Samsung 980 PRO NVMe SSD Uses TLC NAND Flash with Half the Endurance of 970 PRO: Product Page

[R0H1T]

I've seen that several times for those who work with HD+ video editing 9-5 day-in, day-out after a few years.


"Best performing" yet your charts say "power efficiency" not performance. So that SK P31 drive you picked out certainly looks to be a good laptop drive, but given that it tops out at just 1TB capacity with TLC hardly puts it in the same market as the heavy-duty Samsung PRO's which max at 4TB (MLC) / 8TB (TLC). People don't buy the huge MLC drives just as ordinary boot drives, they are for pros with higher data usage.


They're hardly unrelated either. That's why warranties and TBW have been falling from 10yr (MLC PRO) to 5yr (TLC EVO) to just 3yr (QLC QVO). There is an underlying reason for that... (Hint: Can you imagine if 2017-era Gold rated PSU's were sold with 10yr warranties whilst 2020-era Platinum rated ones of the same brand were sold with just 3yr warranties. Would certainly raise a few eyebrows, eh?...)


I would have thought that's obvious - the fewer P/E cycles, the faster a drive will burn through it's flash-"writeability" given the same load. On top of that, 2-bit MLC -> 3-bit TLC -> 4-bit QLC also reduces the voltage state overhead vs voltage drift / cell leakage making the latter drives highly unsuitable for rarely powered up external backup drives (since they won't be able to silent background refresh when unpowered and the time between power ups could be long enough to cause problems with a QLC that has barely 7% overhead per voltage state that won't be there on MLC that has 33% or 4.5x the overhead.

Yes & I picked that drive specifically because it is by far the most efficient one I've seen to date, less heat could mean better "endurance" & certainly leads to a more consistent performance. It is a PCIe 3.0 drive so it's nowhere near what the top drives in this segment can do & you realize they cost a ton more, don't you? The 980 pro isn't even out yet, I'm willing to bet that it'll near about top the charts across the board perhaps with the exception of class leading efficiency. Now if you've been following the PC market you'll also know that chasing the last 1% or indeed last 0.1% of performance costs way more than the first 90~99% even now.

I'll ask again, can you link me to studies which map out let's say a 1000 possible use cases for these drives & covering the NAND "endurance" part? You know why that's not happening ~ the answer's in my last reply!

 

[BSim500]

Yes & I picked that drive specifically because it is by far the most efficient one I've seen to date, less heat could mean better "endurance"

Unless they run outrageously hot, thermals don't make anywhere near the difference as cell exhaustion or having to write to cells in multiple "passes" of SLC cache -> TLC/QLC transfer vs the single pass of SLC/MLC.

I'll ask again, can you link me to studies which map out let's say a 1000 possible use cases for these drives & covering the NAND "endurance" part?

Exactly 1,000, and not 998 or 65,536 use cases? As explained previously, use cases for the high-endurance PRO's are pretty much anyone who needs to write a lot (video editing, multi-track lossless audio, certain databases, 3D rendering, etc). Surely you can figure it out that the guy who spends all day editing 4K video vs granny who boots up Windows to check her e-mail then shuts down after 5 minutes are obviously going to have different needs?...

 

[R0H1T]

Unless they run outrageously hot, thermals don't make anywhere near the difference as cell exhaustion or having to write to cells in multiple "passes" of SLC cache -> TLC/QLC transfer vs the single pass of SLC/MLC.

Any verifiable evidence for that?

As explained previously, use cases for the high-endurance PRO's are pretty much anyone who needs to write a lot (video editing, multi-track lossless audio, certain databases, 3D rendering, etc). Surely you can figure it out that the guy who spends all day editing 4K video vs granny who boots up Windows to check her e-mail then shuts down after 5 minutes are obviously going to have different needs?

Right ~ let's day I'm running multiple VM's on that drive & have it properly cooled. The drives doing fantastic, yet at some point involving a read heavy task it locks up & just goes kaput. It did say 1.5 PB in writes, can you then definitively say that the lower "endurance" rating killed it & nothing else?

The point, my point is that you can't have it all ~ a drive having chart leading performance, which "lasts" 10 years, is cool to boot & covers itself in glory for everything you throw at it. Even enterprise drives compromise on something. Rated NAND writes are the last thing anyone should worry about on most of these consumer drives, unless you have a specific use case where heavy writes are to be expected. In which case there's better, albeit more expensive alternatives.

