QLC NAND on the way...

[Elixer]

with 100 – 150 P/E cycles.

Once this hits mainstream, seems capacity will be higher, and prices, should be cheaper, but, the controller that is using QLC NAND would be much more complex than previous controllers.

I am guessing this will be mainly used for flash pen drives, and maybe some bottom of the barrel NvME/SATA units.

Can't say I like this trend.

 

[cbn]

planar QLC?

 

[Glaring_Mistake]

planar QLC?

Nope, 3D QLC NAND.
Planar QLC NAND would probably not be... very good.

 

[Glaring_Mistake]

with 100 – 150 P/E cycles.

Once this hits mainstream, seems capacity will be higher, and prices, should be cheaper, but, the controller that is using QLC NAND would be much more complex than previous controllers.

I am guessing this will be mainly used for flash pen drives, and maybe some bottom of the barrel NvME/SATA units.

Can't say I like this trend.

Not sure how we consumers will get to see QLC NAND in use but apparently some companies want really large QLC NAND drives (at least Facebook would like some 100TB drives using QLC NAND).
Quoting Tomshardware on use: "QLC NAND will be used only to store infrequently-accessed WORM (Write Once Read Many) data, such as archival and cold data, so random write performance is a minor concern."

 

[Elixer]

That use case makes sense, let's just hope it don't penetrate the consumer market very much.

 

[VirtualLarry]

Yeah, sounds like a consumer product using this tech would be a pretty poor application of this tech.

Then again, SSD prices continue to go up, strangely, while they continue to severely cost-cut.

I want some decent MLC SSDs, with DRAM buffer for mapping tables, and a SATA6G interface. And I want a 1TB one, for $100-120.

 

[cbn]

I want some decent MLC SSDs, with DRAM buffer for mapping tables, and a SATA6G interface. And I want a 1TB one, for $100-120.

+1

 

[Glaring_Mistake]

That use case makes sense, let's just hope it don't penetrate the consumer market very much.

Sure, good use case as long as they get the retention right.
Otherwise, not so much.

 

[bononos]

Not sure how we consumers will get to see QLC NAND in use but apparently some companies want really large QLC NAND drives (at least Facebook would like some 100TB drives using QLC NAND).
Quoting Tomshardware on use: "QLC NAND will be used only to store infrequently-accessed WORM (Write Once Read Many) data, such as archival and cold data, so random write performance is a minor concern."

I'll bet these type of storage will end up in cheap flash drives and counterfeits.

 

[fleshconsumed]

Eh, I'm still sticking with MLC for my OS drives and traditional spinners for media/data storage. I view QLC as a middle ground between MLC drives and spinners. Datacenters may benefit from QLC, higher speed than spinners and lower power consumption, but I don't see any real use for QLC for home use, not unless they bring the cost down to below that of the spinners. Datacenters may justify higher cost through lower electricity costs, home users primarily care about the up front cost. To be honest everyone kept saying years ago that flash technology will eventually become cheaper per TB than spinners, but I don't see that happening, so I fully expect to keep my spinners running for a long long time.

 

[cbn]

Update:

http://www.anandtech.com/show/11590/toshiba-768-gb-3d-qlc-nand-flash-memory-1000-p-e-cycles

(1000 P/E cycles for Toshiba 3D QLC).

 

[Glaring_Mistake]

Update:

http://www.anandtech.com/show/11590/toshiba-768-gb-3d-qlc-nand-flash-memory-1000-p-e-cycles

(1000 P/E cycles for Toshiba 3D QLC).

Damn it, you just managed to get that in before I completed my comment.

Anyway, I think Kristian Vättö once said 3D QLC NAND was supposed to have similar endurance to 15nm 2D TLC NAND but since then I have only seen estimates of like 100-150 P/E.
If it does behave similar then voltage drift may not be that difficult to keep under control.
Also means we may relatively soon see it used in for example SSDs.

 

[cbn]

Anyway, I think Kristian Vättö once said 3D QLC NAND was supposed to have similar endurance to 15nm 2D TLC NAND but since then I have only seen estimates of like 100-150 P/E.

Yep, the first Anandtech article on Toshiba 3D QLC also mentioned 100 to 150 P/E cycles---> http://www.anandtech.com/show/11585/western-digital-announce-bics4-96-layer-nand

 

[cbn]

I wonder what kind of controller support we will see for Toshiba 3D QLC?

SATA 6 Gbps?

 

[cbn]

Marvell specifically mentions 3D QLC support for their 88NV1160 dram-less PCIe 3.0 x 2 controller:


http://www.marvell.com/company/news/pressDetail.do?releaseID=8416

• NANDEdge™ LDPC error-correction technology boosts SSD endurance and reliability and supports 15nm 2D TLC and 3D TLC/QLC

 

[VirtualLarry]

Given the current (high) prices for PCI-E M.2 SSDs, I might be willing to give a 3D QLC PCI-E M.2 SSD a shot for consumer usage, with one of those Marvell controllers. Doesn't sound so aweful. At least, I'm not putting it through insane workloads.

Though, I'm still very impressed with my first PCI-E M.2 SSDs, some Samsung SM951 AHCI 128GB models. Even at the smaller NAND capacity of those drives, they performed AMAZING well, and because of the AHCI interface, they installed with Win7 64-bit using stock drivers.

