SanDisk Focusing More On Desktop and Mobile SATA SSDs, Extreme II Series Tested 71
MojoKid writes "Odds are, if you've purchased anything that uses Flash memory in the last 20 years or so, you already own a piece of SanDisk technology. The company has been in Flash storage since the late '80s and manufactures products used in everything from smartphones to digital cameras. Even though it enjoys a long history in the Flash memory business, SanDisk is perhaps not as well known for its Solid State Drive (SSD) solutions for desktop and mobile PCs. However, SanDisk recently expanded their product stack with new, high-performance SSDs that leverage the company's own NAND Flash memory and Marvell's popular 88SSS9187 controller. The new drives are SanDisk's Extreme II family of SSDs targeted performance enthusiasts, workstations professionals and gamers. The initial line-up of drives consists of 120GB, 240GB, and 480GB models. Performance specifications for the three drives come in at 545MB/s – 550MB/s for reads with write performance from 340MB/s to 510MB/s, depending on density. In the benchmarks, SanDisk's Extreme II SSD showed it has the chops to hang with some of the fastest drives on the market from Samsung, Corsair and OCZ."
you've got to be kidding me (Score:5, Insightful)
This blurb could not possibly have been written by a regular human interested in technology, unless there is a SanDisk fanclub I was previously unaware of.
Re:you've got to be kidding me (Score:5, Insightful)
As somebody closely following the development of SSD technology (that we use for our database servers) I would have to respectfully disagree.
You see, our results testing SSDs against PostgreSQL 9.1 showed that SSDs improved performance by at least 90%. In other words, queries, particularly the large, nasty, 10-table joins with combined inner, outter, and meta-table joins that our vertical application is rife with, take 10% or less time to run. That result isn't just dramatic, it's a game-changer. But the truth is that even that isn't enough. Being able to saturate a 6 Gbps SATA III link in a random access read-load is fine and dandy, but write performance is also a very big deal, especially since our system is highly transactional and transaction wait states are painful.
In short, unlike CPUs, SSD technology is still immature enough that every bit of good news counts quite a bit.
Re: (Score:2)
Personally, I'm more interested in durability. And I'm very suspicious that planned obsolescence plays a big part in this market.
Re: (Score:2)
The part that bugged me is the idea of fixing an obviously badly broken application by just throwing more hardware at it. Although doing this with hardware of dubious reliability does seem like a recipe for disaster.
Re: (Score:2)
I must ask, do you consider research into quantum computing to just be 'throwing more hardware' at NP-hard computing problems?
That's a completely stupid analogy. NP-hardness is an ultimate limit (for now, at least) but there's probably lots of sloppily coded databases which can enjoy great performance increases today by just someone doing sane software optimization.
Re: (Score:2)
Using a JOIN is not a sign of a badly broken application; on the contrary, it generally indicates you have well-normalized data. We've shied away from using them in the past because they're inherently reliant on data scattered across the disk, and until SSDs came around accessing the non-sequential data proved too expensive in many cases.
This exactly the kind of problem that you *should* throw hardware at, because this kind of hardware (storage with near-zero latency) is the right tool for the job. We've ju
Re: (Score:2)
It's like people are afraid of using a tool (SQL) to do what it does best: join relevant slices of data from a large set of data! (generally called a "database")
When did it become preferred to use SQL more like a NoSQL like file store? No wonder some people are claiming that NoSQL is "better"!
We use ORM for editing data, EG: CRUD. But for the types of complex reports that our customers ask for, there's no higher god than a good, well-written SQL query. Even though some queries take a while (5 seconds isn't
Re: (Score:2)
I'm confused like you, if you're going to use SQL, use it properly. If you want a big flat file of data, just write it out as a pickled Python object or something.
Properly normalized data with properly maintained indexes is efficient and fast despite the people with crap drives trying to prove otherwise.
