I'd love to see results of the SATA SSD's on various RAID controllers, such as a 3Ware Card ( http://www.3ware.com/products/index.asp) . Can the slowness of the SSD in some tests be made up in striped RAID 0 configuration?

I wonder if maybe taking two 16GB SSD drives and striping them on one RAID 0 Card (for the OS), and then taking three WD Raptors and striping them on another RAID 0 Card would give the OS and data the best performance?

If you get a good RAID card that does all onboard processing, there should be little offloaded to the mother board CPU, and with PCI-e or PCI-X there should be plently of bandwidth on the bus right?

Hmmmm? Of couse you still have to make sure there is a place to put the graphics card.

Even the 16GB SSD drives look pricey. While we are talking fast configs (and pricey), maybe the fastest rig would be with two 16GB SSD drives striped (OS) and then some SAS drives striped (for installed apps and data\games) using a unified serial controller: http://www.3ware.com/products/serial_ata2-9690sa.asp

Why do my dreams always cost so much. I thought dreams are supposed to be free.

I'm not certain I'm reading the same article, based on some of the overly positive replies that I've gone through in this comments thread.

From the tests that you've run, this SSD is only marginally better than the higher-quality mechanical drives, and in many tests that you show results for, it's worse (or much worse.)

Why on earth would there be any interest (even from the gamer/hard-core enthusiast) for this technology given how poorly it performs (at the present) against run-of-the-mill hard disk drives?

Certainly there might be niche applications (low power consumption is the only one that really sticks out to me) that can benefit from this technology, but I don't see general-purpose applications gaining such a beneft.

My theory is that the web server test is returning many small requests. I doubt this trend would continue if the web server test was vs. larger files served...

A further analysis should include a "file-size" dimension.

Further, I now wonder about the database results. If you're doing a lot of very quick, lookups reading small data sets, or performing data locking in transactions, I think the solid state might work well. If you're querying for large datasets, I doubt it.

I wonder about small database transactional writes... vs. large transactional writes.

ps: Also might it be better to deck out your webserver with ram and cache everything? With new systems, 128GB of RAM is actually possible..

What do y'all think?

At the IDF, Intel is already touting an SSD with twice the write performance of existing SSDs (though actually, if you're paying attention, that line above it about "10X-50X IOPS" is more impressive).

http://www.hothardware.com/articleimages/item1027/big_gelsinger_22.jpg

You know even windose will create and use RAMdisks.

Set up one with 2G or so and see how fast that puppy is.

I'm glad they're pushing up capacity, but price isn't quite in the right place just yet.

Does anybody know what the mean time to failure of this drive is (or SSD drives in general)?

It would be interesting to see some compile times with the flash drive vs standard drives. The random access time should help but the write speeds might counter some of that. However when running multiple builds on a drive it could make a difference vs a standard drive.

Somebody buy me one, please?

I'll name my nextborn child after you.

Hmmm... definitely better than some past SSD, but even ignoring the price it seems mostly interesting for some specific needs.
What struck me more was just how well the WD 750GB HD did in most of the tests.
edit: great review!

Hmm, I see an idea here that may make this drive work a lot better...

How about a DRAM write cache? I know it takes up more space, and increases power use, but it might be worth the tradeoff - you can write to DRAM a hell of a lot faster than you can write to flash. So, in a workstation workload, that should help a lot (because it can write to flash during periods of idle...) even an 8 MiB cache would help. Remember, it keeps up with the pack in read access, even without a read cache.

As far as thickness issues go, wouldn't a stacked BGA architecture solve this problem without all the problems one would normally encounter with stacked TSOP technology?

Somebody get it out of the way to ask whether there are limited writes on SSDs, and then ubergerbil will tech pwn ur *ss.

APART FROM COOLING FANS, hard drives are the only mechanical devices inside a modern PC

Lets not forget optical drives :p

i would totally buy this if i could find a 95% off coupon for it.

lost my interest after the first benchmark where it wasnt in first place.
i can put up with slightly slower load times for games & system boot when in most other normal situations a $100 hard drive is faster than this thing.
i dont see the point at all.

Price will drop 50% each year. Capacity will climb, though I don't know how necessary that is now that we're over 100GB -- archival and media storage can remain on spinning disks. Speed should improve as well, though they're going to have to find ways to jam more chips into the same space. Assuming they can do that, doubling the transfer speed next year is trivial, quadrupling the year after that. Try that with HD tech.

You know, there really shouldn't be much power used by this at idle. The NAND by itself isn't using any, so there certainly shouldn't be such a difference vs the 8GB IDE SSD simply due to the size. It looks to me like the SATA interface by itself uses a watt. I wonder if the entire stack -- motherboard, BIOS, SATA controller, OS -- fully implemented the latest ACPI (3b) and AHCI (1.2) specs, if the SATA interface could be put to sleep and then woken up (looking much like hot-plugging the SATA controller, I would guess) to eliminate that power drain. I wouldn't be surprised if Windows wouldn't be too happy with having the interface for the system drive going away periodically, however.

Still, 1W is a lot in a mobile application.

what i like most about TR reviews is that they understand that the enthusiast is actually riding a fine line between server tech and desktop tech. for us to REALLY be ahead of the curve, we need to be aware of hardware like this.

Thanks for the review!

I'd like to see this SSD tested with an NVIDIA or AMD based motherboard. At least some Intel chipsets have performance issues with SSDs, as demonstrated by AnandTech's review.

I want to live in a world where Macbooks come stock with a hybrid 2.5" drive: 4GB NAND w/ 1 80GB platter. OS X would treat the NAND portion as the "fast sector" of the hard drive, and migrate commonly read files there, while intelligently copying or writing large files onto the platter.

Maybe someday, after robotic beings rule the world.

No review of the Mtron yet? That's the only really impressive flash drive out there at the moment.

http://www.anandtech.com/storage/showdoc.aspx?i=3064