Slashdot Mirror
Samsung SSD 850 EVO 32-Layer 3D V-NAND-Based SSD Tested
MojoKid writes Samsung just took the wraps off a new family of mainstream solid state drives, targeting the market segment previously occupied by its popular SSD 840 EVO series. The new Samsung SSD 850 EVO series is the follow-up to the company's current flagship SSD 850 PRO, but the new EVO is Samsung's first to pack 32layer 3D VNAND 3-bit MLC flash memory. The move to 32layer 3D VNAND 3-bit MLC flash brings pricing down to the .50 to .60 per GiB range, but doesn't adversely affect endurance because the cell structure doesn't suffer from the same inherent limitations of planar NAND, since the cells are stacked vertically with the 3D VNAND. The new 850 EVO drive performs well with large sequential transfers and also offered very low access times. The compressibility of the data being transferred across the Samsung SSD 850 EVO had no impact on performance and small file transfers at high queue depth were fast. Small file transfers with low queues depths, which is what you'd expect to see with most client workloads, were also very good. The Samsung SSD 850 EVO drives also put up excellent numbers in trace-based tests like PCMark 7.
Anytime the price and reliability of SSD improves it makes it more viable for end users and business work stations. If I had a bigger budget, every workstation would currently have an SSD.
They're not stacking silicon wafers on top of each other. Rather, they're putting more layers of oxide, semiconductor, etc onto each wafer in order to produce the 3d stacking. Yes, it's more complex. But it's a pretty mature technology.
Help find a cure for cancer. Join the [H]orde
.50 or .60 what per GiB?
Quarts? Furlongs? Solar masses?
According to AnandTech, they move from 20nm back up to 40nm for this tech!
if you're here you should be able to tell the difference between reviewing the first entrant in a new product category and every subsequent device.
My God, it's Full of Source!
OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
Well, with so many layers your process control must be extremely good which means a more mature process. Here's a good illustration, they're like little wells with regular gates down the sides.
Live today, because you never know what tomorrow brings
Fuck that.
Why no PCI-E model?
With increased density from 32 layers (despite larger feature size) why don't they have a 2tb (or 1920gb) model yet?
Anandtech wrote:
Initially I was told that the 850 EVO would come in 2TB capacity as well, but later on Samsung opted against it due to the limited demand.
Most likely because there's no savings whatsoever, if a 1TB drive is $500 then 2TB is probably like $980. They scale almost perfectly, all you need is an extra SATA port and you'd get a lot better performance with two in RAID0.
Live today, because you never know what tomorrow brings
No, actually for flash it does NOT have to be extremely good. Flash memory production can handle lots of bad bits due to process errors. The controller firmware will simply map-out the bad bits, simple as that. If 1% of the flash cells are lost due to process errors, it just isn't a problem.
The same cannot be said for cpu and ram logic.
-Matt
The summary fails to mention the 5 year warranty, which is obviously quite fantastic. It was only a few years ago many hard drive manufacturers were cutting back from 3 years to 1. A quick survey of amazon indicates many HDDs are currently offering a 2 year warranty. I'd be peeved if a drive died at 2 1/2 years. 5 1/2, not so much.
(Clippy pops up)
Hi, I see you're trying to write in English!
Can I help?
Did you mean to say "I'm trying to sell Samsung tech and want you not to realize this is a PR puff piece"?
-- Tigger warning: This post may contain tiggers! --
The question is, were they already at 1% before they started stacking 32 on top of each other, because the drive firmware sure as hell can't cover up for 32%* of the drive being screwed.
It is a shame I already made a comment and can't give you mod points.
TFA says:
The move to 32-layer 3D VNAND 3-bit MLC flash brings pricing down to the .50 to .60 per GiB range, but doesn't adversely affect endurance because the cell structure doesn't suffer from the same inherent limitations of planar NAND, since the cells are stacked vertically with the 3D VNAND.
which didn't make sense to me. Luckily Anandtech has a non-gibberish explanation:
Rather than increasing density by shrinking cell size, Samsung's V-NAND takes a few steps back in process technology and instead stacks multiple layers of NAND cells on top of one another. ...In the floating gate MOSFET, electrons are stored on the gate itself - a conductor. Defects in the transistor (e.g. from repeated writes) can cause a short between the gate and channel, depleting any stored charge in the gate. If the gate is no longer able to reliably store a charge, then the cell is bad and can no longer be written to. Ultimately this is what happens when you wear out an SSD.
With V-NAND, Samsung abandons the floating gate MOSFET and instead turns to its own Charge Trap Flash (CTF) design. An individual cell looks quite similar, but charge is stored on an insulating layer instead of a conductor. This seemingly small change comes with a bunch of benefits, including higher endurance and a reduction in overall cell size. That's just part of the story though.
V-NAND takes this CTF architecture, and reorganizes it into a non-planar design. The insulator surrounds the channel, and the control gate surrounds it. The 3D/non-planar design increases the physical area that can hold a charge, which in turn improves performance and endurance.