 

[ssdpro]

I am only interested in performance in a "Pro" product meant for a desktop. The problem with TLC is the performance plummets under any load to nearly SATA levels. It tries to only "write" to the top layer and gradually move so you get bursts that are high and averages that are abysmal. TLC has its place and works for the novice user that reads that "max spec" but doesn't know the real spec is about 30% of that. See charts:

 

[BSim500]

Any verifiable evidence for that? Right ~ let's day I'm running multiple VM's on that drive & have it properly cooled. The drives doing fantastic, yet at some point involving a read heavy task it locks up & just goes kaput. It did say 1.5 PB in writes, can you then definitively say that the lower "endurance" rating killed it & nothing else?

If it's properly cooled then it wouldn't randomly lockup. And page 7 of the Anandtech article you linked to showed the SK31's drive temps hit 83c, ie, they aren't any lower than any previous SSD tested using same methodology, so I genuinely have zero idea what point you're trying to make in falsely believing a lower power consumption for laptops = lower operating temps or better reliability. They aren't remotely the same thing. Nor does a lower temp TLC / QLC = some "magic beans" compensations for inherently inferior flash endurance...

 

[R0H1T]

If it's properly cooled then it wouldn't randomly lockup.

And you're saying this based on?

Nor does a lower temp= some "magic beans" compensations for inherently inferior flash endurance...

Lower temp (i.e. better efficiency) does mean consistently good performance, that's the key point why I linked to the Sk Hynix charts ~

 

[Searing]

Ugh the MLC vs TLC wars again. I'll just say that I find MLC to be completely useless and offer no advantages at all (remember if it costs twice as much, you can buy twice as much TLC storage and solve all your issues). TLC is very mature. QLC on the other hand...

 

[Makaveli]

I am only interested in performance in a "Pro" product meant for a desktop. The problem with TLC is the performance plummets under any load to nearly SATA levels. It tries to only "write" to the top layer and gradually move so you get bursts that are high and averages that are abysmal. TLC has its place and works for the novice user that reads that "max spec" but doesn't know the real spec is about 30% of that. See charts:
View attachment 167330

Nice!

My drive has a SLC write cache of about 333GB once that is used up it will hit about 600 mb/s for sustained writes.

However I don't think I've ever written 333GB of data to the drive in one shot is usually much lower so I've never see the performance outside the cache. Other TLC drive have smaller cache and the hit on those will show up sooner.

 

[BSim500]

If it's properly cooled then it wouldn't randomly lockup.

And you're saying this based on?

Observable reality. If a drive is properly cooled, it won't throttle. Period. There's nothing special whatsoever about the SK31 that makes it immune to throttling if you stuck it underneath a horizontal Mini-ITX board and if you're seriously claiming that a TLC drive with a rated 1,000 P/E cycles running at 83c will last twice as long as an MLC drive with a rated 10,000 P/E drive at say 87c like for like, to allow you to pretend there's no difference between TLC / QLC vs SLC / MLC for P/E cycle endurance, simply because it draws literally milliwatts less power, then it's up to you to prove your claim, not everyone else to disprove it. In the mean time, here's a review of an 970 PRO (MLC) drive peaking at 55c load temps, rendering your "oh yeah, well my 83c SK31 TLC drive will last longer than everyone else's MLC drives because it draws less power and runs cooler" point rather moot, not to mention the blatantly obvious that someone paying for an expensive drive will have no trouble sourcing a half-decent $10 M2 cooler even if your claim were true. I honestly didn't think it needed mentioning that light-duty casual SSD users may run out of space before an MLC drive dies of cell exhaustion is completely irrelevant if they aren't the target market for whom the high-endurance MLC drives are aimed at in the first place, but here we are...

 

[R-T-B]

QLC?

I never mentioned QLC.

TLC on professional focused products is dubious.

Honestly, it's not if the node size is large enough. But then you lose the space savings. So yeah, I guess it is A LITTLE dubious since you have to question what their motive is in the first place.

My main point here is though, "professional" does not need mean "absurdly long lifetime" like it has in the past. But Samsung can and should knock the price down accordingly. I don't have confidence that will happen however.

 

[R0H1T]

Observable reality.

More like (your) warped reality, FYI a drive can die for any number of reasons & my post had no mention of throttling. Also I've never claimed anything about (NAND) endurance & temps ~

less heat could mean better "endurance" & certainly leads to a more consistent performance

Unlike you who's just posting somewhat contradicting statements ~

Unless they run outrageously hot, thermals don't make anywhere near the difference as cell exhaustion or having to write to cells in multiple "passes" of SLC cache -> TLC/QLC transfer vs the single pass of SLC/MLC.

And then this

There's nothing special whatsoever about the SK31 that makes it immune to throttling if you stuck it underneath a horizontal Mini-ITX board and if you're seriously claiming that a TLC drive with a rated 1,000 P/E cycles running at 83c will last twice as long as an MLC drive with a rated 10,000 P/E drive at say 87c like for like

The observation is wrt efficiency, heat & (consistent) performance. Now unless you're living in some part of the world where the natural law doesn't apply, your points literally make zero sense!

claiming that a TLC drive with a rated 1,000 P/E cycles running at 83c will last twice as long as an MLC drive with a rated 10,000 P/E

Let's see, the one claiming SK Hynix 4d NAND has a P/E cycle rating of 1000 & 970 pro having 10000 P/E cycles, yup that's me! Wonder where you're pulling this from

In the mean time, here's a review of an 970 PRO (MLC) drive peaking at 55c load temps

Wonderful a review from AT, another from Guru3d. Let me have what you're sipping right now.

I honestly didn't think it needed mentioning that light-duty casual SSD users may run out of space before an MLC drive dies of cell exhaustion is completely irrelevant if they aren't the target market for whom the high-endurance MLC drives are aimed at in the first place, but here we are...

There is no such things as high endurance MLC drive in the consumer space! The drives having MLC are first & foremost geared towards better & consistent performance, it just so happens that MLC have better endurance. You want better "endurance" drives get an enterprise one ~
The Samsung 983 ZET (Z-NAND) SSD Review: How Fast Can Flash Memory Get?

but here we are...

Yup here we are, making fantastical claims about how 3d/4d NAND, 96L TLC, 128L TLC, xxL QLC or MLC behave all in a manner which is consistent with your hypothesis devoid of any real world studies to back them up

 

[BSim500]

<Snipped a bunch of mindless troll-bait.>

I genuinely do not understand why you're so hyper-triggered, but you really are sounding out of mind at this stage:-

- "I've never claimed anything about (NAND) endurance & temps" - You literally said in post #51 "less heat could mean better endurance" in direct quoted response to what everyone was very obviously comparing (NAND lifespan based on P/E cycles between MLC vs TLC vs QLC). Again, when people do compare drives then it's pretty damn obvious we mean "like for like", not "when doing MLC vs TLC vs QLC P/E based NAND endurance comparisons, let's start counting random controller failures that have zero to do with NAND" or "If I hold a hair dryer up against an MLC drive whilst pointing a fan at the TLC one, that could complicate NAND lifespan comparisons". No sh*t, Sherlock...

- "Wonderful a review from AT" - Dude, that was your own link from your own post #45...

- "Higher endurance consumer MLC doesn't exist, it just so happens that MLC have better endurance. Consumer SSD's are not about endurance at all but performance". Perhaps you better e-mail Samsung and tell them they've gotten their marketing materials wrong considering that "Pro performance and Endurance" has been literally right there at the top of their SSD PRO range pages going back 5 years with the words "designed to handle heavy workloads on workstations and high-end computers with IT heavy users in mind" literally underneath pushing endurance as a primary sales pitch...

Sorry dude but you'll have to find someone else to troll with these mindless "arguing for the sake of arguing" word games. It's pretty obvious who buys MLC drives and your own claims have already been debunked by Samsung's own website in nice big size 64 font. Two people here other than myself have also tried to explain it to you, and if you still can't figure it out, then far from "curing the forum of misinformation", you've ironically ended up the bigger purveyor of it...

 

[lexluthermiester]

I never mentioned QLC.

Didn't say you did. I included it for reference and comparison to help with a certain perspective.

Honestly, it's not if the node size is large enough.

That's actually a very good point. The larger the process, the more durable a NAND gate becomes. If TLC was made on the 20nm or 16nm process for example, durability would be far greater than the current 7nm(or is it still 10?) because there would be more material to "wear" through before failure.

"professional" does not need mean "absurdly long lifetime" like it has in the past.

Have to strongly disagree with that. When I see a physical product focused on the "professional" market sector I expect and demand greater quality & durability for the extra price being paid. Otherwise it becomes a meaningless word that loses it's distinction.

Let's see, the one claiming SK Hynix 4d NAND has a P/E cycle rating of 1000 & 970 pro having 10000 P/E cycles, yup that's me! Wonder where you're pulling this from

Please review the following in the section "Write Endurance";

Flash memory - Wikipedia

en.wikipedia.org

Those are general numbers but are accepted as effective standards in the industry. Endurance numbers will vary somewhat from one manufacturer to another.

 

[R-T-B]

Otherwise it becomes a meaningless word that loses it's distinction.

It doesn't lose it's distinction if it's stronger in endurance than the other products in the lineup, but it certainly makes it less enticing vs the previous gen.

