Cherryville - SSD 520

[Ao1]

http://vr-zone.com/articles/intel-ss...510/13646.html

The SSD 520 "Cherryville" family is slated for this November, the 4th of November to be precise. That is when one can expect market-availability. The second important bit of information is that SSD 520 is poised to displace the SSD 510 "Elmcrest" series, that's currently out in the market, available at 120 GB and 250 GB capacities.

SSD 520 series, according to fresh information, is built around the idea of providing more capacity choices to the consumer, higher throughput, and lower latencies, compared to SSD 510 series. The new SSD 520 series will be available in capacities as high as 480 GB and as low as 60 GB; provide trasfer rates of 520 MB/s sequential read, 500 MB/s sequential writes (compared to 500/315 MB/s on SSD 510), and 40,000 IOPS read/write 4K random access throughput.

In related news, we hear the codename of the successor to SSD 520 "Cherryville" is codenamed "King Crest", and is slated for the second quarter of 2012.

No mention of the controller yet.

[Computurd]

jeez that is impressive write speed

gah wonder at the controller...is it gonna be an SF?

[Wendy]

gah wonder at the controller...is it gonna be an SF?

£50 says it is.

[bluestang]

Please, oh please, let it be Intel has designed their own controller!

[johnw]

£50 says it is.

No bet, unless you give me at least 4 to 1.

[lowfat]

If they go w/ a third party controller again then I will be severely disappointed.

[squick3n]

Aren't those write speeds sort of a giveaway that it's going to be SF?

[johnw]

Aren't those write speeds sort of a giveaway that it's going to be SF?

It is not proof of Sandforce, but 500 MB/s write speed makes it extremely likely that they are doing compression. In which case, between Sandforce, or Intel's own new controller with compression, I think Sandforce is more likely. But it still could be something else.

[Ao1]

40,000 IOPS read/write seems low for SF, but maybe Intel rate their performance specs differently. (i.e. used state, sustained performance and something other than zeros).

It’s looking unlikely that it will be an Intel based controller, so I guess that only leaves Marvell or SF. Hopefully it’s not the later.

[Wendy]

The suggested capacities also point towards SandForce.

[Ao1]

OCZ may as well pack their bags if it is SF, unless Indilinx can pull something out the hat quickly.

[Anvil]

One can only hope that they have developed their own controller.
Their 1st gen controller has lasted for 3 years or so, should be a healthy investment for them developing a new one.

I'm not sure that Intel would base their specs on compressible data although the 500MB/s write points in that direction.
If it was to be the SF2 controller I don't know what to think, based on the ongoing issues I'd expect a delay, if they get the issues straightened out I'd surely order some. (I'll do that anyways )

[bluestang]

Indilinx Everest?

[Ao1]

According to Fudzilla the King Crest successor to Cherryville, will use 20nm NAND. It looks like Cherryville will only be around for 6 or so months.

[Gilgamesh]

According to Fudzilla, Intel will buy Sandforce


I hope for an Intel controller 6Gbs that write 500MBs without compression

[squick3n]

"So we don't believe that Intel will acquire this controller making manufacturer and that just as everyone else, it plans to use the controllers and make its drives much faster."

I think the article just says Intel "might" buy them, which seems like an odd non-story for a website to run

[johnw]

The suggested capacities also point towards SandForce.

No, it points towards redundant flash with an 8 (or 16) channel controller. That does not necessarily mean Sandforce. Intel already has a controller that uses redundant flash, although it uses 10 (or 5) channels in the 320.

[johnw]

I'm not sure that Intel would base their specs on compressible data although the 500MB/s write points in that direction.

The fastest incompressible data writes on an SSD are on the Samsung 830, at about 400MB/s. That is 8 channels. We know the 520 will use 25nm IMFT flash, though, not toggle flash. I think that toggle flash is a little faster for writes than 25nm IMFT flash. But if IMFT flash actually were as fast as toggle flash, then 10 channels might be enough to hit 500MB/s (10/8 * 400 = 500). Or if IMFT 25nm flash is slower, I guess 16 channels (plus interleaving) might be enough to hit 500MB/s, but only on the highest capacity SSDs since it would take a lot of flash die to fully utilize 16 channels plus interleaving. Probably 64 to 80 die.

[CedricFP]

What do higher sequentials do for the average consumer (ie, no video editing, not many large files, etc.)?

[Ao1]

I wonder if the Cherryville will be ONFI 3.0, which supports up to 400 MB/s compared to 133 MB/s for ONFI 2.0.

Right now I can’t imagine SF are in a good bargaining position. For sure Intel could not afford to release a product with more bugs than a bait store, so unless SF gave Intel complete access to their technology to enable them to develop their own bespoke proprietary f/w I just can’t see it happening.

I just don’t get why Intel need to use 3rd party controllers, when they employ world renowned experts. I’m sceptical they will use SF. My guess is Marvell.

[Ao1]

Looks like Intel have been busy behind the scenes:

http://vr-zone.com/articles/intel-ch...012/13708.html

[Anvil]

Nice to have something to look forward to

I wonder how long it will take until we know a bit more on the Cherryville controller.

[Christopher]

I'm still hoping for a November 4th release, whatever the controller.

[Computurd]

yes the cherryville is going to change things quite a bit. the landscape will change a bit for sure.

[squick3n]

Do you know something, or is that a guess? I kind of thought the Samsung 830 was supposed to change things as well?