The final piece of the V-NAND puzzle is to stack multiple layers of these 3D CTF NAND cells. Since Samsung is building density vertically, there's not as much pressure to shrink transistor sizes. With relaxed planar space constraints, Samsung turned to an older manufacturing process (30nm class, so somewhere between 30 and 39nm) as the basis of V-NAND.
By going with an older process, Samsung inherently benefits from higher endurance and interference between cells is less of an issue. Combine those benefits with the inherent endurance advantages of CTF and you end up with a very reliable solution. Whereas present day 19/20nm 2-bit-per-cell MLC NAND is good for around 3000 program/erase cycles, Samsung's 30nm-class V-NAND could withstand over 10x that (35K p/e cycles).
The question is, were they already at 1% before they started stacking 32 on top of each other, because the drive firmware sure as hell can't cover up for 32%* of the drive being screwed.
If 1% of each of the 32 layers is broken, 1% of the drive is broken, not 32%.
That explains why I was able to get a sweet deal on the 840 EVO 500GB for $219 and Micro Center. I ain't complaining, I'm thrilled in fact =).
Life is not for the lazy.
Many laptops don't have space for two SSDs, and even with a desktop, it's cumbersome to have two drives, instead of everything on one.
And never mind 2TB, 3D SSDs should manage at least 10TB right?
Why OpalCalc is the best Windows calc
Technically 10TB is probably possible but if you do the math, do you really want a $4-5000 SSD in your laptop? Or let me rephrase it, you want it but would you pay the price of a cheap used car before you even get to the rest of the machine? You know you wouldn't. And to be honest, I think if you're the kind of user where 1TB is not enough then 2TB is probably also not enough. If we could see a sales breakdown I'm guessing even 1TB is rare.
Live today, because you never know what tomorrow brings
The lease expired on my work laptop, and the new one has a 256GB SSD instead of the 320GB spinning disk the previous one had. It's not enough :-) Specifically, it's not enough to keep my ~60GB of music on, along with the actual work stuff, so that's temporarily off-loaded to an external drive, plus I had to off-load a lot more stuff for the "move almost all your stuff to the new machine" software to have working space.
And unfortunately, the IT department won't let me crack it open and add an extra spinning disk inside it. The state of the art in SD memory cards seems to be that 64GB cards are cheap, but 128GB cards are really expensive, so I'll probably wait six months for 128GB cards to get cheap and install one. 128GB USB3 flash sticks are getting to be cheap, but I can't leave one of them plugged in all the time.
Bill Stewart
New Fast-Compression-only CPR http://preview.tinyurl.com/dy575ks
Where did you find cheap SDxC cards for 128-256GB? When I looked online a month or so ago (plus in Fry's today), they were reasonable up to 64GB, then expensive above that (except for no-name Chinese brands on Amazon that had reviews saying the capacities were fake.)
For USB2/USB3 flash sticks, they seem to be cheap up to 128GB, but with most laptop designs, that's going to stick out of the case, so I'd prefer SDxC cards that can stay installed, as long as I'm not using them for high-speed applications. (If I really believed that ReadyBoost accomplished anything, I'd be tempted to get a 16GB USB3 stick just for that, but I assume that makes a lot more difference on a spinning-disk machine.)
The cheapest ones at Fry's today were $40-45 for either 64GB SDxC or 128GB USB sticks. Since I've got just about 60GB of music I had to offload from my work laptop (new one had SSD that's smaller than the old hard drive), 64GB isn't quite enough so I'll wait around for Moore's Law to catch up.
Bill Stewart
New Fast-Compression-only CPR http://preview.tinyurl.com/dy575ks
They're not stacking silicon wafers on top of each other. Rather, they're putting more layers of oxide, semiconductor, etc onto each wafer in order to produce the 3d stacking. Yes, it's more complex. But it's a pretty mature technology.
They're actually doing both. The 120GB drive only has a single NAND package (the PCB is tiny!) which contains 8 stacked 3D NAND dies. The NAND package has eight chip select pins, one for each die, and the controller interleaves requests to each die to achieve parallelism.
You're young. Early PCs cost 4-5k. Individual hard drives were in the $1000 range back in the 80s.
For someone who absolutely needs 10TB of zero-wait storage in a 2.5" form factor, 4-5k is not a big deal. Because pretty soon it will be $2000, then $1000, then $500.
Inexpensive enterprise SSD is having a big impact on how you spec out servers now. Do you build something with a bunch of 15k RPM drives in a RAID 0+1 array, short-stroked and end up with about 1TB of useful space? Or do you simply put 2x1TB in a RAID-1 array in a much smaller unit?
I paid about $650 per drive last week for 1TB enterprise quality SSDs. I expect them to be below $400 by this time next year. By 2016, I suspect you will not be able to buy a 15k RPM SAS drive as the enterprise SSDs are crushing them from above on price/performance.
Wolde you bothe eate your cake, and have your cake?