Slashdot Mirror
SSDs vs. Hard Drives In Value Comparison
EconolineCrush writes "SSDs hardly offer compelling value on the cost-per-gigabyte basis. But what if one considers performance per dollar? This article takes a closer look at the value proposition offered by today's most common SSDs, mixing raw performance data with each drive's cost, both per gigabyte and as a component of a complete system. A dozen SSD configurations are compared, and results from a collection of mechanical hard drives provide additional context. The data are laid out in detailed scatter plots clearly illustrating the most favorable intersections of price and performance, and you might be surprised to see just how well the SSDs fare versus traditional hard drives. A few of the SSDs offer much better value than their solid-state competitors, too."
It says: "A few of the SSDs offer much better value than their solid-state competitors, too."
Is that meant to be "SSDs"?
While a pretty comprehensive article, nowhere do they actually talk about reliablity and longevity of these drives in their value calculations. That's a pretty important factor for me, and has been one of the reasons (besides price) that I haven't seriously considered one yet.
I don't know about the cost/benefit for most people, but we're all running SSDs for our laptops now, and it's definitely worth it.
Once I realized that I could fit on a 64GB SSD comfortably if I didn't keep my ENTIRE photo collection on my laptop, it was a pretty easy decision to make to try them.
And after some testing, I've decided that it's enough worth it for us that we're all using them. In most cases it isn't a bit noticeable difference. But for some things it really does make a difference, and not having to wait for them is a big gain. The things that are a lot faster are: booting (rarely, but you're entirely "down" while doing it), opening big apps like OpenOffice, re-opening firefox or thunderbird when they flake out, and doing big find/grep jobs. Searching through e-mail and the like? Great.
For a long time, CPU increases were way outpacing the disc performance gains. We how have CPUs that are faster than most of my staff can really take advantage of on our laptops. But disc performance, even at 7200 RPM, was often the bottleneck.
So, we've traded volume for performance, and been very happy with it.
Sorry, but until I can get an SSD and not have to spend almost the same amount of money again for a drive to store media and games on, no deal. They are just way too expensive per GB, and I'd rather pay for one HDD to get a lot of space than pay for a HDD PLUS an SSD just to get a speed increase with only slightly more space.
I'm afraid that people jumping big-time on the expensive SSD bandwagon, though, will not encourage makers to decrease prices as fast as if people would have actually smartly waited until they were a decent price to size ratio.
"Intel X25-V Intel PC29AS21BA0 32MB 40GB $110"
I stopped reading there. Everything else was at least three times as expensive and I'm not spending half as much on a damn SSD as the whole computer which is the new Mac mini model at $750CAD.
I doubt you would. I have a 40 GB Fujitsu MPG3409AT-E hard disk from 2001 that is still running yet the so called best Seagate Pulsar - the "first enterprise-ready" SSD failed after less than a year of database usage.
Bottom line: Do not trust SSDs.
Super Star Destroyers are better value?!?!
While most every hard disk supports and respects proper cache flush semantics, SSDs typically trade performance for data integrity. Although it should be a standard feature, very few SSDs include a capacitor to prevent filesystem/data corruption in the event of power loss.
Unfortunately, the vendors are very secretive about SSD internals, and the algorithms they choose to employ can also have a significant effect on data integrity. At this point in time, there is far too much blind faith required, and many vendors definitely do not deserve it.
I've seen, and have been able to reproduce reliably, hard disks losing their internal cache data, claiming to have written it to platter when in fact it was not. And I am /not/ talking about battery-backed RAID cache, OS write cache, or anything of that nature; I am speaking specifically of the internal hard disk cache.
When we figured out what was going on, needless to say we were all a bit shaken. But the lesson is learned: your storage needs to have a battery backup system.
Nemilar http://www.techthrob.com - Visit Me!
I think the consumer trend is pretty clear with respect to SSDs (enterprise-level I think is still uncertain). Consumers like the speed and the battery savings (laptops being incredibly popular now) that SSDs provide, but of course there is no way you are going to get the sheer quantity of storage space that you can get with hard disks.
Consequently, a lot of companies are marketing "home storage servers." I've seen Lenovo, Acer, Asus, etc... all come out with small 4 or 5 bay boxes, usually running Windows Home Server, all aimed at the mid-range consumer market. It makes complete sense to put the platters in a box, where you can keep network-accessible massive storage, and to put the fast, low-power SSD into your client machine.
