Wednesday, April 8th 2026

SK hynix Begins Shipping 321-Layer QLC NAND cSSDs, Dell First Customer

SK hynix has started shipping its PQC21 cSSD, the first product built on its 321-layer QLC NAND flash technology. The product is available in 1 TB and 2 TB capacities, with Dell as the first customer starting this month. QLC (Quad-Level Cell) stores four bits per cell, which maximizes storage density per unit area but has historically come with a write performance penalty compared to TLC (Triple-Level Cell). SK hynix addresses that with SLC caching, frequently accessed data is written to faster SLC-mode regions first before being transferred to the QLC cells, smoothing out the performance gap for typical workloads.

According to market research from IDC (International Data Corporation), the QLC NAND is expected to go from 22% of the global cSSD market in 2025 to 61% by 2027, so the timing of this launch makes sense. SK hynix says it will expand beyond Dell to other major customers as production scales up.
On the broader NAND market, Samsung is still on top with $6.6 billion in revenue, up 10% QoQ, though its share slipped to 28% as bit shipments dropped during a process transition. SK Group (SK hynix and Solidigm combined) had the strongest quarter of the top five, with revenue up 47.8% QoQ to around $5.21 billion and market share climbing to 22.1%, driven by mobile NAND and enterprise SSD demand.

Source: SK hynix
Add your own comment

22 Comments on SK hynix Begins Shipping 321-Layer QLC NAND cSSDs, Dell First Customer

#1
lexluthermiester
SLC caching can only do so much. It can't solve the QLC NAND degradation problems either.

Still, 321 layers is impressive. Now let's see a TLC version.
Posted on Reply
#2
Chrispy_
Yuck.

What's the raw QLC write rate and how many TBW for a 1TB drive, for example?
I'm sure it's enough to last the warranty period for a thin-and-light that primarily runs webapps on Windows, but nobody wants to buy the absolute bare minumum of quality or performance.
Posted on Reply
#3
bonehead123
Chrispy_nobody wants to buy the absolute bare minumum of quality or performance.
Nope, not doin that !

WD4Me4Eva, hahaha :D
Posted on Reply
#4
TheinsanegamerN
lexluthermiesterSLC caching can only do so much. It can't solve the QLC NAND degradation problems either.

Still, 321 layers is impressive. Now let's see a TLC version.
Yes sir I would like the 32TB M.2 SSD please.

Please.
Posted on Reply
#5
Chrispy_
TheinsanegamerNYes sir I would like the 32TB M.2 SSD please.
"That'll be $4000, thanks very much. Oh, and minimum order quantity is 10,000 units and you'd be the 118th AI datacenter customer in the queue for an order so the lead time is approximately 18 months if you complete payment today"
Posted on Reply
#6
Wirko
bonehead123Nope, not doin that !

WD4Me4Eva, hahaha :D
W.D. loves Q.L.C. too, more than ever before.
Posted on Reply
#7
kondamin
Chrispy_"That'll be $4000, thanks very much. Oh, and minimum order quantity is 10,000 units and you'd be the 118th AI datacenter customer in the queue for an order so the lead time is approximately 18 months if you complete payment today"
Just 40 million and I'd be in 118th place?
Posted on Reply
#8
Chrispy_
kondaminJust 40 million and I'd be in 118th place?
That was an hour ago. Now it's 44 million and you'd be in 123rd place.
Chop chop, get that order in before we run out of allocation from TSMC for our SSD controllers!
Posted on Reply
#9
Athena
I wonder how much spare area we are up to with these? 15%?

then I wonder how much work the controller is doing for error correcting, it gets way more complex with the higher and higher density they are hitting
Posted on Reply
#10
bonehead123
Athenait gets way more complex with the higher and higher density they are hitting
"Complex is as complex does"
"Density is as density does" - Forest, Forest Gump (well, sorta :))

At some point in the near future, surely they will reach a point of diminishing/no returns on the layers & their densities, yes ?
Posted on Reply
#11
chrcoluk
AthenaI wonder how much spare area we are up to with these? 15%?

then I wonder how much work the controller is doing for error correcting, it gets way more complex with the higher and higher density they are hitting
This can be tested to a degree by writing data, then not refreshing the data for a few months, and see how fast it can read the data back, if its slow then its probably heavy error correction at work.

