Marvell's 3rd Generation 88SS9187 SSD Controller
by Anand Lal Shimpi on March 14, 2012 12:23 PM EST
Today Marvell is announcing its third generation SSD controller with the 88SS9187. Its predecessor, the 88SS9174 was used in many drives including Crucial's m4 and Intel's SSD 510. The 9187 appears to be mostly a feature update, adding support for the following:
- SATA 3.1
- On-chip RAID (allowing for failure of defective NAND block, plane, die or device)
- DDR3 memory interface (up to 1GB external DRAM supported)
- ~500MB/s sequential write speed
- 6Gbps SATA limited sequential read speed
- "best-in-class" random read/write speed
Details are scarce but it sounds like the 9187 is simply a higher performance (potentially lower power?) version of its predecessor. Marvell is expecting the 9187 to be used by manufacturers in Ultrabooks among other designs this year. Marvell's SSD controllers have always been priced competitively, but they've often required a lot of custom firmware work to get the best performance/compatibility/reliability out of them. I see no indication that Marvell's business model has changed, so you can expect to see the 9187 used similarly to its predecessor.
Facebook
Twitter
RSS
24 Comments
View All Comments
iwod - Thursday, March 15, 2012 - link
Interesting, but isn't Apple already the world largest NAND buyer? Why buy another package when they could easily build them themselves? I mean it is not like HDD, SSD are simple controller and NAND soldered together.And as we have already seen Apple do make their own SSD Firmware, so they will have to work on their own SSD firmware regardless of which manufacturer.
And honestly, why another samsung parts in Apple when there are many great alternatives.
Kristian Vättö - Thursday, March 15, 2012 - link
Apple is definitely a big buyer of NAND but I'm not sure if they are the biggest. Building an SSD is relatively easy but doing it well is not. I think there are enough of these bad SSD stories showing that it's not as easy as one would think. Especially reliability is hard to achieve and that's what matters to Apple.Apple's purchase of Anobit may eventually lead to an Apple-designed controller but right now we don't know. Validation alone can easily take a year, and design can take even more so it will take a while before we know what Apple has been planning.
Apple uses custom firmware in their drives but we don't know who wrote it. Apple could have paid for Samsung/Toshiba to customize the firmware to suit their needs. Or it could be the basic FW with a few Apple tweaks. It's totally different to make a few tweaks to a very mature FW compared to almost building the FW from a scratch.
Apple buys lots of stuff from Samsung (and Toshiba) so they are logical choices. Creating a new relationship always has its risks and I bet Samsung offers fairly competitive pricing for Apple, plus their SSDs have proven to be very reliable.
alan1476 - Thursday, March 15, 2012 - link
Now maybe Marvell will turn its attention to the crappy SATA ports on the LGA 1155 and LGA 2011 not to mention the embarrassing LGA 1366 cards. You cannot use an SSD on them becuase it wont run at the rated speeds, so you are left with 2, yes only 2 Native Intel SATA ports that work correctly with SSDs.magnetar - Thursday, March 22, 2012 - link
I agree 1000%, the so-called "Marvell 91xx SATA III" chipsets, that can only be used with one PCI-E lane, so are limited to 5Gb/s, instead of the 6Gb/s of true SATA III interfaces like those found on most Intel 6-series chipsets.I cannot understand how Marvell, and mother board manufactures, have gotten away with calling these SATA III speed capable interfaces. SSD and mother board manufactures forums are filled with questions about why their owners SATA III capable SSDs are not operating at their full speed. I cannot think of one mother board review anywhere that has even mentioned it, much less anyone calling Marvell out on this, although those of us out in the "wild" are very familiar with it.
Is any SATA interface that surpasses the SATA II speed limit by one MB/s, qualified to be called SATA III? Or do the magic words "up to..." make it alright?
Marvell does freely state the 91xx chipsets operate on one PCI-E lane, which I have seen when their Option ROM runs on my mother board, and displays "... Marvel 91xx... 5Gb/s...". Or read their product overview document on their website for these chipsets. Can someone tell me what I am missing here?
Marvell does have their 88SE92xx series SATA III chipsets available now, apparently, that can be connected to two PCI-E lanes. They also have Marvell's SSD caching feature called HyperDuo, combining a SSD and HDD. Two of the chips have four "SATA III" connections, so does that mean we'll get twice as many under-performing SATA III ports?
While I have two SSDs with Marvell SSD controllers (Crucial M4's, Intel 510) which IMO are great products, I won't take Marvells' word about their SATA III interface chip's performance until I see multiple tests that verify they perform as they should. Hopefully someday a PC hardware review web site will have the nerve to print the truth about these SATA chipsets.