The problem arises when you need to access what's on that home NAS while you're out on the road. While I think many people have the upload bandwidth for streaming music, I don't think that exists for video (at least, not in the United States, or at least not where I live). So sites like hulu, etc.. will remain popular in that regard for the time being.
Nemilar http://www.techthrob.com - Visit Me!
About 5 months ago I bought a $700 250G SSD for my laptop and ditched the spinning disk. The system is overall faster, and for someone who's used HDs since the 286 days and floppies before then, the performance is oddly different (almost always better). The big bonus though is that my laptop takes about 10 seconds to boot (once past the BIOS) while it used to take about a minute. This has changed the way I use my computer, and is enough to justify the swap. I do have a few other systems I occasionally use, and apart from the OLPC XO-1 (which has its own performance characteristics that are different again from anything else I've seen), it's now kind of irritating to use spinning disks and feel those delays again. As the costs go down, I imagine anyone who's tasted SSDs will spread the technology very broadly among their friends.
For every problem, there is at least one solution that is simple, neat, and wrong.
OTOH, with spinning drives, you have to wait for the platter to spin back around to the head a few times before you can flush that cache. Maybe SSDs don't need anywhere near as big a capacitor to keep 'em going during that last-gasp writecycle?
Come to think of it, anyone have any public literature about how HDDs manage to pull this off? There may not be much talk about it in SSD-maker circles, but I doubt there's a lot of talk about it in HDD-maker circles either. It's not like it's easy for benchmarkers to test for it, and while it's presumably important to server guys, it's not the first thing on the spec sheet... and the retail market probably never thinks about it.
I dropped my latitude d620 on the concrete floor my desk sits on, and it crashed the hard drive instantly. For the replacement drive I let a friend convince me to shell out for the SSD. It's amazing. I no longer have to worry about bad sectors, my battery lasts longer, the machine is cooler, it's quieter, and the OS loads in like 5 seconds to usable state with virus scanner etc.
I have a couple slow terabyte hard drives in my old system I use for a media system/home file server, but for systems I actually use for work or play I'll never have any non solid state drive as a primary drive again.
I read a lot about SSDs before buying one, and one thing I don't see stressed very often when it comes to analyzing the 'Value' of an SSD is how the life of a computer can be prolonged by adding one.
I have a three-year-old T61 Thinkpad. I would typically replace a machine like this in another year or two. However, when I added an SSD (Intel's X25-V, for about $100) to my machine, the performance boost is so great that I believe I've further extended the life of this laptop by at least another year or two.
When viewed in this light, upgrading to an SSD becomes a no-brainer for a wide variety of use cases. The lowered TCO is THAT significant.
If your really a budget consumer, and are using the hard drive to get crap done then at the cheapest rate a laptop replacement SSD from newegg is going to cost you like 80 dollars more for a 64 gb SSD than a 500gb hard drive. If your time is worth 50 bucks an hour on the market, and your boot time is reduced by 2.5 minutes your ROI is at break even in around 3 work weeks according to my head math.
Don't chase dimes with dollars.
It's been obvious to me for a while that drive manufacturers are missing the boat with adding a flash backup area the size of their RAM cache and some caps to give it the ability to save the RAM to flash. This would allow you to return the 'written' status to the OS much faster, _and_ be safe in the event of a power failure.
For more points, add more flash and smarts and use the flash as a cache for 'hot' portions of the drive.
Awesome furniture, accessories and cabinetry in Santa Rosa, CA: http://humanity-home.com/
We have started to deploy more multimedia intense apps and found most of our 3+ year old laptops where dogs at running them..
We then did some side by side benchmarks between an old laptop with the HD replaced with an SSD vs a new laptop with a new normal HD. Guess what? In MOST tests the old laptop performed BETTER than the new one, despite the new laptop having a faster CPU and main board...
Guess what, although they cost WAY more than a new normal HD per GB, they are WAY cheaper than a new laptop!
EA David Gardner -"... but the consumers have proven that actually what they want is fun."
Hard drives tend to use the momentum stored in the spindle itself to at least park heads after a power failure (especially for laptop drives that park away from the media). This presumably works by powering the drive's rails through the motor controller's protection diodes. I'm not sure if they also use it for last-gasp writing of write-cached data, though. i guess it depends on whether the write controller can handle media that is losing speed.
