Given the massive increase in productivity any SSD provides over mechanical drives ~$250 is more than worth the price if you don't already have an SSD.
Yeah, sand >> rust, and I couldn't imagine switching back now.
However, I now prefer to buy SSD models that have been available for more than 6 months.
I've owned 6 SSDs and 4 of them failed (3 OCZ, BTW). I've seen some really crazy failure modes. I don't buy OCZ anymore :/
One SSD I had, it came from the store booting into a busted Win98 install. I managed to install an entire new OS, including a soft reboot. OK, OS install complete.
After a hard reboot (power off), the drive would boot back up into the broken Win98 install. That was my first inkling of just how fundamentally different SSDs were internally compared to standard HDDs. I had trouble even coming up with a reasonable hypothesis to explain that failure mode. See e.g: http://www.dslreports.com/forum/r25491097-Dell-Laptop-and-SS...
It would now not surprise me to learn that other types of SSD failures could result in data loss you might not neccesarily notice at first.
But definitely, backups to HDD. 256GB backs up fast, so it's not too bad.
The last I checked the MTBF on SSD's was substantially longer than for hard-drives. Barring firmware bugs you should probably strike that off your list of reasons for not using a SSD.
Even if you do think SSD's are horribly unreliable, they're so fast that they're still worth using. Just put your OS and programs on a SSD and store your data on a HD. If the SSD fails, you've lost nothing you can't reinstall easily. I started doing this a couple of years ago and there's absolutely no going back.
We've had 3 SSD failures out of 5 SSDs, and 0 magnetic failures out of 8 magnetic drives in our office over the last 2 years. We definitely keep only OS, programs, and VMs on SSDs.
Given the difference in reputation between OCZ and not-OCZ I couldn't imagine buying them on anything other than a "way cheaper than everything else" basis.
Unremarkable benchmarks plus no cheaper than models that have been on the market for some time? Double-pass.
Are you saying OCZ has a bad reputation? I haven't followed things closely, but I used a Vertex 2 and now use a Vertex 3 based on their good reviews at http://www.anandtech.com/tag/storage
They're fast drives but in the past have had the highest failure rates and firmware problems. Anecdotally Samsung and Crucial drives are much more reliable, while the Intel drives are probably the best because of Intel's verification.
For those curious about high-end SSD's. I took the bait and it really wasn't worth it.
I paid $300 for a Vertex 3 when it first came out. It advertised 500MB/s up and down with SATA6.
I upgraded from one that was around 150MB/s.
I'm a developer, on my computer all day and to be honest, I rarely noticed the difference between the two.
The difference between HDD and SSD is massive. But between different SSD tiers, it's not super noticeable unless you are a heavy writer/reader for maybe something like video editting.
Kind of like 24bit music vs 16bit.
Spend the cash on size and reliability, not performance.
The feeling you get from a good SSD is largely due to reduced IO latency (both read and write).
The ability to absorb random writes (to a degree) is huge too. It's why the 2008-vintage Intel X25-M caught the industry by surprise. Other SSDs at the time had extremely poor random write latencies.
Bandwidth is secondary, unless you are talking about an enterprise/database deployment (where queue depth is deep and you want great bandwidth and latency).
Pedants will note that bandwidth and latency are related (see Little's Law).
To be fair, if you only look at sequential access numbers on any disk your likely to be disappointed unless you happen to have a very sequential workload, which isn't terribly common.
Loading large applications, copying large files, loading the OS, some types of video editing, etc may benefit from sequential performance but for the majority of real world workflows 4k random I/O per second is a much better metric (though certainly not universal). IO/sec is much more important in the area of dev work, database performance in particular. The current generation of SSDs are massively faster in IO/sec than older models, much greater improvement than the sequential numbers indicate. That said, you still may not notice a difference if what your doing doesn't hammers a database that is 'safe' (guaranteed consistency on disk) or if your working set doesn't outstrip the ram you've allocated to the db.
However, a 128G boot drive for Windows is cramped without creative use of junctions - I've been there with 80G X25M, it's worth paying over the odds for the space.
Fiddling around too much with junctions can break some installers, for example, that expect the MoveFile() API to work for directories whose paths start with the same drive letter, or do weird things with NTFS hard links.