 

[lexluthermiester]

It doesn't lose it's distinction if it's stronger in endurance than the other products in the lineup, but it certainly makes it less enticing vs the previous gen.

It does if it's much less durable than previous iterations of the product lineup, which in the case of the above mentioned product in the article, it is. Even if just a "ballpark" estimate, anywhere near half durability is unacceptable if Samsung wants to continue calling it a "pro" product and charge similar prices as before. Samsung's new lineup simply doesn't meet the mark of what can be considered a professional product with this very much lesser standard and is no longer worth the price they're asking.

 

[R0H1T]

BSim500 Considering you quoted me first, perhaps you should ask yourself that question? As for the endurance/temps reply from me, are you really pretending to be daft or that you really are? I said could ~ which was to your own response in my previous post where I said that there's not enough data to conclude how temps (can) affect all types of NAND.

For instance we have no idea how operating drives (3D NAND) at high temps for an extended period of time affects the NAND "lifespan" or indeed what happens when you're using drives for read/write heavy tasks with say good airflow.

The scenarios I had in mind were as ~ persistent writes, likely with mid/high temps vs say a large number of reads/writes & then the drive idling after that. Tell me what does your crystal ball say about Endurance for either of them?

^Dude you brought up the point about 83C temps for Sk Hynix from the AT review, why are using numbers from Guru3d to compare it to 970 pro! Next you'll tell me pick any random review & compare the numbers from it to any other, for a truly apples to apples comparison you must have a controlled environment wherein the reviewer has used the same system, workload(s) & ideally same ambient temp to test the drives! And let's see the BS you're posting from AT ~

The power management feature set of the SK hynix Gold P31 is fairly typical. The warning and critical temperature thresholds are only a degree apart, but realistically, this SSD isn't getting anywhere near those temperatures without a lot of outside assistance. The power state transition times claimed by the P31 are pretty quick.

Note that the above tables reflect only the information provided by the drive to the OS. The power and latency numbers are often very conservative estimates, but they are what the OS uses to determine which idle states to use and how long to wait before dropping to a deeper idle state.

So do me a favor, next time you quote me at least read the drivel you've posted first heck I'd be glad if you didn't quote me

Also when you respond to me with statements bring up real life numbers, not something you conjure up in your mind. I'm an authority on everything SSD is not an argument, at least not when you're pretending to ask BS questions (to points which I never made) & answering them yourself!

As for your last point ~ keep the trolling part to a corner where you can deal with it as it seems to be a hobby of yours. I only engaged with you hoping you'd have serious replies, but then looking at your fantastical claims with exactly Zero Facts backing them up I have to say this is not worth my time so Ciao, this is my last reply to you

 

[lexluthermiester]

The scenarios I had in mind were as ~ persistent writes, likely with mid/high temps vs say a large number of reads/writes & then the drive idling after that. Tell me what does your crystal ball say about Endurance for either of them?

No crystal ball is needed. Physics will tell you that high temps + heavy work load = degraded performance & diminished device life span.

 

[R0H1T]

Physics will tell you that high temps + heavy work load = degraded performance

Yeah that's exactly what I said.

less heat could mean better "endurance" & certainly leads to a more consistent performance.

Unlike the poster I was replying to.

& diminished device life span.

True for most electronics or ICs though specifically this part is unclear especially wrt rated endurance or lifetime writes of drives & the various types of NAND we have on the market today. Like I said there's literally a trillion other use cases or possible combinations involving workload, application, system wherein to cover all of them for a manufacturer is nigh impossible. This is why lower endurance numbers are a reality ~ they are however as a general rule of thumb highly conservative for most drives on the market today!

 

[lexluthermiester]

Yeah that's exactly what I said.

Unlike the poster I was replying to.

Oh, I am sorry. Misunderstood you. It seemed like you were arguing against that point.

 

[olstyle]

Considering NAND exhausting p/e cycles is the only thing that can fail on a SSD, endurance is at least 90% of reliability

That's just plain wrong. All bigger drive series failes that occured were firmware issues and the sometimes happening "suddenly does not boot at all" error is a controller issue.
I have never come around anyone actually exceeding write cycles = running out of rearranged sectors.

 

[Octopuss]

So the advertised SSD speeds are basically speeds of the cache (be it that few GB SLC buffer or whatever), right?

I don't like this, not because of the durability, but because when I buy something that's supposed to transfer stuff at X MB/s, I expect X, not X/10 if I copy large enough file.

 

[olstyle]

Once the Cache has a size which is bigger then 10 times your normal file size it's kind of academic.
That's a bit like saying my VRAM is not fast because once it's exceeded the system RAM is used.

