Sign in to follow this  
Followers 0
CrazyElf

Is SSD progress starting to level off?

4 posts in this topic

Does it seem like SSD progress is beginning to level off?

Most current SSDs remain MLC planar, at around 16-20nm in size. There's a few value-oriented TLC SSDs as well. But at some point, it is looking like the die shrinks are going to start levelling off. Perhaps maybe ~10nm is the end?

Does it seem like 3D NAND too is no easy solution over planar? It does seem like Samsung in particular is stuck at 40nm for now with 3D NAND for at least the next couple of generations, and I'd imagine the other vendors would encounter similar technical difficulties. It may very well be that the benefits were somewhat overhyped. It's not clear I think whether it will scale down very well past 40nm either.

Performance-wise, I think we are seeing NVMe drives start to enter the market. Whether that will impact much save the high end though remains open to debate.

So is that it, are we seeing SSD performance level off then? Likewise, is the rate of price drop going to level off as well in terms of rate of cost declines in cost:gb?

Share this post


Link to post
Share on other sites

I don't think Samsung is "stuck"-- from my guess, it's by choice. Using a larger process means better individual cell lifetime, faster page program times (at least typically), less interference between cells...

http://www.anandtech.com/show/7237/samsungs-vnand-hitting-the-reset-button-on-nand-scaling

Given how easy it is to make more layers and stack them, I assume Samsung believes they can get enough near-term scaling at 40nm with enough layers (32 now, I think?)...

As far as seeing SSD performance chance, I think we're going to have to wait a few more months as NVMe with PCI-e links starts replacing the SATA 6gbps interface.

As far as pricing, IIRC there were a few articles a year or two back about how much it costs to produce NAND. Obviously they're out of date now, but the jist was the bulk of the SSD cost was coming from the cost of the NAND itself, which is very unlike a harddisk (where the platter cost/head cost does not directly affect the direct price of the drive nearly as much), so that bit of scaling as far as price is probably still true...

Share this post


Link to post
Share on other sites

The issue here is that past 16nm (Micron's current MX100 is 16nm), I'm not too sure how much smaller NAND can get. How small could it get? Maybe 10nm? 7nm?

It is not so much whether there will be progress or not. The issue is if the rate of progress is slowing down.

On the 3D NAND front, I guess we'll have to wait and see. I wonder how easy it will be to shrink the 40nm NAND. As it stands, the 850 Pro is competitive yes with high end drives, but it didn't really bring the kind of revolution expected by some. Only other option is to add more layers to the 3D NAND. Not sure about that one - whether more layers = technical problems (could be).

But right now, yeah I agree that large NAND is good in that it has a lot of life.

Another point is that the SSD industry is already pretty cut throat. Margins are already pretty thin. On one hand this means lower prices as is, but on the other, it does mean that prices have less room to fall.

Edited by CrazyElf

Share this post


Link to post
Share on other sites

Dang, I can't find the article I was looking for, but this reference (again, not the precise one I was looking for):

http://www.anandtech.com/show/7237/samsungs-vnand-hitting-the-reset-button-on-nand-scaling

The root cause is basic physics. To store 2 bits per cell in present day MLC NAND, we’re talking about counting a couple dozen electrons. Go to 3 bits per cell (TLC NAND) and the difference between levels shrinks further. Scale down to 14/15/16nm NAND and everything gets more difficult. We’re heading towards a situation where the difference between levels in a NAND cell are a matter of a handful of electrons.

Shows that shrinking NAND further is going to be tough...

http://www.anandtech.com/show/8066/crucial-mx100-256gb-512gb-review/2

Says IMFT has 3D NAND of their own at the moment, but looks like they're pushing 16nm MLC and thinking about an even smaller process. SK Hynix also has 3D NAND as well, ditto for Toshiba/Sandisk. This statement sounds speculative:

Toshiba and SanDisk believe that their 2D NAND is more cost efficient at the moment and thus see no need to move to 3D NAND yet. Their just announced 15nm NAND will be their last 2D NAND node but we probably won't see it in products until late this year or early next year, so that will cover the market for about a year before being replaced by 3D NAND.

And yes, as referenced somewhere in the many links I've posted, I believe Samsung has clearly stated additional layers aren't an issue, at least for the next version of 3D NAND (getting more than one version down the road, tho, I wouldn't be as sure-- I admit I just don't know, nor am I deep enough in the industry to know).

As it stands, the 850 Pro is competitive yes with high end drives, but it didn't really bring the kind of revolution expected by some.

I think that's more a SATA 6Gbps limit than anything else? Drives before the 850 Pro have already seemed to hit a performace ceiling-- many are very similar in performance-- that makes me think the SATA 6Gbps interface is a significant reason for performance limits. Going to a better interface stack with NVMe will help, but won't entirely remove the SATA 6Gbps limits...

We'll see where M.2 and SATA Express goes in the next generation. :D

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!


Register a new account

Sign in

Already have an account? Sign in here.


Sign In Now
Sign in to follow this  
Followers 0