SandForce Unveils SF-2000 Series SSD Processors

[Imperceptible]

These chips feature a 6 Gigabit-per-second SATA host interface, industry applauded DuraClass Technology, an unprecedented 60,000 sustained random read/write IOPS (Input-output Operations Per Second) and sustained sequential read/write performance of 500 Megabytes per second.

http://techpowerup.com/132368/SandF...ors-Enabling-Up-to-500-MB-s-over-SATA-6G.html

Exciting times are ahead!

 

[dmoney1980]

thats awesome !

 

[Edrick]

Finally, a good SATA 6 SSD controller.

 

[FishAk]

Here we go! Look out Intel.

60K read and write IOPS? I gota wonder if that's write, I mean, right.

500Mb/s sequential... Wow!

 

[Zap]

What, you guys don't read Anandtech?

 

[aggressor]

One was posted earlier than the other

 

[FishAk]

Thanks, Zap, I hadn't noticed Anand had his article up.

From what Anand says, SF really will have very impressive writes. Just wait till they get hold of some of that 25nm nand.

 

[Mark R]

Here we go! Look out Intel.

60K read and write IOPS? I gota wonder if that's write, I mean, right.

Yowsers. That's pretty fast. In fact, it's 2x the speed of the X25-E.

I wonder if that's with conventional MLC. The presentation says 'cMLC' - which I presume means conventional MLC (as opposed to eMLC) - but I suppose it could be a typo - as I've not seen cMLC in use as an abbreviation anywhere else. The eMLC results should be more interesting - particularly for write performance. Even so, that's a pretty astonishing performance.

What's more is that sandforce quote the lowest bit-error rate of any major manufacturer - and they have reduced it further with this latest controller. Presumably, that's a result of their unique ECC algorithms and internal redundancy "RAISE". It's worth pointing out that to-date, all SSD controllers use generic ECC which assumes random bit errors (or random bit-pair errors for MLC), with the exception of Sandforce and Fusion IO. These two companies have performed their own internal analysis of error patterns in different types of flash memories, and found big difference between manufacturers. As a result, they supply customised firmware, with an ECC algorithms tuned to the precise memory used.

 

[=Wendy=]

Great speeds, if they are realised in the real world.

@Mark R
Could the cMLC mean consumer grade MLC?

 

[Zap]

Yowsers. That's pretty fast. In fact, it's 2x the speed of the X25-E.

How long as the X25-E been out? That's like comparing the future Sandy Bridge to a socket 775 Xeon.

 

[Burner27]

Don't some of the 'special sauce' SF-1222 based drives already achieve 50,000 IOPS? How is 60,000 IOPS much of an improvement for the SF2000 chips?

 

[FishAk]

50,000 IOPS is only on reads.

 

[Voo]

How long as the X25-E been out? That's like comparing the future Sandy Bridge to a socket 775 Xeon.

Well the rumoured values for the Postville refresh don't look that much better.

If SF really leaves Intel behind like those numbers would suggest, that'd be a really formidable task - they have less than 100 engineers and probably less than .1% of Intels budget.

 

[BTRY B 529th FA BN]

What, you guys don't read Anandtech?

And then read the comments at the bottom of the page, those are more interesting.

Me, the top factor in the triangulation of my decision making for an SSD drive these days is RELIABILITY.

Whats the stock market abbreviation for Sandforce anyways?

 

[Mark R]

Whats the stock market abbreviation for Sandforce anyways?

It's not a public company.

It's privately owned, with large investments from 2 venture capital firms (Storm ventures and Doll Capital Managment). I don't think these firms cater to small investors either - they seem to be an introduction service between millionaires with cash to gamble and new businesses - rather like Shark Tank, but on a rather larger scale.

Still, if Sandforce continue their rapid growth then, as Anand hints, they may choose to take the company public.

 

[Zap]

And then read the comments at the bottom of the page, those are more interesting.

Sometimes they are. There's some interesting discussion about current generation Intel versus Sandforce, and how Sandforce drives totally beat the Intel drives in specifications and synthetic benchmarks, but often fall behind in real-world usage. This came about from talk about how the SF-2000 totally whips Intel's announced G3 specs. Of course those are controller specs versus actual drive specs, so the numbers will likely end up being closer.