 

[RoyTyrell]

I’m sorry but had to jump in here... R0H1T... are you seriously trying to suggest that ssd cell burnout is an unproven or controversial phenomenon?

Yeah, you really shouldn’t be commenting on tech boards. This is very basic. To deny it is like trying to argue 2+ 2 = 5

Facts matter. The “endurance” of an ssd drive is 100% correlated to is write cycles and is an extremely important metric for “pros” of all types for whom actual data storage is important.

 

[Parn]

Isn't using TLC one of the methods to increase max capacity? If so then why does 980pro still top out at 1TB? Guess it's time for me to move on from samsung SSDs for my main rig.

 

[R0H1T]

No, in the interest of this topic going off the rails & me having having to repeat everything 10x over here's what I said ~

NAND "endurance" or lifetime writes ~ which are guaranteed by the manufacturer are generally highly conservative for most, if not all consumer drives. You're likely to encounter an issue with the firmware &/or controller than simply running out of P/E cycles. There are no long term studies demonstrating how modern 3D NAND (or SSD) is affected by temps or indeed the usage pattern of users ~ your internet surfing grandparents or what some call themselves "power" users! TLC is a proven tech & is about just as "reliable" as MLC which is to say that your drive will not "wear" itself out just because it's TLC instead of MLC unless you're doing some atypical "server grade" writes. For 90% consumer workload TBW ratings are meaningless, the last 10% ~ are you sure it isn't the temps or firmware/controller that killed your drive instead of it exhausting its P/E cycles? If you can't answer that with any authority you have no right claiming whatever's being peddled in the last few pages!

And lastly here's the P/E numbers from the the first consumer 3D NAND drives, unlike the poster who's just making things up here's the facts ~

Endurance: Close to Planar MLC NAND
The big question with every new NAND generation is the endurance. We already saw 6,000 P/E cycles in the SSD 850 Pro and an amazing 40,000 P/E cycles in the SSD 845DC Pro, which proved that V-NAND provides substantially better endurance over today's planar NAND nodes. However, endurance was never really an issue with planar MLC NAND except in the enterprise space, so the 850 EVO with its TLC V-NAND offers a much more interesting insight to the capability of 3D NAND technology.

To test endurance, I put the 120GB 850 EVO through our usual endurance test suite. Basically I just used Iometer to write 128KB sequential data at queue depth of 1 to the drive while monitoring the Wear Leveling Count (WLC) and Total LBAs Written SMART values. The 'Current Value' of the WLC SMART value gives the remaining endurance as a percentage (starts from 99), whereas the 'Raw Data' value indicates the number of consumed P/E cycles. In order to estimate the endurance, I had to find the spot where the increase in 'Raw Data' value decreases the 'Current Value' by one.

It appears that TLC flavor of V-NAND is rated at about 2,000 P/E cycles. The raw WLC value seems to be based on the user capacity (i.e. 120GB = 1 P/E cycle) because just going by it puts the endurance at ~2,133 P/E cycles (128/0.06), but that doesn't add up with the raw NAND capacity and total data written. However, the estimated total write endurance (which is just 15,260/0.06) suggests that the NAND itself is rated at 2,000 P/E cycles, which would make sense as the number of P/E cycles is usually an even thousand and it's also inline with the increase that the 850 Pro saw (from 3,000 cycles in the 840 Pro to 6,000 cycles).

While write endurance in client workloads was never truly an issue even with planar TLC NAND, the doubled endurance in TLC V-NAND makes it practically impossible to wear out the drive before it has become totally obsolete. Only some very extreme workloads could wear out the smaller capacities before the warranty runs out, but the 850 EVO is a wrong drive for such workloads in the first place. All in all, there should be absolutely no reason to worry about the endurance of the 850 EVO, especially given the endurance ratings Samsung is giving to the 850 EVO (75TB for 120/250GB and 150TB for 500GB/1TB).

The 1000 P/E cycles numbers are for QLC & even there you won't find latest gen QLC P/E cycles being listed by SSD makers because they aren't relevant for most consumer workloads. Modern multi layer NAND are way more resilient than typical NAND, especially the TLC kind. So next time when challenging someone on facts, better find something other than alternate facts

What this thread & forum needs is people updating their knowledge & definitely not trying to sound obnoxiously like they know everything about NAND or SSD's just because they see ZOMG only half the TBW writes

Yeah, you really shouldn’t be commenting on tech boards. This is very basic. To deny it is like trying to argue 2+ 2 = 5

Next time you suggest something like that, at least have the decency go through the entire thread not just the last few replies!