Are these solid-state hard drives better than conventional hard drives? Are the prices coming down? How much longer until they’re comparable in price to regular hard drives?
I’ll be needing a new computer in the next year, and I get the impression that I should wait for this.
I wouldn’t be surprised if they were much cheaper within a year. Right now, they are still much more expensive than platter drives. I buy PCs for my company and am anxiously awaiting the right price point to make them standard. I don’t have long-term reliability data, but I’ve been testing one for about 4 months and have had no problems.
It will happen sometime in the next few years but solid state drives still have problems with high read-write data cycling. Businesses cannot afford to take risks with that so conventional hard drives will be around for quite some time. You run into the problem of comparing what is technologically cool versus what people (and especially businesses) actually need. Conventional hard drives still win on almost all accounts for practicality even for casual users. Experimenters build their own computers with solid state drives but the benefits are nebulous for most routine tasks right now. It would take a revolutionary solid state storage device to make that happen quickly.
I have heard it said–but cannot find a cite–that SSDs fail more gracefully than conventional drives. Specifically, that when they fail, they can no longer be written to, but can still be read. Is there any truth to this?
It would have to become a lot cheaper as Solid State drives degrade ultimately. If it got to the throwaway price point they would definitely replace hard drives.
Hoops I think that is true, I do not know from experience.
I’m not sure. I think the main advantage would be that SSDs are considered to be in a constant state of slowly failing bit by bit, and file systems for them are created to handle that by marking off the bad areas and working around them.
The sections which fail also fail when you write data over them. Reading doesn’t cause them to fail. So assuming that we’re not talking about someone hitting the device with a hammer, but rather just natural write cycle failure, then you can continue to read a section up until it fails with no worry that it will ever fail, until you write new data to it (and after that write, it will verify that the data was written, and if not write the data elsewhere, ad infinitum until there just isn’t anywhere non-broken to write to.)
Sage Rat That wouldn’t lead to a better image at the end, it would lead to a fried drive with an extremely fragmented data set with artifacts from old data.
I don’t think you’ll see them comparable to magnetic drives in a year. They’re almost an order of magnitude smaller and almost an order of magnitude more expensive than conventional drives right now. I expect they’ll come down, but they won’t come down that fast. And frankly, the speed improvements seem to be a mixed bag. Some of them are noticeably faster than magnetic hard drives, but many aren’t. And on some systems the bottleneck isn’t at the drive itself, but somewhere else along the SATA system. I’m waiting for these to become common, cheap, and reliable too, but they still seem pretty hit-or-miss for the cost right now–enough so that I don’t see it changing in mere months.
I work for an electronic distributor that represents (among many other suppliers) one of the SSD vendors. Things will get better with time but right now one of the problems is the time it takes to write to an SSD. SSD’s read times are typically better than HD’s so, for example, a laptop based on SSDs will boot Windows faster than it would off an HD. But the longer write times, for various ‘benchmark’ tests are sufficiently slower to limit widespread adoption.
[ Note; I’m talking above about read and write times not lifecycle issues. ]
But it will get better…
Another big plus for SSDs is they withstand shock/vibration better.
DVDs have a better chance of replacing CDs than SSDs have a chance of replacing HDDs. Cost factors aside, there are the limitations in read/write cycles as noted above. Also, the largest SSD I find on newegg atm is only 512GB, and it’s $1575. The largest HDD is 1.5TB (3x capacity) and can be had for about $120. I can also buy more than 10 of those HDDs for the cost of one SSD. (And still have money left over to start putting together a RAID array)
On a per GB basis, the SSD is about 3 per GB. the HDD is about .07 per GB. Even with the price of SSDs dropping, the SSD is still about 43 times more expensive than the HDD.
The price point isnt here and according to last years predictions we should have been at the price point by now which makes me assume that these things will probably never compete on a gibabyte/dollar basis with conventional drives. Drive manufacturers will drop the price on their higher capacity mechanical drives as needed.
