:

Case A (OP):
Manufacturer supplies with 10% OP,
You format it, and get 900 GB of available formatted space, with 100 GB OP.
And let's say it has a fixed pSLC cache of 200 GB, provided by the manufacturer.

Case B (extra OP):
As above, but you increase TOTAL OP to 20%.
You now have 200 GB OP, and 800 GB formatted available space.
And the same factory supplied pSLC cache of 200 GB.

Now you do various benchmark tests on Case A and B.
And in both Cases, you do your tests on each drive after they've been filled with various amounts of sequential data: empty, 100GB, 300 GB, 500 GB, 700GB
That is filled BEFORE being tested.
To investigate the effects of using your drive when it gets fuller.

And I am very sure you will see quite large differences in the results as a function of:
- % OP
- % full before testing
As well as before the pSLC is full, and after

There are many reasons for it, some quite complex:
(1) OP is not the same as empty but formatted space in a number of ways:
- With Case B, having the extra OP means you've limited the AVAILABLE space to write to, as well as read from
OP is like being reserved.
It's not treated the same as formatted, even if not used
- The NAND designated as OP can be distributed amongst the dies as the controller logic determines.
Ideally in an optimum way.
As compared to empty but formatted space, which after it's started to be written to by random writes, is up for grabs

(2) When the drive becomes full, and when the spare area allocated by the OP is used up, then the drive needs to perform grarbage collection at the same time as writing
This is why high demand drives have more OP than basic consumer drives for people who don't run them very hard.

There are more, but let's leave it there for now.

I know the common refrain from most review sites is that most of their readers are gamers, and don't need high performamce SSD's.
Or that steady state tests, and writing big files, or lots of small files is unusual to most of their readers.

I'm not sure that's really the case, or that they are not interested in seeing the results.
And there are users who actually do use their computer for more than a source of flashing RGB's.

But there does seem to be a dearth of Review Sites that adress the interests of those who place heavier demands on computers.

 

[cucker tarlson]

There is a general lack of Reviews on 2TB SSD's (and I'm obviously only referring to variants where that exists).
I'd bet that there's at most a 30% coverage (or worse) of 2TB SSD's.
I'll welcome disagreements with any data to back them up.
But my survey shows that.

And I'd argue that with the continued drop in SSD prices, and the general acceptance of SSD's as both boot, internal data, abnd external drives, that the "sweet spot" for SSD's, is not at 1TB, but somewhere in between, 1TB and 2TB.

As in, if people had the information, and understood the potential benefits of a 2TB, then they'd buy one.
Or two.

People are pretty fixated on performance, even if they don't actually NEED it, and then at buying time, jump at what seems a bargain, without thinking if it's really the best deal for them.
If more people understood that a SSD will last longer, and run faster if it's given more overproportioned space, and not filled to capacity, then the idea of buying a 2TB and then giving it extra OP will be more common.
And you could then get the performance of a more expensive model, say Intel or Samsung, where you can only afford 500GB or maybe 1TB, and instead potentially pay LESS, and get a 2TB drive.

So I'm talking about increasing OP to 20-25% from the initial 9% that's usually provided, and then only filling what's left to say 80% at most (which is the max you should be filling one in any case).

So it means a nominal 1TB goes from an effective 740GB size (using the usual 9% OP, and 80% max fill) to 600GB (25% OP and max 80% full).
And using same idea, a nominal 2TB drive goes from 1440GB to 1150GB.

And if you REALLY needed 1TB useable space, a nominal 1TB doesn't give you that, anyways.

But if there are good reviews that consider this, you can find a 2TB drive that is both fast as well as cost effective compared to a high end drive.

If you've got a laptop, you're better off with an internal 2TB SSD than a 500GB, or even 1TB internal, and a 2TB external HDD because you need the extra capacity.

About the only issue is that Reviewers want to be first off the block with a new model release, and the 2TB sizes aren't always released at the same time as the 500 and 1TB ones.
But then they should do the 2TB when it comes out, and add it to the Review.

And if the excuse is they can only review what the Manufacturers give them, then I think if a Review site cannot get the 2TB, variant then it shows you're not being taken very seriously by the Manufacturer. So maybe need to up your game.
Or at worst, buy one, and then re-sell on eBay.
Costs you maybe $50 in the end?
Until at least you can convince Manufacturer's you're worthy of doing the bigger drives.

Feedback?

not every ssd has that 80% fill cap

 

[WarthogARJ]

not every ssd has that 80% fill cap

Hi Cucker,
I'm not sure what you mean exactly, but thanks for the feedback.

And i'm not saying there are any hard and fast actual values for where you stop gaining some benefit from extra OP, or from leaving empty space.
I just used 20% OP, and 80% full as a rough ballpark.

And indeed, you point reinforces my oint.
The ACTUAL values very likely differ from drive to drive.

And unless the Manufacturers release more information, and that probably means from all three: SSD "manufacturer", controller and NAND (they are rarely the same company), and in fact keeps a lot of it a secret, then we need to have thorough Reviews that allow us to determine these ourselves.

I find the general tone and level of a lot of review sites a bit demeaning, where they assume that their readers aren't that interested in the more intensive behaviour of SSD's. It's a bit like rading a car review of a potentially very fast car, where the reviewer says, "yeah it WILL go up to 300 km/h, but none of you readers actually ever NEED to go that fast, so don't bother about it".

And in the same analogy, they don't actually mention how many cylinders the car has, or any engine characteristics. Or about the gears.
Which is what happens when you get an SSD Review where they don't mention and then discuss details about the controller and NAND.

Or if the car (or SSD) has the marketing handle of being "fast", then they need to say what governs that.
Which for a car, is not so much engine size but handling.
And for an SSD, is often its cooling capacity, which afffects when it is thermally throttled.

And the "handling behaviour" of an SSD, is more how it deals with mixed loads, data sizes (4kB, 128kB etc), and QD's.

The "acceleration" and "top speed" ability of an SSD is maybe its behaviour under a sustained load.
A Horsepower vs RPM curve for an SSD is more like Throughput vs time or data written and or read.
Not an EXACT physics analogy, but more the same idea of what happens as you feed gas (or data) to the car (or SSD), and see what happens to the power (or data read/written speed).