Well, the problem is that we don't is it in-house or third party. The speeds look a lot like SandForce but that doesn't automatically mean that it's SandForce. Here is an article claiming that Hitachi is using in-house controller in their SSD400M lineup (I would be willing to bet that it's the same controller with different NAND and possibly firmware):
Majority of the enterprise products have had in-house tweaked, if not developed controllers.
Not a coincidence that the enterprise products are only pretty much coming from companies that are absolute behemoths in the IT space (Intel, Seagate, Hitachi, etc)..
If there are no difference in SLC and MLC manufacturing, then SLC should really double the price per bit compare to MLC. Assuming we are talking about 2 Bit MLC here.
But the current SLC cost 3 to 4 times as much.........
1. SLC enters production later so most of the time, you will have SLC based on older process node. That means bigger die size or smaller capacity, which means you get less capacity out of a single wafer, making manufacturing more expensive. For example now, all SLC SSDs have been based on 3Xnm NAND, while MLC SSDs are mainly 2Xnm.
2. While the actual manufacturing is the same, the quality is not. Just like in CPU manufacturing, some of the dies are higher quality than the other - after testing, one is sold as i5-2400 while other is sold as i7-2600. This is the same with NAND, not every die is the same. Since SLC is solely for enterprise, it needs to be high quality. It needs to withstand higher temperatures (this affects the P/E cycles) and be good enough for enterprise load. So while SLC and MLC dies can come from the same wafer, I would suspect that the best dies are picked for SLC production. Of course, there are Grade 1, Grade 2 etc. but I think OEMs that use SLC SSDs want the highest quality, not some cheap stuff.
3. SLC based SSDs often deploy enterprise-grade controllers and firmwares. They are more expensive than their mainstream counterparts so that raises the total price of the SSD as well. Also, enterprise-grade SSDs tend to go through stricter validation process to ensure that it's bug free, which isn't free either.
There will also be a TLC (triple-level cell) NAND article out soon, which will talk about NAND manufacturing a bit more :-)
I doubt it. One characteristic of SF controllers is random write being about the same, or even higher, than random read. It's all about the compression. These numbers don't reflect that.
Samsung claim to be the worlds largest NAND manufacturer yet the majority of SSDs all use IMFT NAND. Word on the forums is toggle is actually faster than sync IMFT NAND. Can you shed any light on why its not more of an even split between IMFT and Samsung used in SSDs? Is Samsung NAND a lot more expensive? Is Samsungs order book already full?
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
14 Comments
Back to Article
FaaR - Wednesday, February 8, 2012 - link
...Which controller do Hitachi's stated figures most closely resemble - if any?Maybe it's possible to figure out if they've bought a 3rd-party controller with a little bit of deductive reasoning. :)
MrSpadge - Wednesday, February 8, 2012 - link
The problem is we can only compare to MLC speeds, which are slower than SLC speeds for the same controller.Kristian Vättö - Wednesday, February 8, 2012 - link
Well, the problem is that we don't is it in-house or third party. The speeds look a lot like SandForce but that doesn't automatically mean that it's SandForce. Here is an article claiming that Hitachi is using in-house controller in their SSD400M lineup (I would be willing to bet that it's the same controller with different NAND and possibly firmware):http://www.zdnet.co.uk/news/storage/2011/08/09/hit...
sanguy - Wednesday, February 8, 2012 - link
It's a native SAS 6Gbps drive and is Hitachi's own controller.EMC has been using their prior SLC models for some time with solid reliability and performance.
They also make a MLC model.
You just don't hear about these on the enthusiast sites as not the intended market.
mlambert - Wednesday, February 8, 2012 - link
It is for enterprise class storage arrays (VSP, FAS, etc).DukeN - Wednesday, February 8, 2012 - link
Majority of the enterprise products have had in-house tweaked, if not developed controllers.Not a coincidence that the enterprise products are only pretty much coming from companies that are absolute behemoths in the IT space (Intel, Seagate, Hitachi, etc)..
iwod - Wednesday, February 8, 2012 - link
If there are no difference in SLC and MLC manufacturing, then SLC should really double the price per bit compare to MLC. Assuming we are talking about 2 Bit MLC here.But the current SLC cost 3 to 4 times as much.........
Kristian Vättö - Thursday, February 9, 2012 - link
I see several reasons for that:1. SLC enters production later so most of the time, you will have SLC based on older process node. That means bigger die size or smaller capacity, which means you get less capacity out of a single wafer, making manufacturing more expensive. For example now, all SLC SSDs have been based on 3Xnm NAND, while MLC SSDs are mainly 2Xnm.
2. While the actual manufacturing is the same, the quality is not. Just like in CPU manufacturing, some of the dies are higher quality than the other - after testing, one is sold as i5-2400 while other is sold as i7-2600. This is the same with NAND, not every die is the same. Since SLC is solely for enterprise, it needs to be high quality. It needs to withstand higher temperatures (this affects the P/E cycles) and be good enough for enterprise load. So while SLC and MLC dies can come from the same wafer, I would suspect that the best dies are picked for SLC production. Of course, there are Grade 1, Grade 2 etc. but I think OEMs that use SLC SSDs want the highest quality, not some cheap stuff.
3. SLC based SSDs often deploy enterprise-grade controllers and firmwares. They are more expensive than their mainstream counterparts so that raises the total price of the SSD as well. Also, enterprise-grade SSDs tend to go through stricter validation process to ensure that it's bug free, which isn't free either.
There will also be a TLC (triple-level cell) NAND article out soon, which will talk about NAND manufacturing a bit more :-)
semo - Thursday, February 9, 2012 - link
2 bits per cell quadruples the effective storage capacity.semo - Thursday, February 9, 2012 - link
Brain fart. Ignore the above commentFunBunny2 - Thursday, February 9, 2012 - link
I doubt it. One characteristic of SF controllers is random write being about the same, or even higher, than random read. It's all about the compression. These numbers don't reflect that.Coup27 - Saturday, February 11, 2012 - link
Samsung claim to be the worlds largest NAND manufacturer yet the majority of SSDs all use IMFT NAND. Word on the forums is toggle is actually faster than sync IMFT NAND. Can you shed any light on why its not more of an even split between IMFT and Samsung used in SSDs? Is Samsung NAND a lot more expensive? Is Samsungs order book already full?helpful - Monday, February 13, 2012 - link
in a word APPLE inc.hingfingg - Thursday, February 16, 2012 - link
** {{w w w }} {{proxy4biz }} {{ com}} *****