Samsung Unveils Massive 32 TB SSD Leveraging 64-layer 3D V-NAND

[btarunr]

At the 2016 Flash Memory Summit, Samsung unveiled a massive 32-terabyte solid state drive (SSD) for enterprise applications. The drive leverages the company's 4th generation, 64-layer 3D V-NAND flash memory, making it the company's highest-capacity SSD, upping the 15.36 TB drive the company unveiled in March, 2016. The drive is built in the 12 mm-thick, 2.5-inch form-factor, and features the SAS 12 Gb/s interface. At the summit, the company announced that it plans to have 100 TB SSDs ready by 2020.



View at TechPowerUp Main Site

 

[P4-630]

Nice, now I can install my OS and all my games on the same drive

 

[Prima.Vera]

^-- Really...

On a different note, if they can produce drives with such a huge capacity, why aren't any "normal" size drives in the offers for reasonable prices, like 1,2,3 or 4 TB?

BTW, Seagate released a 60TB drive on the same day:

http://technology.inquirer.net/50843/60tb-ssd-unveiled-by-seagate-at-flash-memory-summit-conference

 

[natr0n]

^-- Really...

On a different note, if they can produce drives with such a huge capacity, why aren't any "normal" size drives in the offers for reasonable prices, like 1,2,3 or 4 TB?

BTW, Seagate released a 60TB drive on the same day:

http://technology.inquirer.net/50843/60tb-ssd-unveiled-by-seagate-at-flash-memory-summit-conference

Just like any company they want to juice us for as long as possible and very slowly trickle in the good stuff.

 

[nemesis.ie]

^-- Really...

On a different note, if they can produce drives with such a huge capacity, why aren't any "normal" size drives in the offers for reasonable prices, like 1,2,3 or 4 TB?

BTW, Seagate released a 60TB drive on the same day:

http://technology.inquirer.net/50843/60tb-ssd-unveiled-by-seagate-at-flash-memory-summit-conference

Agreed, and with this kind of density we should be able to have M.2 style format drives of larger capacities and/or cheaper prices too.

Seagate's 60TB is 3.5" though, so the Samsung is a little better physical space-wise and two would be 64TB.

 

[ZeDestructor]

^-- Really...

On a different note, if they can produce drives with such a huge capacity, why aren't any "normal" size drives in the offers for reasonable prices, like 1,2,3 or 4 TB?

BTW, Seagate released a 60TB drive on the same day:

http://technology.inquirer.net/50843/60tb-ssd-unveiled-by-seagate-at-flash-memory-summit-conference

Just like any company they want to juice us for as long as possible and very slowly trickle in the good stuff.

Are any of you man enough to pay the about 30c per GB (that works out to 18000USD for 60TB, 9600USD for 32TB, 4800USD for 16TB, 1200USD for 4TB, 300USD for 1TB) those drives would cost?

No? I thought as much. Fact is, NAND isn't particularly overpriced, not even in the enterprise space - it just scales linearly with capacity. True, you can get real cheap drives at 20c per GB, but then you're not getting as good a controller+firmware combo.

Agreed, and with this kind of density we should be able to have M.2 style format drives of larger capacities and/or cheaper prices too.

Seagate's 60TB is 3.5" though, so the Samsung is a little better physical space-wise and two would be 64TB.

M.2 is fundamentally limited in it's capacity to 8 NAND packages (4-6 if you demand full power-loss protection as the capacitors eat up the space) right now due to space constraints, and that's not changing anytime soon, so no, you can't just magically make magically reach parity with the 2.5" and 3.5" drives on M.2.

Capacities will grow.. maybe a 2TB M.2 drive this year, but 2.5" (and to a lesser extent 3.5") is where the real big drives will come.

 

[nemesis.ie]

M.2 is fundamentally limited in it's capacity to 8 NAND packages (4-6 if you demand full power-loss protection as the capacitors eat up the space) right now due to space constraints, and that's not changing anytime soon, so no, you can't just magically make magically reach parity with the 2.5" and 3.5" drives on M.2.

Capacities will grow.. maybe a 2TB M.2 drive this year, but 2.5" (and to a lesser extent 3.5") is where the real big drives will come.

I wasn't/am not expecting parity, but the increase in layers (higher density) does mean more capacity for the same number of packages.

 

[dj-electric]

Honestly, 2-4TB ssds should've already been in most homes today.

These demonstrations of power tell you how dense and available these just are.

It's time. We had enough from the wd Seagate cartel of hdds

 

[ZeDestructor]

I wasn't/am not expecting parity, but the increase in layers (higher density) does mean more capacity for the same number of packages.

True, but will consumers pay for it?

Honestly, 2-4TB ssds should've already been in most homes today.

These demonstrations of power tell you how dense and available these just are.

It's time. We had enough from the wd Seagate cartel of hdds

As I asked literally 2 posts above yours, are you gonna pay the more or less flat 30c per GB (1200 for 4TB, 600 for 2TB) that those drives are asking for? I highly doubt it.

Fact is, the consumer market on the whole isn't willing to pay for it, so nobody produces it. Simply because they can't afford it (small outfits like PNY, Corsair and such), or the really big guys (Intel, Micron, SanDisk) don't care for it cause servers are way more lucrative. Samsung is the exception that pushes that far, because they can afford the relatively small loss on the consumer-side (basically a tiny amount of R&D in tweaking their firmware for consumer loads and managing a few extra SKUs) in exchange for market visibility.

In 2-3 more years, sure, you'll have your 2-4TB SSDs for consumers, but right now 480GB and 240GB is where the numbers are made, and 1TB is kindof up and coming.

EDIT: To add on that, up until very, very recently (48-layer stacked NAND), it was literally impossible to build drives over 1TB in the 2.5" 7mm form factor because one simply could not fit enough packages in the volume. This all changed a year ago for Samsung, and right about now for the rest of the industry.

 

[nemesis.ie]

True, but will consumers pay for it?

The higher density should mean lower cost per GB as time goes by (and R&D is recouped and they have made money from enterprise and perhaps have the "next thing" there) so should mean we get the higher capacities for less per GB (eventually) in the consumer space. That's the way things have always gone.

The first step is to cram it in, the next to lower the price to end users. It may not take 2-3 years, the amount of data people need to store seems to only be growing, I'm sure 4K video will fuel that too once we see more content and 4k action cams etc. appear.

 

[ZeDestructor]

The higher density should mean lower cost per GB as time goes by (and R&D is recouped and they have made money from enterprise and perhaps have the "next thing" there) so should mean we get the higher capacities for less per GB (eventually) in the consumer space. That's the way things have always gone.

The first step is to cram it in, the next to lower the price to end users. It may not take 2-3 years, the amount of data people need to store seems to only be growing, I'm sure 4K video will fuel that too once we see more content and 4k action cams etc. appear.

Unlikely. Would have already happened if that was the case. Right now, Samsung just ships fewer packages if they can get away with it instead of going bigger. The decline in pricing will continue on more or less the same path as we've been o n for a while since the bits produced per time is more or less constant - stacked NAND is produced by simply stacking planar NAND, no not much improvement.

As for video, most consumers won't care, or will do a small jump to a bigger external HD. For the vast majority, it'll go to the cloud then be deleted from local storage.

For the few of us professionals and prosumers, we'll just have to suck up the pricing of real server gear.

 

[nemesis.ie]

Lets just agree to disagree and leave it at that and we'll see what transpires over the next year or two.

 

[ZeDestructor]

Lets just agree to disagree and leave it at that and we'll see what transpires over the next year or two.

Just telling it how the market is moving based on numbers from the various cloud providers (think Google and Amazon) and various opinions in various developer and user circles. As for pricing, that's just from observing trends and known pricing.

I'd love it if things changed to massive local storage, but like many, I've kinda given up and resigned myself to effectively building and running my own personal cloud instead - associated costs and all.

 

[nemesis.ie]

J

I'd love it if things changed to massive local storage, but like many, I've kinda given up and resigned myself to effectively building and running my own personal cloud instead - associated costs and all.

I'm more interested in the price coming down further, the problem is if they reduce the number of packages the speed may come down too, so we need a higher speed per package too.

It'll be interesting to see what things will look like with e.g. a single xpoint (or similar tech from someone else) package in a year or two,

 

[ZeDestructor]

I'm more interested in the price coming down further, the problem is if they reduce the number of packages the speed may come down too, so we need a higher speed per package too.

It'll be interesting to see what things will look like with e.g. a single xpoint (or similar tech from someone else) package in a year or two,

Based on AT's tests of all the 850 EVO SKUs besides the 4TB version (120GB, 250GB, 500GB and 1TB here) it seems that the loss of packages isn't that much of an issue. The bigger issue seems to be the lack of free space to do GC and other array maintenance well at the lower capacities. The relevant tests that show that are all the tests comparing between the 500GB and 1TB variants (4 vs 8 packages, near-identical perf), and the consistency and sequential tests of the 120GB (1 NAND package) vs all others.

EDIT: on the subject of XPoint, we just won't see them in the cheap SATA parts, and I doubt at anything less than 4x the cost of NAND, which makes the entire argument moot. Would be great as a caching layer in front of all-flash arrays though...