Cost is not much of a worry for me. Reliability and random write performance is all I care about. It is my understanding oxide munching flash memory has improved somewhat but is still a problem. Intel is touting a 228 YEAR MTBF for some of their SSD products. Such claims seem to me to be rediculous and unrealistic.
Paradoxically ssd disks can take more energy than spinning platters when activly moving bytes based on process technology and how well the a/d's are designed. Obviously idle power consumption on SSD side is a huge plus.
I'm going to wait a few years for the technology to mature a bit more before jumping on the SSD bandwagon. Really hoping to eventually see a technology thats not flash based with more DRAM like characteristics take over for persistant storage.
...but I think I've heard enough from the existing comments.
I'm currently working on my Dell m1530, and it feels about as hot as the pan I use to fry eggs. It also doesn't have the best battery life, especially if I'm trying to watch movies or catch up on work while traveling. It sounds like switching to an SSD will help on both of those fronts.
I just hope that the price of SSDs drops by the time I'm in the market for one. I'm not entirely sure I'd like to drop $400 or $600 in addition to the $1000 for my next laptop...
coding is life
If you have a 10 minute boot up, and people on the road visiting clients, several times a day, (and standby is disabled because of security concerns with disk encryption) then a 3 minute boot can pay for itself in a few months.
If your laptops take 10 minutes to boot, you've got much bigger problems...and how is standby a concern with disk encryption? If you wake the machine, you should have to enter a password.
What are you storing that requires this level of paranoia with so many client visits? Clearly not defense.
Please help metamoderate.
I dropped a Seagate Momentus XT 500 GB in my Macbook Pro for $130 the other day. It has a 4 GB SSD-like swap-space on it and it's totally boss. You don't get the performance of an SSD, but you do get better than average performance for not much more. http://www.anandtech.com/show/3734/seagates-momentus-xt-review-finally-a-good-hybrid-hdd
The minute I found that 100% CPU load with long operations didn't generate heat along with crunching speeds that were doubled from my prior hard drive, it convinced me that I had just become far more productive for my 3D work. It didn't take me long to understand that more speed made me a lot more productive. It was so good, I put 2 SSDs in my i7 Mac Book Pro using MCE's Optibay so I could get the 2nd SSD in the CD/DVD bay (which I never use on the road).
Comment removed based on user account deletion
If you have a 10 minute boot up, and people on the road visiting clients, several times a day, (and standby is disabled because of security concerns with disk encryption) then a 3 minute boot can pay for itself in a few months.
If your laptops take 10 minutes to boot, you've got much bigger problems...and how is standby a concern with disk encryption? If you wake the machine, you should have to enter a password.
What are you storing that requires this level of paranoia with so many client visits? Clearly not defense.
Google for cold boot attacks - there are issues with disk encryption + standby. Posting as AC cause im lazy.
i dont know about you guys but i dont like the sounds that HDDs make. newer HDDs are much quieter but still audible, so i got a sweet SSD. now my PC runs nicely without moving parts with exception to my media storage HDD drive that spins up when i need it and DVDRW drive when it has media. yep, no fans or water cooling on anything, just silence.
i love my SSD.
No, they don't do the last-gasp writing. It simply takes too long to do it and it's too risky as the speed is uncontrolled and there's always a danger of overwriting critical areas by accident (servo tracks, firmware regions, control data, etc) which would render the drive unusuable.
In fact, this sort of power down is designated as "emergency stop" - the momentum is used to turn the spindle motor into a generator, and the power is dumped into the voice coils directly. It's quite a violent procedure and most drives are severely derated. I've seen one rated to 50,000 load-unload cycles, but only 10,000 "emergency unloads". It's just that all those pieces slamming into each other start wearing out the mechanical bits.
Full disk encryption + AV + bloatware = 10 minute boot easily.
There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
Nope, they just lose whatever writes are in cache, that's why RAID controllers disable write caching on the drives themselves (one good reason to only use supported drives with servers, the manufacturer has made sure this process actually works).
There are 4 boxes to use in the defense of liberty: soap, ballot, jury, ammo. Use in that order. Starting now.
The test is very unfair on small SSDs like the Intel X25-V because it doesn't look at overall price, only $/Gb. Hardly anybody is going to install a small SSD as the only drive in a machine. Most people would combine them with a big hard disk so the final score would be a blend of the scores for the SSD and the second hard disk.
eg. I just rebuilt my machine with an X25-V for the OS and applications. The X25-V gives the machine amazing boot up times and near-instant application load times - way faster then my old Velociraptor. As an overall performance enhancement it's a complete no-brainer for $110.
