Hacker News new | past | comments | ask | show | jobs | submit login
Toshiba Introduces MacBook Air 'Blade-Type' SSDs to Mass Market (macrumors.com)
34 points by there on Nov 8, 2010 | hide | past | favorite | 18 comments



That is awesome. It is an utter waste of space to put SSD's into a 3.5 or 2.5" hard drive case.

This also opens up the door for some very compact PCs. Rather than having flash memory soldered onto the motherboard, I would like to see slots on regular PC motherboards for this.

Lots of questions: Is this some standardized connector, or has Toshiba created a defacto standard? Is this just an alternate SATA plug, or is the SATA daughter controller integrated onto the motherboard itself?

Edit: It's just standard mSATA. Controller on each drive. From the horse's mouth: http://www.toshiba.co.jp/about/press/2010_11/pr0801.htm?from...

mSATA is a standard: http://gizmodo.com/5364485/msata-its-like-sata-but-smaller


Putting an SSD in to a 3.5 or 2.5" hard drive enclosure is indeed a waste of space, but talking 'mSATA' to what is essentially memory seems to be a waste of speed.

All that work to stuff data down a serial pipe emulating a harddisk that isn't even there.


Flash has certain disadvantages that favor using a controller instead of direct addressing like DRAM. To make SSDs fast, you have to do wear leveling, command queueing, and parallelization of requests across all flash chips. Oh and when a cell fails, you have to remap spare cells. Doing this stuff in software would be a giant pain.

You can get raw flash chips, but it's hard to get decent performance and reliability out of them. That sort of setup is usually for embedded systems where writing is rare and read performance isn't important. Firmwares for switches and routers, for example.


While I understand your comments and that this can get hairy, also consider a softmodem-like implementation as a counter-example.

A softmodem (variously also called a Winmodem) is a partial or full software implementation of a modem, and helped reduce the parts counts and the costs of modems.

Ugly or not, it's a near certainty that we'll see a software (or firmware) implementation of this; some sort of "softssd" or "winssd" scheme.


You'd need custom file systems, etc., plus no PC BIOS will boot from something that doesn't resemble an ATA device. The whole ecosystem is based on layers of legacies, you should know that by now!


now imagine these babies filling in a rack cabinet.


I've been expecting this from the moment I first saw an exploded view of an SSD. I understood the case allowed for the SSD to be used as a replacement drive, but figured it was only a matter of time before they'd be installed in slots on the motherboard. Next prediction: RAM/SSD soldered right onto the motherboard. [Yes, I realize I'm stating the obvious.]


Hate to steal Apple's thunder, but the first eeePC (of all things!) was, I think, the first to do this, and many non-HDD netbooks now use this design. The MBA ones just seem to be longer, allowing more flash chips. Previous examples not only have cribbed the mini PCIe's connector but also the entire form factor (5cm or 7cm long), despite not using the PCIe protocol. Interestingly, the MBA still seems to use the PCIe connector, just the form factor has changed.

So far the various netbook designs are electrically incompatible. I don't know where the MBA one fits into all this.

EDIT: looks like they standardised the interface as "mSATA" back in June.


Hate to steal EeePC's thunder, but this has been done in embedded PCs on mini-itx form factor computers since flash storage was available, pre-sata connectors. It looked like a 1-inch nub that fits in the IDE connector. It was perfect for embedded PCs since power requirements were always very low for flash memory and an easy way to save a watt.


Indeed, but weren't/aren't those typically encased in a plastic housing, not unlike CompactFlash cards?


Those drivers never had the same form factor! even in the same eeepc they used to differ.

my eeepc 1000 has 40GB of SSDs. 8gb in fast SSD drive, and 32 on another, much slower. The 8gb measures some 8cm. the 32gb measures 7cm.

There were also dell mini9 that measured some 10cm if i'm not mistaken.

Those drivers use some non-standard pata (newer asus) or sata (older asus and dells) over a mini pci-e connector.

I have no idea what was standardized as mSATA, may be something completely different, or could be the sata version of those mini-pci-e


Anyone know how the Toshiba controllers on these and the MBAs perform, relative to the Intel and recent Sandforce controllers?


Anand does some performance tests:

http://www.anandtech.com/show/3991/apples-2010-macbook-air-1...

summary: pretty good!


MBA benchmark http://arstechnica.com/apple/guides/2010/10/116-macbook-air-...

Using X-Bench, we saw the 128GB model—the APPLE SSD TS128C—give the following numbers:

Seq Uncached Write: 134MB/sec (4k blocks) Seq Uncached Write: 158MB/sec (256k blocks) Seq Uncached Read: 68.5MB/sec (4k blocks) Seq Uncached Read: 165.8 MB/sec (256k blocks)

Random Uncached Write: 32.18MB/sec (4k) Random Uncached Write: 116.5 (256k) Random Uncached Read: 8.27 (4k) Random Uncached Read: 108.56 (256k)



Apparently, either poorly or pretty good! depending on who you ask, and how you interpret the following review.

http://www.anandtech.com/show/3991/apples-2010-macbook-air-1...

I guess it's a glass half full, glass half empty kind of thing.


I'm wondering when we'll see servers with hot swap versions of these.





Consider applying for YC's Spring batch! Applications are open till Feb 11.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: