Samsung 64GB DDR 2.0 NAND Flash in Production!

Ohhhhh Baby! The next year with SSD is going to be very very interesting. I believe we'll see prices continue to drop to a $1.50-.60ish for our Crucial's and the premiums, Intel and Vertex 3's, dropping to $1.80ish per gb.

The most exciting part is the announcement of the new NAND flash that Samsung is unveiling. This may help me understand why they were not debuting a SATA III SSD. Me thinks we'll be seeing a Vertex 3 killing in the next 6 months.

Samsung's 64Gb toggle DDR 2.0 NAND flash memory with 400Mbps transfer rate hits production -- Engadget

discuss

 

Nothing to discuss, when I test these personally I will have something to add.

But Vertex 3 killing? They've been doing that to themselves for a while now.

 

nothing to discuss? he forgot the 400Mbps speed per chip.

this would result, f.e. on an intel g2 controller with 10 chips in an 80gb ssd with 500MB/s max speeds.

so those are quite nice chips indeed.

 

davepermen,

yeah, the 'theory' is always so rosy, huh?

 

Sounds impressive but how does this work? The DDR1.0 32GB NAND have 133 Mbps but a NAND DDR2.0 with the double size have 3x the speed. Does this mean that a 64 GB SSD with the "old" NAND had 266 Mbps and this new 2.0 chips get 400 Mbps? Or a 256 GB SSD with 1.0 NAND have 1064 Mbps (8x) and the same SSD with 2.0 NAND have 1600 Mbps (4x)?

 

A little more complicated than that: it depends on the controller (how many channels) it depends on how many chips are stacked (per package) and it depends on the capacity of the SSD in question (and somewhat affects how much spare area is allocated).

So, with a ~256GB SSD with 10 channels we have just saturated the brand new SATA3 interface (by a lot).

Bring on LightPeak!

 

What does 10 channels translate to? 10 NANDs? And will you get less NAND chips on a 256GB since the 2.0 is twice the size? Or can you use small capacity chips but with the same technology? I am asking because dave talked about 10 x 8GB NANDs on a 80GB SSD.

 

Each 64Gb nand chip is 8GB's. Each also has a throughput of (400/8) 50MB/s maximum (theoretical).

Each nand chip can be stacked and give you added throughput per stack.

Each chip/stack can be attached to each available channel (dictated by the controller used).

So, if we can put 4 nand chips per stack (at 32GB size) and use 10 channels for 320GB total size we are approaching/surpassing PCI-E speeds (depending on how many lanes we use/need) - not the far humbler SATA3 we can utilize now with our SSD's.

The point is: small capacity chips with the same (new) tech will never be made (they'll cost the same, essentially) and yes, you will get less nand chips on a given sized SSD.

The performance difference in SSD's today is how many stacked chips they use and how many channels they have available. Like I've found out the hard way - small SSD's are simply headaches waiting to happen - 600GB's and up are when SSD's will become a real threat to current HDD's.

See:
AnandTech - OCZ Vertex 3 (240GB) Review


So, with at least 8 channels, multiple nand 'devices' per package and the benefits of interleaving, we should be able to easily hit double what SATA3 is spec'd for right now.

In a theoretical world.

 

Aha ok thanks for the explanation. I thought that if you devided the 400Mbps between 8 x 8GB chips you would get 50Mbps for each or around 5-6MB/s per chip.

 

No, each chip gets 50MB/s by itself - when interleaved on a single package/device it should get ~66% faster - so, about 83MB/s. Now, consider stacking two of these (we are at 32GB now for this 'channel'), and we should be hitting close to 139MB/s if the interleaving/stacking scale perfectly. Now, multiply that by the 8 or 10 channels SF or Intel offers and we are easily double what SATA3 can provide.

I'm not an authority on this - just taking what Anand stated and interpolating it to the new DDR2.0 nand spec's.

(Yeah; 'interpolating' is just a guess. ).

 

now i'm curious how the actual controllers can scale with faster flash? they could just clock them higher

anyways, it's a nice speedbump, and it will result in higher performance at some point. so i'm happy. progress

other than that, yeah, the bigger the ssd, the less anyone has to care about anything. the current offerings (of all sizes) are great for geeks, people that care about the products. for a full hdd replacement, the bigger ones will be awesome. then again, i don't want a full replacement anymore. i'm very glad ssds forced me to move away from local storage, anyways. made life much more simple.

and in two years, if they're still on schedule (according to an about one month old report, they are), memristors from hp will walk in and stomp down everything
then, even tiller will be happy

 

davepermen, you have a report that could make me happy and you don't even provide a link.

I thought we were friends?


Btw, yeah - they could clock them faster (didn't OCZ do that in late/early 2009/2010, but they don't really want to (simply uses more power which is 'not' what SSD's are supposed to be about - for just a little more performance).

 

na, you're a liar => you can't be a friend



oh they don't care about power consumtion for the highest end per se. those are for performance-systems, not for the batterylifehungry ones anyways. well, as a first step, that is ment. over time, there would be new controllers and they would be fast AND efficient.

but i could definitely see an ocz vertex 3 high speed edition coming out soon-ish, if those samsung chips work with sandforce. and it would be a nice ssd for those that want the highest performing system (overclocking the cpu and all, multi gpu setups).

you don't know memristors yet? or what?

 

Okay, I'm a liar (lol...) - and yeah, I do know about memristors - but what about a link to their latest predictions of when they'll take over the world?

 

yeah, you still spread fud about hdds faster than ssds

hm.. can't find it in my history currently. most likely it was on a different system, and it was about am month ago. message was more or less "we're on track, 2013 still target". can't find it in google reader, so it wasn't news per se.

no further info, sadly.. i'd love to be betatester of a memristor drive and memristor ram (so standby uses 0 power).

i know just one thing. sata3 is stupid. and it'll be a limit very soon. memristor or not

 

But SSD are slower than HDD's for me (again: not in 'snap', but 'productivity'). No lies (wait for more details after this weekend).

Now, I am calling you a liar!

From the March 1, 2011 blog here:


See:
HP Blogs - HP nanotechnology research looks to sustain HP ser... - The HP Blog Hub


So, I un-friend you too.

 

well, i've read news. i have not verified them so i'm not a liar, per se.


oh can't wait for after that weekend. have you failed another ssd test? seriously, how dare you to fail, again?

 

Haven't failed yet - and it's not me that fails - it's these wimpy SSD's that can't stand to do real (storage) work.

In late 2009 eBoostr was giving me 'snap' - if that is all SSD's are about (still) then we'll have to wait for your imaginary HP memristor 'tech' to revamp the world as we know it.

 

so what single hdd brings you > 150MB/s read speed? or 0.065ms latency? or same write? etc..

what crap ssd do you test with?

 

Who cares about 'specs'? And lying benchmarks?

My tests are a little more real world than that. And get 'scored' with a stopwatch.

And the crap SSD's I'll test with are the same crap you are enjoying too - Intel.

 

well, i have 13 ssds by now, most of them in use. i know about both specs and reality.

and the reality is that i get exactly the benchmark-supported numbers: everything much faster than on a hdd, within the spec-number limits.

i can even take those specs, compute on paper what the differences will be, do it on hdd and ssd, and the outcome is exactly as computed.

so specs work for me, they work very well. they are more or less what real world is. if not, then it's a faulty disk.

so, if you, again, fail, then you seriously have troubles. massive ones except you've chosen a drive tha does, by spec, not allow to outperform the hdd in exactly the point you want (say the one i'm on right now, it has 100MB/s write speed at max. you can beat that with a hdd over parts of the hdd)

 

I knew that Samsung was going to make Sata III SSD's , It appears in the foregoing link is the nand flash chip they are going to use,,no doubt other SSD makers will be buying these chips as well. Samsung is upping the ante,,,lol ,,

"Get ready for it speed freaks, Samsung's toggle Double Data Rate 2.0 MLC NAND chips are now in production -- an industry first. The 64 gigabit flash chips manufactured using 20nm processes boast an impressive 400Mbps transfer rate. That makes these toggle DDR 2.0 chips about three times faster than toggle DDR 1.0 (a 133Mbps interface) or ten times faster than the 40Mbps SDR NAND flash in widespread use today. Look for 'em in future teardowns of tablets, SSDs, and smartphones"

http://www.engadget.com/2011/05/12/samsungs-64gb-toggle-ddr-2-0-nand-flash-memory-with-400mbps-tra/

Nice find and post permen.

Samsung is diversified,,they make Laptops,SSD's ,TV's,HDD's,phones,DVD players,ect, 24 Product lines--with all kinds of models in these.

http://www.samsung.com/us/#

No doubt these chips have a future in samsung products where they can be used,Yahoo,Yippie,,,!!!!

Cheers
3Fees