For the price of a big SSD you can probably get an X25-V (boot drive) plus a 300Gb Velociraptor (video editing and/or your hardcore games) plus a 1.5Tb HDD (for your torrentz and AVIs). Beat that for price/performance!
No sig today...
high altitude computing. I was reading an article about mountaineers with laptops, when you get up around 15 or 16 thousand feet the air pressure is so low that the Bernoulli effect no longer works properly in your hard drive, so your drive makes lots of nasty noise and is more prone to failure. With SSDs you just have to worry about the lack of oxygen damaging your brain and your internal organs, but not about endangering your data or the performance of your laptop.
cheap labor conservatives - they want to keep you hungry enough to be thankful for minimum wage.
[Citation needed]
Requesting link for proof that most HDDs or at least some specific ones park the heads after unexpected power-loss.
Doing large scale tax evasion? :)
The article sounds like an "industry sponsored report" designed to sell more SSDs because, presumably, the manufacturers are not selling as many as they would like to.
Personally, I don't see what value SSDs bring based on how expensive they are currently:
1. Reliability? - A responsible computer user will still need to maintain backups of SSDs in the same way that they currently do for hard disks. Sure, the failure rate of SSDs may be lower but, ultimately, every SSD will eventually fail - and because it's a new technology, people do need to be extra vigilant for previously unforeseen problems that may only appear after millions of them have been sold. The price of three hard disks (a mirrored pair and a backup disk) is still far cheaper than one SSD.
2. Battery life? - I cannot argue with this one except to say it's still cheaper to buy a couple of spare laptop/netbook batteries than it is to buy an SSD.
3 - Bootup/operational speed - I'd certainly be impatient waiting 5 or 6 minutes for a computer to boot up but I'm not sure my life is that busy that waiting 30 seconds for a hard disk as opposed to 3 seconds for an SSD matters that much to me. In my 30 years computing experience, machine speed comes from avoiding bottlenecks and good OS optimisations - yes, a faster SSD helps with the hard disk speed bottleneck but that still leaves things like the amount of memory, CPU power, OS bloat and fragmentation to consider.
I'm certainly not dissing SSD, it's a logical progression to the hard disk, but for the current prices of them, there's not enough benefit to me that justifies replacing my hard disks with them.
Gentoo Linux - another day, another USE flag.
this is [original research], but if you pop open a few drives, unlock the head, pull the head out to the middle of the drive, then rotate the platter as you would normally rotate it--the heads go back and park.
You'd probably be surprised how many drives don't actually flush the data and lie to you. There was a Slashdot story about a tool that the Livejournal.com guy wrote. It tests whether your drive tells the truth. http://brad.livejournal.com/2116715.html
"Hey, the normal metric makes us ssd manufacturers look bad. Quick, invent a new metric and hire someone to send it to /. !"
We can tell you're lazy, because apparently you haven't bothered to Google for "cold boot attacks" yourself.
So now you should Google for "cold boot attacks", read until you understand them, then come back and explain how "standby" (i.e. suspend-to-RAM) is more dangerous than simply be on and running normally (hint: it's not).
Whether the machine is running normally or in standby mode an attacker can power-cycle the machine and access the encryption key from the uncleared RAM. There's literally no difference in the attack if the machine is standby versus normal operating mode. And if you have 10-minute boot times I guarantee that your users are not powering off their machine every time they leave it to get coffee.
The only power mode you have to worry about it suspend-to-disk, because a poorly designed encryption system might not dismount the encrypted volumes before the RAM is copied to disk, which would result in on-disk encryption keys. So if you wanted to disable that power mode, and/or verify that your encryption solution properly dismounts volumes before hibernating, that would be reasonable.
Pros:
Lower power consumption, good for laptop battery life and energy bills
Immunity to low ambient pressure
Random reads are WAY faster than on HDDs. WAY faster.