Re: (Score:2)
Re: (Score:3)
Didn't we JUST have an article about the new MacBook Air's PCIe SSD getting 700-800MB/s in both read and write? Now a pseudo-press release claiming that 500MB/s is keeping up with the competition? All you're proving with ~500MB/s is that you can keep up with SATA.
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
There is actually no reason PCI connected storage should cost more than SATA connected storage. There is also no reason for it to be limited to Apple. Others will get around to it eventually and SSDs in their current form will only be around to support legacy systems.
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:3)
Re: (Score:2)
Good article stubbed its toe once (Score:1)
Quite informative piece, but Hothardware spells it "terra-bytes".
Re: (Score:2)
Endurance (Score:5, Interesting)
Personally I couldn't care less if the write throughput is 300, 400 or 500 MBps, but the write endurance of 80 TB, combined with the pseudo-SLC intermediate cache look pretty promising for home use. Intel 335 only specs 18 TB endurance.
Re:Endurance (Score:5, Informative)
I can't vouch for the modern SSDs, but I've had an X25-M in my netbook for years, it's used every day, and it's now reporting that it's down to 99% of its write capacity. At that rate the netbook will be in a museum long before the SSD dies.
Re: (Score:1)
I wouldn't try too hard to convince the "Derpability" crowd. There seems to be an vocal minority idea that if a SSD can't beat a platter drive in every conceivable way, then it's clearly unusable in any market for any reason. Forget the fact that we invented RAID as a way to mitigate the fact that hard drives are the least reliable part in a computer, this SSD can't work because it has a possibility of failure.
Re: (Score:1)
I wouldn't try too hard to convince the "Derpability" crowd.
I would. There's a known solution to derpability problems, which may start becoming commonplace as feature sizes decrease to 8nm. The endurance of an SSD can be increased dramatically by designing the chips to bake [arstechnica.com] their floating gates periodically.
Re: (Score:2)
"Forget the fact that we invented RAID as a way to mitigate the fact that hard drives are the least reliable part in a computer,.."
"We" didn't "invent RAID", "we" had nothing to do with it. Credit the creators of RAID for what they did. You clearly weren't part of that team.
No, RAID was not invented " to mitigate the fact that hard drives are the least reliable part in a computer". RAID was developed to enable lower cost drives to be useful in higher performance applications. Part of that is the need to
Re:Endurance (Score:4, Informative)
The X-25M had a write endurance of 7.5 TB for the 80 GB, and 15 TB for the 160 GB. If you've got the 160 and write an average of 100 MB per day, that's only 0.18 TB in 5 years, or barely more than 1% of the endurance. If on the other hand you wrote 10 GB a day (about two DVDs' worth), that would be 18 TB, and the drive would be likely shot. An HDD has infinite endurance. It can die from various failure modes, but not from "using up" the magnetic storage medium.
It completely depends on how much data you write.
Re: (Score:2)
Because the heads pivot back and forth on the same bearing in the same spot for ages, I'd guess that HDDs have a maximum amount of seeking they can do before they fail, whereas SDDs have a maximum amount of writing they can do...
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
"An HDD has infinite endurance."
No it doesn't, it just doesn't have the same parts that are being measured in this specific way. All devices, mechanical or electrical, can "wear" out.
It is also important in heavy load applications, which inherently exist in these asinine discussions, to normalize the wear rate of a device to the rate it is handling load. It wouldn't matter if an SSD wore out in 5 years in a given application if a HDD couldn't do the job at all.
Re: (Score:2)
Would you really expect capacity to run out linearly? Write balancing means that you should expect every bit of flash to be written roughly the same amount on average. That means once you start losing the ability to write to some cells, you can expect more cells to fail in short order.
Re: (Score:2)
Would you really expect capacity to run out linearly?
I believe Intel's capacity reporting in the SMART data for the drive is roughly linear. If I remember correctly, there's a separate indicator for the percentage of spare blocks in use after existing blocks begin to fail.
