Do you think we will see any of the NAND manufacturers using the node after 15/16nm?

[cbn]

Do you think we will see any of the manufacturers using the node after 15/16nm for NAND production?

10nm planar NAND?

Toshiba?

http://www.anandtech.com/show/9113/...t-3d-nand?_ga=1.192737762.93334132.1463738514

I'm not surprised that Toshiba's first 3D NAND product is a 48-layer part because the company has the smallest planar NAND node (15nm), so anything less than 48 would likely not have been cost efficient enough when compared against the 15nm node. There has been a lot of semiconductor analyst chatter about the optimal first generation high volume 3D NAND layer count and the consensus seems to be that you really need more than 32 layers to be competitive with modern planar NAND nodes because 3D NAND requires a whole new set of manufacturing tools that aren't exactly cheap. Toshiba did clarify that planar and 3D NAND will exist in parallel for some time, which suggests that even with 48 layers 3D NAND may be more expensive to manufacture than 15nm planar NAND (although that may change as the process matures and yields improve).

 

[Hellhammer]

Samsung and SK Hynix are already on it: http://www.custompcreview.com/news/samsung-mass-produces-14nm-nand/27476/

 

[BSim500]

10nm planar NAND?

Estimated 10nm planar TLC/QLC Spec Sheet:-

P/E Cycles : 2
Unpowered data retention : 3 hours
1TB Cost : $800 ($50 NAND + $750 500-layers of error correction) :biggrin:

 

[hojnikb]

Estimated 10nm planar TLC/QLC Spec Sheet:-

P/E Cycles : 2
Unpowered data retention : 3 hours
1TB Cost : $800 ($50 NAND + $750 500-layers of error correction) :biggrin:

I doubt they will bother with TLC, let alone QLC. It's pretty tough making a reliable 16/15nm TLC flash, let alone something smaller.

 

[cbn]

14nm SK Hynix TLC has arrived:

http://www.goplextor.com/Product/Detail/S3C#/Spec

(Thanks Glaring_Mistake for pointing this out in the other thread)

 

[VirtualLarry]

My buddy was telling me that he had read that SSDs no longer care about performance, and are totally in a "race to the bottom", to make them cheaper and cheaper, while retail prices continue to rise, due to NAND shortages (ALLEGEDLY).

 

[dave_the_nerd]

My buddy was telling me that he had read that SSDs no longer care about performance, and are totally in a "race to the bottom", to make them cheaper and cheaper, while retail prices continue to rise, due to NAND shortages (ALLEGEDLY).

Well... that's half the truth.

 

[cbn]

I wonder with 14nm NAND out if we will see the return of the Phison S10 controller? (A controller that can only use planar NAND.....not 3D NAND).

 

[Glaring_Mistake]

I wonder with 14nm NAND out if we will see the return of the Phison S10 controller? (A controller that can only use planar NAND.....not 3D NAND).

Maybe, but I don't think so.

While I think Phison S10 does a pretty good job at managing voltage drift a lot of manufacturers would likely want LDPC ECC for this NAND (since endurance does not seem to be that great).
And since for SATA drives we seem to be transitioning to DRAMless controllers and/or controllers with fewer channels the Phison S11 would probably be more likely to be used.

Then there's the question of how many drives we will see using it though seeing as we're in a NAND shortage there may be those willing to stick with 2D NAND a bit longer.

 

[cbn]

While I think Phison S10 does a pretty good job at managing voltage drift a lot of manufacturers would likely want LDPC ECC for this NAND (since endurance does not seem to be that great).

That makes sense to me.....perhaps with 14nm MLC NAND?

With that mentioned, I believe SSDs pairing Phison controllers with non-Toshiba NAND have been pretty rare. (Partriot memory being the exception I believe).

 

[Glaring_Mistake]

That makes sense to me.....perhaps with 14nm MLC NAND?

That might work.

With that mentioned, I believe SSDs pairing Phison controllers with non-Toshiba NAND have been pretty rare. (Partriot memory being the exception I believe).

Yes, that is pretty rare, like it has happened but it is almost used exclusively with Toshiba NAND.
Patriot Memory is an exception however and I think is also the only one to use it with non-Toshiba TLC NAND.
Though if you go by the SMART-values it may not have been a very good combination (but I don't know if I really trust the Patriot Blast's SMART-values).

 

[Glaring_Mistake]

I wonder with 14nm NAND out if we will see the return of the Phison S10 controller? (A controller that can only use planar NAND.....not 3D NAND).