My SA510's have been like this, they fast when freshly written but can get very slow even after only a month.

Problem is benchmark tools just read back freshly written data, so its not a problem easily exposed.
Posted on Reply
#12
_roman_
the storage just has to deliver the review and first day performances.

long time endurance does not matter for certain companies like hp or dell. the device shall be defective or slow or malfunctioning in year 3 or 4 or 5.

Dell should solder those storage and dram to the motherboard to improve short and long term profit
Posted on Reply
#13
Bwaze
bonehead123"Complex is as complex does"
"Density is as density does" - Forest, Forest Gump (well, sorta :))

At some point in the near future, surely they will reach a point of diminishing/no returns on the layers & their densities, yes ?
We're already deep into this territory, at least for consumers.

I bought Super TalentUltradrive ME 64GB SSD back in 2009 for 160 EUR, and for a decade I could double the capacity every two years for the same amount. Until 2020, when M.2 format arrived, and then we hit a brick wall, and half a decade later we're actually going backwards. QLC RAMless drives that top at 2 TB?
Posted on Reply
#14
Scrizz
chrcolukThis can be tested to a degree by writing data, then not refreshing the data for a few months, and see how fast it can read the data back, if its slow then its probably heavy error correction at work.

My SA510's have been like this, they fast when freshly written but can get very slow even after only a month.

Problem is benchmark tools just read back freshly written data, so its not a problem easily exposed.
Enterprise drives generally have some sort of data refresh. This refresh keeps the information fresh and ready for use. Thankfully, your understanding and experience with consumer grade SATA SSDs does not apply to enterprise drives.
The following paper should be enough to give an idea of one of a myriad of solutions in the enterprise space. :toast:
ieeexplore.ieee.org/document/6378623
BwazeWe're already deep into this territory, at least for consumers.

I bought Super TalentUltradrive ME 64GB SSD back in 2009 for 160 EUR, and for a decade I could double the capacity every two years for the same amount. Until 2020, when M.2 format arrived, and then we hit a brick wall, and half a decade later we're actually going backwards. QLC RAMless drives that top at 2 TB?
yeah keep a QLC dramless SSD far away from me. :laugh:
Posted on Reply
#15
KLH Techsystems
I had a QLC SSD included as part of a mini PC I got a while back. Atrocious performance if you wrote more than a few gigabytes continuously, but for basic desktop usage (i.e. what I got the mini PC for), it was perfectly fine. The bigger concern is whether the cost advantage of QLC is being leveraged for or against consumers.

Outside of the consumer market, the QLC enterprise drives I've seen actually have decent sustained write performance. Not as good as TLC drives, but nowhere near the sheer garbage of consumer QLC SSDs. I personally would be more than willing to go with QLC enterprise drives for a cost and/or capacity advantage.
Chrispy_"That'll be $4000, thanks very much. Oh, and minimum order quantity is 10,000 units and you'd be the 118th AI datacenter customer in the queue for an order so the lead time is approximately 18 months if you complete payment today"
Jokes aside, $4000 would be pretty good for a 32 TB M.2 SSD, I'd probably pick up 4 or so, allowing me to relegate all hard drives to cold storage / backup duty. If I remember correctly, there were even 30.72 TB QLC U.3 enterprise SSDs available for around $3000 before the storage apocalypse.
Posted on Reply
#16
Wirko
lexluthermiesterStill, 321 layers is impressive. Now let's see a TLC version.
Not that it isn't impressive but Samsung has 280-layer TLC V-NAND v9 in later variants of the 990 Pro. Not sure when it came to the market, GoogAI says mass production of chips started in April 2024, TLC first, QLC later.
Posted on Reply
#17
Athena
bonehead123"Complex is as complex does"
"Density is as density does" - Forest, Forest Gump (well, sorta :))

At some point in the near future, surely they will reach a point of diminishing/no returns on the layers & their densities, yes ?
I'm sure there is a hard limit, but, some NAND companies are projecting 1000 layers
heck, SK Hynix seems to be heading toward PLC NAND
then after that, perhaps HLC (hexa-level cell), or maybe a new storage type will be made

Posted on Reply
#18
EsaT
_roman_the storage just has to deliver the review and first day performances.

long time endurance does not matter for certain companies like hp or dell. the device shall be defective or slow or malfunctioning in year 3 or 4 or 5.

Dell should solder those storage and dram to the motherboard to improve short and long term profit
And actual data retention/"non-volatility" after some wear wasn't tested in some of those endurance tests of review sites.
With pretty much nearing analog 16 charge/voltage levels for cell QLC isn't going to be good in that.

Pretty sure manufacturers know retention behaviour, but it's even more sure those numbers are under lock and key.
Athenaheck, SK Hynix seems to be heading toward PLC NAND
then after that, perhaps HLC (hexa-level cell), or maybe a new storage type will be made
Gotta get back to floppy drive's writing speeds and DRAM's data retention...
Posted on Reply
#19
_roman_
i agree on that. i saw personally the first dead sata m2 2280 drive last year. more will follow.

the argument ssds do not die because of flash based is therefore most likely gone.
Posted on Reply
#20
Chrispy_
AthenaI'm sure there is a hard limit, but, some NAND companies are projecting 1000 layers
heck, SK Hynix seems to be heading toward PLC NAND
then after that, perhaps HLC (hexa-level cell), or maybe a new storage type will be made

We probably won't have to deal with PLC as consumers; They've been trying to make it work for almost a decade already but the hits to endurance, performance, and controller complexity simply don't offset the savings of 5 bits per cell instead of 4 when you're only talking about pitifully tiny 2GB, 4GB, or 8GB consumer capacities.

If you're talking about a 200TB all-flash storage server then maybe it's coming soon but the problem with those $100K+ appliances is that they need to five-nines, enterprise-grade, mission-critical reliability with automatic drive replacement under warranty within a 4-hour window. Even QLC is risky for that (I currently have one all-flash Pure Storage array based on QLC drives and so far it's been good, but it's also only a couple of years old Time will tell and all those Intel 660p QLC drives have not fared well as storage array members. They spawned a 5 year period of me avoiding QLC like the bubonic plague!)
Posted on Reply
#21
chrcoluk
ScrizzEnterprise drives generally have some sort of data refresh. This refresh keeps the information fresh and ready for use. Thankfully, your understanding and experience with consumer grade SATA SSDs does not apply to enterprise drives.
The following paper should be enough to give an idea of one of a myriad of solutions in the enterprise space. :toast:
ieeexplore.ieee.org/document/6378623


yeah keep a QLC dramless SSD far away from me. :laugh:
Some consumer drives do this as well, but for some reason they dont do it to every drive. I guess lessons not properly learned from the Samsung 840's.

I am pretty sure the 4TB SA510 is QLC. :laugh: Opened one of them up and posted pics on here, there is a mysterious big Q engraved on the nand chips. :laugh: Could have got a red or 870 EVO for less than 20% more, obviously not now though, really annoyed with myself.
Posted on Reply
#22
remixedcat
Nomad76QLC NAND is expected to go from 22% of the global cSSD market in 2025 to 61% by 2027
when will we see PLC NAND and disposable SSDs. I'm sure SK is frothing at the mouth rn. Single digit endurance marketed in a way that will make it seem better cuz smaller size for eugenia cooney level thin laptops. but hey it'll be super secure right? so secure that the disk will self destruct like in those cheesy movies.
_roman_Dell should solder those storage and dram to the motherboard to improve short and long term profit
don't give em any imore deas. they tried that crap w xps and ppl didn't take it well even with TLC NAND. so many videos I've watched dellpartspeople on YT he felt very strongly about those.
Posted on Reply
Add your own comment
Nov 12th, 2026 04:43 CST change timezone

New Forum Posts

Popular Reviews

TPU on YouTube

Controversial News Posts