There are a ton of corner cases that can produce weird results, so it's possible that that result is correct (but probably not representative of real-world performance). If you read a different review that uses a different workload driver, it will probably produce different 4KB write numbers.
Do your research if you're planning to soup up a Macbook with an OCZ SSD. Whichever party is to blame, they often don't play well together. Mine randomly freezes for up to 30 seconds every couple of days. Others have reported worse symptoms. By the time I get another SSD, I'll probably be getting a new computer, but for anyone else thinking about it:
1) Don't buy OCZ, OR make sure it has a SATAII, not a SATAIII interface, (That may apply to other brands if OS X is to blame; I'm not an expert on the issue)
24 comments
[ 5.1 ms ] story [ 70.1 ms ] threadHowever, I now prefer to buy SSD models that have been available for more than 6 months.
I've owned 6 SSDs and 4 of them failed (3 OCZ, BTW). I've seen some really crazy failure modes. I don't buy OCZ anymore :/
One SSD I had, it came from the store booting into a busted Win98 install. I managed to install an entire new OS, including a soft reboot. OK, OS install complete.
After a hard reboot (power off), the drive would boot back up into the broken Win98 install. That was my first inkling of just how fundamentally different SSDs were internally compared to standard HDDs. I had trouble even coming up with a reasonable hypothesis to explain that failure mode. See e.g: http://www.dslreports.com/forum/r25491097-Dell-Laptop-and-SS...
It would now not surprise me to learn that other types of SSD failures could result in data loss you might not neccesarily notice at first.
But definitely, backups to HDD. 256GB backs up fast, so it's not too bad.
Even if you do think SSD's are horribly unreliable, they're so fast that they're still worth using. Just put your OS and programs on a SSD and store your data on a HD. If the SSD fails, you've lost nothing you can't reinstall easily. I started doing this a couple of years ago and there's absolutely no going back.
At this point I would recommend buying SSDs that have proven track records; don't buy cheapest on the basis that any SSD is better than any HDD.
Unremarkable benchmarks plus no cheaper than models that have been on the market for some time? Double-pass.
SH/SC says "Do not buy drives from OCZ or last-generation (SATA300) models, they are not reliable." and refers people to this list: http://www.behardware.com/articles/881-7/components-returns-...
It's possible that they're the entire cause of the "SSDs are less reliable than spinning disk" conventional wisdom.
I paid $300 for a Vertex 3 when it first came out. It advertised 500MB/s up and down with SATA6.
I upgraded from one that was around 150MB/s.
I'm a developer, on my computer all day and to be honest, I rarely noticed the difference between the two.
The difference between HDD and SSD is massive. But between different SSD tiers, it's not super noticeable unless you are a heavy writer/reader for maybe something like video editting.
Kind of like 24bit music vs 16bit.
Spend the cash on size and reliability, not performance.
The feeling you get from a good SSD is largely due to reduced IO latency (both read and write).
The ability to absorb random writes (to a degree) is huge too. It's why the 2008-vintage Intel X25-M caught the industry by surprise. Other SSDs at the time had extremely poor random write latencies.
Bandwidth is secondary, unless you are talking about an enterprise/database deployment (where queue depth is deep and you want great bandwidth and latency).
Pedants will note that bandwidth and latency are related (see Little's Law).
Loading large applications, copying large files, loading the OS, some types of video editing, etc may benefit from sequential performance but for the majority of real world workflows 4k random I/O per second is a much better metric (though certainly not universal). IO/sec is much more important in the area of dev work, database performance in particular. The current generation of SSDs are massively faster in IO/sec than older models, much greater improvement than the sequential numbers indicate. That said, you still may not notice a difference if what your doing doesn't hammers a database that is 'safe' (guaranteed consistency on disk) or if your working set doesn't outstrip the ram you've allocated to the db.
Fiddling around too much with junctions can break some installers, for example, that expect the MoveFile() API to work for directories whose paths start with the same drive letter, or do weird things with NTFS hard links.
Anyone can shed light on this? Can this benchmark be flawed somehow? (I myself am trying to convince whether SSD is good or not).
1) Don't buy OCZ, OR make sure it has a SATAII, not a SATAIII interface, (That may apply to other brands if OS X is to blame; I'm not an expert on the issue)
2) Take out the optical drive and put it there.