That said, they do compete for the performance crowd, but SSDs have really lost their glow since anandtech and others wrote about them earlier this year. Specifically:
The arent power savers. Turns out they almost use as much power as a regular drive.
They slowdown over time. This could be fixed with a custom filesystem.
They fail like normal drives - usually unreadable and unrecoverable. The ability to to do emergency recoveries on failed flash media looks to be pretty difficult if not impossible.
Random read times are impressive, sequential writes and reads, not so much.
No one is yet sure how safe these are to run in a RAID array. Enterprise is hesitant to put all their data on these things. We may not know how well they can be trusted for years.
Some people think that as the technology evolves the big bottleneck will be SATA. The final version of these drives might just be something that plugs straight into your PCI slot.
Turns out most people dont care about drive performance as to pay a 20X premium. Its like bluray vs DVD. Sure bluray is nice, but is it worth it?
Mmno. It just means that the overall useable space shrinks. I’ve made a device driver and file system for SSDs, so I’m confident that what I said is accurate. Probably you misunderstood me at some point.
Any change to the filesystem verifies, after the write, whether the written data matches what was written. If it doesn’t it marks that block as bad and writes the data elsewhere. There’s no loss in data just, like I said, slowly the amount of free space on the device shrinks until you can’t make any further changes successfully.
To augment what Sage Rat said in a reply, a solid state drive doesn’t care how fragmented the data is. Unless the drive software has some oddities, it takes no more time to read or write data at one storage location than any another. Memory pointers are faster than rotation and head seeks and not subject to physical delays.
And who cares about “artifacts from old data”? If nothing points to a memory block, it will never be accessed. It becomes invisible to most applications. The total storage capacity just shrinks.
SSDs fall in price with Moore’s Law. So, you can expect them to be twice as cheap or twice as capacious in 2 years. It’s not such a hot rate of progress.
(That said, a year or two ago the drives had a high mark-up over the cost of the flash chips. This markup is now gone, thanks to competition. The prices have flattened out, and they won’t get cheaper soon. Performance, however, may still improve, particularly for low-cost drives.) EDIT: This is for drives on Newegg. Drives built into laptops still often have heavy markups. Those markups will indeed fall over the next year.
If I understand it right, the gracefulness of the SSD failure is the combination of the factthat one small block will fail, but the rest of the drive will still work, and that failure can be discovered as the data is being written?
As oppposed to a magnetic hard drive, which typically fails without warning and often in a way that trashed the entire drive.
Well, a magnetic hard drive will also have small-block failure, but often it will be discovered in reading a block that was written to a while ago - meaning that the data in that block was lost. That may mean that you only lose a small file or a part of a big file - it may mean more difficulties if that block was part of a directory, or big problems if it was part of the disk’s master allocation table.
Magnetic drives can also break in much bigger ways, which I think might have to do with the spinning platters bumping against each other or something like that.
ETA: Okay, from a quick check on wikipedia I got it wrong. The infamous hard disk failure is a ‘head crash’, meaning that the magnetic sensor that reads or writes to the platter collides with the disk itself. Nasty. Dust is apparently a factor in these too.
Wait, they have 512 GB SSDs? Why aren’t we seeing bigger capacity iPods/Zunes/Zens etc.? I have so much music and prefer the highest resolutions that I need at least 300GB, and I don’t want a plattered device.
Because they’re really expensive. Last I checked, a 512GB SSD cost somewhere around $1500.
Also they aren’t very compact. Most HDD-based MP3 players use 1.8" drives, and I don’t think you can get a 512GB SSD in that size. So a 512GB SDD MP3 player would be larger than the current iPod Classic.
Not from what I’ve seen. A fast SSD that holds the OS is one of the best ways to speed up your system nowadays. Much more effective than adding RAM or a bit more CPU speed.
Also, as far as reliability, hard drives aren’t exactly stellar. That’s why businesses use clusters that make the drives redundant.