My thought is that I don't care whether my next SSD is the "fastest" or not. It just has to be reliable and cheap. Gimme an SSD that is no faster than the existing Intel G2 and just as reliable, but make it cost less than half the current $/GB and I'll be putting BIG SSDs into all my computers instead of the current small boot SSD and larger data SSD.

 

[Dark Shroud]

Very impressive over all. Between Sandforce & OCZ's new HSDL connection for SSDs things are getting interesting. I really wish Intel had gone with SATA 6gbs. Who knows maybe Intel will surprise us with a model running on Light Peak.

Just wait till they get hold of some of that 25nm nand.

Is anyone besides Intel even making 25nm nand?

 

[LokutusofBorg]

Is anyone besides Intel even making 25nm nand?

Samsung and Toshiba.

 

[FishAk]

Didn't OCZ show off a 25nm SF drive at IDF?

Even though Intel doesn't actually make any 25nm flash themselves, I think with their muscle, (do they actually own IM Flash Tech?) they could prevent anyone else from releasing 25nm drives before they do.

 

[LokutusofBorg]

The I in IMFT is Intel. They get 49% of the production from the IMFT plant, Micron gets 51%.

Even if Intel refused to sell, SandForce drive-makers could buy from Micron, or Samsung who isn't far behind. I don't know what stage Toshiba is at in their production ramp. "Intel marketing" would have zero effect getting Samsung to refuse to sell, and I doubt they'd even have much leverage in getting Micron to not sell, unless they had some agreement ahead of time.

You can bet these companies have known about each others plans for a lot longer than us.

 

[flamenko]

Didn't OCZ show off a 25nm SF drive at IDF?

Even though Intel doesn't actually make any 25nm flash themselves, I think with their muscle, (do they actually own IM Flash Tech?) they could prevent anyone else from releasing 25nm drives before they do.

Only Sandforce showed working 25nm in present footprint at IDF

 

[adamantinepiggy]

The I in IMFT is Intel. They get 49% of the production from the IMFT plant, Micron gets 51%.

Even if Intel refused to sell, SandForce drive-makers could buy from Micron, or Samsung who isn't far behind. I don't know what stage Toshiba is at in their production ramp. "Intel marketing" would have zero effect getting Samsung to refuse to sell, and I doubt they'd even have much leverage in getting Micron to not sell, unless they had some agreement ahead of time.

You can bet these companies have known about each others plans for a lot longer than us.

Micron is making a good amount more per NAND chip on their own Crucial line of SSD's; versus selling them at market prices. If the SSD side of things take off further, expect to see even more of Micron's NAND production stay in-house.

 

[taltamir]

The claimed speed is 500MB/s. However:

We'll get to the how in a moment, but let's start at the basics. The overall architecture of the SF-2000 remains unchanged from what we have today with the SF-1200/SF-1500 controllers.

SandForce’s controller gets around the inherent problems with writing to NAND by simply writing less. Using real time compression and data deduplication algorithms, the SF controllers store a representation of your data and not the actual data itself.

How could they even come close to getting such speed improvements merely by adding compression? and how bad will it perform with already compressed files?

Also:

Both ONFI 2 and Toggle NAND add another bit to the NAND interface: the DQS signal. The Write Enable signal is still present but it’s now only used for latching commands and addresses, DQS is used for data transfers. Instead of only transferring data when the DQS signal is high, ONFI2 and Toggle NAND support transferring data on both the rising and falling edges of the DQS signal. This should sound a lot like DDR to you, because it is.

I thought it was saying the architecture didn't change, then the article is saying that it has added DDR like capability (which should double the speed... which is inline with the claimed improvements in speed)

 

[Dark Shroud]

Micron is making a good amount more per NAND chip on their own Crucial line of SSD's; versus selling them at market prices. If the SSD side of things take off further, expect to see even more of Micron's NAND production stay in-house.

This is kind of what I see with Intel & Micron producing a lot of volume but using it all themselves simply because of the supply constraints in the market.

Hopefully Sandfoce will drive down Intel's prices. I want a large capicity SSD for my software & games. I can handle not having a top of the line Sandforce performance as long as I'm not paying Sandforce prices.

 

[taltamir]

the huge majority of the SSD price right now is the price of the NAND chips.
the reason why bigger drives are so much more expensive is due to stacking... each "chip" is actually 1 to 16 dies stacked in 3d, typically 2 or 4 die. It gets very expensive to stack beyond 4.