Samsung's amazeballs 3D V-NAND SSD not THAT much better than predecessor

El Reg asked Samsung how the 3D V-NAND SSD's speed compared with the same company's SM843T 960GB SSD, a high-performance SSD used in data centres and servers released just a few months ago. The reply may surprise some hardened storage veterans. 3D V-NAND, announced on 7 August, stacks layers of NAND vertically to increase the …

[Piro]

I thought they were using 30nm class flash.

That's how they got the endurance: they went back in tiiiiime. Then stacked them up to increase density.

[Steven Raith]

Piro is correct

"We are also asking ourselves, what is the lithography of the NAND cells in 3D NAND: 24nm, 21nm or less? Or more?"

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

The whole point is that when you get down to sub 20nm scales for flash, you start to get cell to cell interference problems beyond a certain point - meaning lower yields and more errors.

The mature 30nm class is cheap (compared to building a new fab), reliable (there have been devices running 30nm flash for years so there are no 'process' problems) and I believe stacking the cells means you get capacity and speed that are also within expected performance for the class - which the rest of the article seems to confirm.

Unless I've totally misread everything I've seen about VNAND, natch.

[Flocke Kroes]

Up to 20mb broad band

Please do not get fobbed off with 'up to' figures. 'At least' is useful, as is 'average sustained'.

[Giles Jones]

More hype from Samsung, they're getting as bad as Apple.

[Piro]

Except not at all. Their ssds are reasonably priced, reliable, and fast.

If you're talking about smartphones, sure, but their ssds have little to do with that.

[Anonymous Coward]

Really a phone part?

There's plenty of space for chips in PC SSD's, phones on the other hand have limited real estate so a stacked part is valuable.

There's also the interesting possibility of stacking flash, a CPU and memory in the same chip creating a true PC on a chip.

[Anonymous Coward]

Re: Really a phone part?

If the above posters are correct, it's not about saving space, but using cheaper fabrication and hopefully then cheaper products. While it's a larger and older "chip", it's now "stacked" and so hopefully they can fit a higher density and greater number of chips on each production run.

But what's the chances the consumer does not see any savings? :(

[Anonymous Coward]

Re: Really a phone part?

PS, the "Raspberry Pi" already stacks a Ram Chip on a CPU. :P

[Anonymous Coward]

Why do we care about performance increases?

Better performance is always welcome, but for most non-enterprise needs flash is more than fast enough. What matters is 1) how much cheaper is it per bit? 2) is its erase lifetime no worse than before?

Since it isn't going to a smaller process, we can assume we're fine on the second point so the only question that matters is the first. The answer is probably that it is more expensive now, but Samsung believes that it will be less expensive (maybe much less expensive) in the future.

If so, that'll be more good for them than us, since flash is a commodity they'll sell at the prevailing price and if theirs costs half what everyone else's costs to make, they'll keep the rest as additional profit.

[Ammaross Danan]

Re: Why do we care about performance increases?

The erase lifetime is about 10x what it was before, lending to about 35k P/E cycles.

Also, am I the only one that noticed the Samsung quote was only regarding write speeds, but that the poor-at-researching author applied a boost to the read speeds, as if they were mentioned? Reads don't go from 500MB/s to 600MB/s just because sequential writes go up by 20%....

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