Re: (Score:1)
SSD reliability is a function of its total capacity, and write cycles per cell. Double the capacity makes for double the time to first cell failure. Ten times the cell write cycle gives ten times longer lifetime. Nice and linear relations, in other words.
Take a consumer drive with only 1000 writes per cell, and only 64 GB capacity, and write 24 GB per day, and you would start seeing the first failures after about 6.3 years, or about 55 TB. So to go all the way down to 18 TB (writing 1 GB every hour for two
SandDisk doesn't support their SSDs (Score:1)
Re: (Score:1)
What do you need firmware updates for? Are there bugs that need fixing in firmware?
Re: (Score:3)
Yes, there are. For example: http://hardware.slashdot.org/story/09/08/03/2051252/Intel-Confirms-Data-Corruption-Bug-Halts-New-SSDs [slashdot.org]
Re: (Score:3)
Re: (Score:1)
Recent sand force based drives are pretty good from all makers. Still, I get the impression that Sandisk just sticks whatever lowest bidder flash chips they can get at the moment in to the drives.
I'd stick to Intel or Samsung. Both of those companies have end-to-end control over the controller chip, firmware, and flash chips that go in to their drives.
Re: (Score:2)
Not firmware updates for the majority of their drives. Spend the extra $5 for a company who supports their products
Actually I'd rather support a company which releases polished products which do not need to be patched up afterwards.
mSATA (Score:2)
no mSATA form factor?
(I am currently deciding which mSATA ssd to put in my x230 alongside the 500GB HD.)
Sandisk's own flash? (Score:2)
From what I knew, Sandisk always sourced flash from Toshiba and provided it in all sorts of interfaces - CF, SD, micro SD, memory stick, even Fuji/Olympus' xD picture cards. So if anything was surprising, it was that they were not in the SSD market earlier
Probably b'cos they've managed to get the flash at the cost levels needed to be able to support SSD at its price points.
I wonder what will happen (Score:2)
Re: (Score:3)
Ah, OCZ, the SSD company with the highest rate of failure I've seen thus far. Granted, the earlier Vertex drives were supposedly responsible for much of that failure, but for the life of me, I cannot understand the popularity of this company.
Now Corsair on the other hand...;-)
Re: (Score:2)
You are a custom PC builder that promotes OCZ? I guess that's how the bell curve gets filled out.
After you use one, you'll understand why you are wrong.
Price per GB still a showstopper (Score:2)
Yeah, SSDs are all the rage now, but they're no match to spinning platters when it comes to $ per GB, which's really what matters to anyone who's not a gamer or DBA.
I'm not going to pay around $1 per GB just to see Windows start up faster.
Re:Price per GB still a showstopper (Score:5, Insightful)
I know I'd never go back for my laptop, personally
Re: (Score:2)
Right now, SSDs aren't a *economic* replacement for hard drives. But, if you're willing to shell out more for performance, then they're viable replacements. I'm running a 512MB SSD in my laptop for the OS (data on a hard drive RAID-1). I know there are organizations using SSDs for high-performance RAID-10s which is a great application if you've got the coin.
Re: (Score:2)
I'm running a 512MB SSD in my laptop
Do you have MS-DOS 6.22 as your main OS?
Re: (Score:2)
touche ;)
Re: (Score:2)
What application would a mirroring SSD RAID configuration be "great" for? SSDs do their own "striping" and error correction internally, so what is the failure mode that's so important that RAID-10 SSD would be "great"?
I find consistently that people who talk about RAID don't understand RAID, whether its /. or anywhere else. Self-proclaimed IT experts and architects consistently demonstrate this failure as well. Just look at the morons who proclaim RAID-5 to be dead and RAID-6 about to be because disks are
Re: (Score:3)
My computers each have more than one drive. It generally makes sense to have an SSD as a boot/software drive, and spinning rust for the large media files.
Cue music (Score:2)