I was a bit less impressed with both my 256GB Intel 600p M.2 SSD (3D TLC, I think), and my Adata XPG SX8000 SSD (3D MLC). The Intel was slower than the Samsung, by a noticeable amount, and the Adata slower by a lesser amount. (Well, maybe not slower, per se, but had some "pauses", which I attribute to the controller, an SMI something-or-other, that is used by both the Intel and Adata PCI-E M.2 SSDs.)

 

[Glaring_Mistake]

Marvell specifically mentions 3D QLC support for their 88NV1160 dram-less PCIe 3.0 x 2 controller:


http://www.marvell.com/company/news/pressDetail.do?releaseID=8416

Thought I had some memory of a controller with support for 3D QLC NAND but couldn't find any.
However with your finding that I checked if the Marvell 88NV1120 and 88NV1140 also support 3D QLC NAND and it appears they do.
It's not really made clear in that article but you can see in this that they do: https://origin-www.marvell.com/storage/ssd/88NV11XX/

Or maybe it is possible that any of their controllers using their NANDEdge ECC may support 3D QLC NAND: https://origin-www.marvell.com/storage/ssd/technologies.jsp

 

[Glaring_Mistake]

I wonder what kind of controller support we will see for Toshiba 3D QLC?

SATA 6 Gbps?

Was looking for some information on Phison's StrongECC and it seems that Phison's StrongECC and LDPC ECC can be used with 3D QLC NAND.
See: https://www.flashmemorysummit.com/English/Collaterals/Proceedings/2016/20160810_FE21_Lin.pdf

So the Phison S11 and E8 may support it as well as some other controllers for memory cards and such.
Which means that combined with Marvell there may actually be a number of controllers already out with support for 3D QLC NAND.

 

[cbn]

For the dual channel controllers, I wonder if we will finally see SSDs (using 3D QLC) with 180GB capacity? (ie, two 768 Gbit/96GB dies)

With SSDs using the Micron 384 Gbit (48GB) 3D TLC dies we didn't see 180GB because OEMs used 3 dies (rather than four) on a four channel controller in order to fit the established capacity of 128GB.

 

[VirtualLarry]

It's not like there havn't been 180GB SSDs in the past, and that it's a 'weird' size for consumers to intuitively understand.

 

[cbn]

It's not like there havn't been 180GB SSDs in the past, and that it's a 'weird' size for consumers to intuitively understand.

Yep, Intel had some 180GB capacity SSDs (eg, 520 Series and 330 Series).....but they also came in lower capacities.

https://ark.intel.com/products/series/66202/Intel-SSD-520-Series

http://ark.intel.com/products/series/67297/Intel-SSD-330-Series

So I am guessing the Micron 3D TLC drives were made from an odd number of dies because the 128GB capacity is so popular.

 

[Azuma Hazuki]

100-150 p/e cycles is USB thumb drive territory. I really hope this only ends up in something like a large NAS array, not as a replacement for current MLC/TLC consumer SSDs.

 

[cbn]

100-150 p/e cycles is USB thumb drive territory. I really hope this only ends up in something like a large NAS array, not as a replacement for current MLC/TLC consumer SSDs.

Anandtech posted an update:

http://www.anandtech.com/show/11590/toshiba-768-gb-3d-qlc-nand-flash-memory-1000-p-e-cycles

So now Toshiba's 3D QLC has an endurance of 1000 P/E cycles.

 

[cbn]

So if 120GB/128GB capacity is very popular.....

Where does that put SATA 6 Gbps SSDs made with this 3D QLC?

DRAM-less dual channel controller coupled to two dies of Toshiba 3D QLC for 180GB capacity vs. DRAM buffered quad channel controller coupled to 3 dies of Micron 3D TLC for 128 GB capacity?

Does the dual channel dram-less controller have enough performance for people that mostly only wanted 128GB SSD but were thinking they might need some extra room in the future?

 

[Glaring_Mistake]

100-150 p/e cycles is USB thumb drive territory. I really hope this only ends up in something like a large NAS array, not as a replacement for current MLC/TLC consumer SSDs.

3D QLC NAND may actually have a higher endurance than that (at least Toshiba/WD's).
Still I have heard that some drives with TLC NAND coming out (or already released) are supposed to be rated at around 100 P/E.

Also, according to the SMART-values of one drive I own that drive would look to be rated at somewhere between 100 and 200 P/E.
Have my doubts as to whether the SMART-values actually happen to be accurate though.

So if 120GB/128GB capacity is very popular.....

Where does that put SATA 6 Gbps SSDs made with this 3D QLC?

DRAM-less dual channel controller coupled to two dies of Toshiba 3D QLC for 180GB capacity vs. DRAM buffered quad channel controller coupled to 3 dies of Micron 3D TLC for 128 GB capacity?

Does the dual channel dram-less controller have enough performance for people that mostly only wanted 128GB SSD but were thinking they might need some extra room in the future?

That may kind of depend on whether there actually is a point where SSDs are deemed to have become too slow to be sold.
But if 3D QLC NAND doesn't just have similar endurance to but also performs like 15nm TLC NAND then we have already seen that people will accept it, even if combined with a dual channel, single core DRAMless controller.
Should it be a bit more difficult to get similar performance out of 3D QLC NAND than that, well - then we'll see.