Also something I found when looking for information on Phison's StrongECC, according to Anandtech the S11 was designed with 14nm NAND in mind, see: http://www.anandtech.com/show/11236...vs-and-gfs-drives-featuring-3d-tlc-and-3d-mlc

 

[cbn]

Does anyone know what capacity the SK Hynix 14nm MLC die has? 14nm TLC die?

 

[Glaring_Mistake]

Short review of the Plextor S3C: https://translate.google.com/transl...or-s3c-px-512s3c-obemom-512-gb.html&sandbox=1

 

[IntelUser2000]

They'll be forced to. I don't think they can scale vertical stacking much above 128 stacks, if that. 3D NAND is supposed to be ~40nm process now? When the vertical headroom runs out they'll have to shrink. So in 5-10 years we'll see 128-stack 10nm TLC NAND. At least they have some time until then to improve techniques like error correction to make it not as horrible as it sounds.

 

[Glaring_Mistake]

I don't think they can scale vertical stacking much above 128 stacks, if that.

Maybe. Like, I can see that it looks to get more and more difficult to increase the number of layers.
But at the same time I recall reading about how they would move from using Floating Gates to using Charge Traps because a Floating Gate would just not work for lithographies smaller than ~20nm.
Yet, as we see in this thread then it seems that SK Hynix has made 14nm TLC NAND with a Floating Gate (I'm assuming it would be newsworthy if it was using a Charge Trap).

3D NAND is supposed to be ~40nm process now?

Like you said, 3D NAND is not bound to a ~40nm process.
Samsung has been using a ~40nm process (and likely still is) but as has been discussed around here recently it looks like Intel/Micron went with smaller lithographies (16 and 20nm for their first gen 3D NAND).
And on what lithographies the 3D NAND from either SK Hynix or Toshiba/WD is made there does not seem to be much mention of from any of them.

 

[cbn]

They'll be forced to. I don't think they can scale vertical stacking much above 128 stacks, if that. 3D NAND is supposed to be ~40nm process now? When the vertical headroom runs out they'll have to shrink. So in 5-10 years we'll see 128-stack 10nm TLC NAND. At least they have some time until then to improve techniques like error correction to make it not as horrible as it sounds.

With string stacking they can make chips with more layers by having two or more dies count as a single chip.....with eight or even sixteen chips per package (assuming the string stacked chips are not too thick)

So maybe with enough total SSD capacity (in the setting of WORM for cloud storage or say 10 GB writes per day in a client drive) a very high amount of error checking will not be required. This even with a smaller lithography? With that mentioned, Intel-Micron is supposed to eventually have four bits per cell on the Gen 2 (ie, 64 layer) NAND so I would think any lithography shrink would have to accomodate four bits per cell also.

 

[Glaring_Mistake]

So maybe with enough total SSD capacity (in the setting of WORM for cloud storage or say 10 GB writes per day in a client drive) a very high amount of error checking will not be required. This even with a smaller lithography? With that mentioned, Intel-Micron is supposed to eventually have four bits per cell on the Gen 2 (ie, 64 layer) NAND so I would think any lithography shrink would have to accomodate four bits per cell also.

Think I would still prefer some strong ECC for QLC NAND even with little wear.
However if it does behave like 15nm TLC NAND then I would be open to the idea that the ECC could be relaxed.

 

[Glaring_Mistake]

Plextor S3C was recently reviewed over at Tomshardware (a bit more thorough review than the previous one): http://www.tomshardware.com/reviews/plextor-s3c-ssd,5133.html
Kind of poor performance though that's not likely to come as much of a surprise.

 

[VirtualLarry]

Plextor S3C was recently reviewed over at Tomshardware (a bit more thorough review than the previous one): http://www.tomshardware.com/reviews/plextor-s3c-ssd,5133.html
Kind of poor performance though that's not likely to come as much of a surprise.

• PlexCompressor

SSD storage capacity is always important. Plextor's PlexCompressor uses custom smart-compression technology that gives you more storage capacity without impacting the user experience and system performance.

I'm wondering what that feature means. Is that some sort of additional driver software, for realtime compression, to make the drive seem larger than its native capacity? Or is that a hardware feature? (Like some cheap chinese USB sticks, I guess...?)

PlexCompressor is the standout feature that some users will find useful for a few reasons. This software will automatically compress files you haven't accessed recently, reducing the amount of space they consume on the drive. This frees up space, and as we've proven in the past, that eases the performance roll off due to data accumulation on the flash.

Hmm, seems like a driver / background app.