Cons:
Finite write endurance (especially the newer die-shrunk SLC NAND, but the die-shrinking affects SLC, too)
Capacity / Expense (especially SLC)
Immature technology - most manufacturers still don't have it right. The original JMicron controller is a good example, having no wear leveling algorithm at all. The Indilinx controller is another, having horrible wear leveling that causes write amplification factors up in the teens. Sandforce is marginally better, but lacks any kind of caching that can be used to improve WA and/or reduce the number of erasures required. They claim WA less than 1 based on compression, but that's only under lab conditions with very deliberately-chosen write patterns.
Also, SSDs completely lack any kind of elegant O/S support. Windows sits there and churns away 15kB/s of writes 24/7, slowing eating away at the write endurance of the drive. It also makes no effort to block write, so it'll sit there and send a few bytes at a time, exacerbating the write amplification problem. TRIM is nice, but only a bandaid to the larger problem. OSX still doesn't support TRIM that I know of. I don't know what Linux would do to one... I haven't tried, and am kind afraid of investing a few hundred bucks to find out.
That works for desktop drives and is a mechanical feature. For 2.5" drives that park away from the media, though, you need to actually kill power while the drive is running to see it in action, and it's an electrical feature that uses the stored energy of the spindle.
Sometimes value isn't practically measured just by numbers/benchmarks. I think this is one of those cases.
I think personal enjoyment and your user experience trumps data of Performance per dollar from a chart. If updating your PC to use SSD storage signifigantly improves your user experince on a day to day basis, it's probably worth it.
As an example, what's the difference between a $10 bottle of wine and a $20 one? You could compare alchol levels, etc., but in the end the taste, and palate (ie. user experience) is what matters. Sometimes it's not really possible to put a value on these things using charts and graphs. Your own opinion and what the value is for that convenience/experince is the true measure.
At least the company I work for has super caps in the SSDs for SAS and high-end SATA SSDs.
I doubt your going to get any such service on your Best Buy specials however.
That tool does not specifically check hardware--it checks the end-to-end functionality of fsync(). If the drive doesn't flush the cache properly it will fail, but failure does not necessarily indicate that the drive is responsible.
For example, it will fail on any Mac, since Apple's fsync() semantics are broken by default. If you want your applications to work properly on the Mac, you need to replace calls to fsync() with fcntl(F_FULLFSYNC), like so:
--- diskchecker.pl 2009-02-09 20:45:43.000000000 -0600
+++ diskchecker-darwin.pl 2009-02-10 13:40:07.000000000 -0600
@@ -14,6 +14,7 @@
use strict;
use IO::Socket::INET;
use IO::Handle;
+use Fcntl;
use Getopt::Long;
use Socket qw(IPPROTO_TCP TCP_NODELAY);
@@ -134,7 +135,7 @@
sysseek F,$offset,0;
my $wv = syswrite(F, $buf, $LEN);
die "return value wasn't $LEN\n" unless $wv == $LEN;
- $ioh->sync or die "couldn't do IO::Handle::sync"; # does fsync
+ fcntl($ioh, 51, 0) or die "couldn't do fcntl(F_FULLFSYNC)";
sendmsg($sock, "post\t$pagenum\t$rand");
In my experience, this is the exception, rather than the rule with hard disks. The most common problem is with USB/1394 enclosures, where the bridge chip typically only implements the barest subset of commands. I would be surprised if you found any relatively modern disk itself which lied, though if you have, it would be appreciated if you could share your findings.
However, I disagree with your conclusion; a battery backup is a band-aid solution. It is best to avoid such hardware. For critical systems, one should either buy pre-qualified hardware, or qualify it themselves.
The fact that my PC has had more than a few power failures and the drives aren't paper weights?
A drive that doesn't park its heads on power loss will instead suffer a head crash. Unless something has changed quite a bit, a head crash usually results in the loss of the read/write heads, and a small part of the media itself. The rest of the media is probably fine (well, aside from being covered in shavings of whatever got carved out of the crash area), but you'd need to mount it in a new drive to read it since your heads are gone.
Now, they might or might not park the heads in a manner that is most suitable for shipping/etc, but drives clearly park them someplace other than on the media itself...
Windows boots refreshingly quick, but was that worth the extra price?
Ummm...this was my point (I think), ie. you can have that speedup with a small/cheap SSD.
A $100 SSD represents immense value for money as a performance upgrade. It makes your whole system more snappy. Good luck getting that much extra performance from a CPU upgrade or whatever.
No sig today...
Yup. In every hard drive design, write current is disabled hard as soon as one of the supply voltage comparators trips.
Emergency retract is absolutely required because if the heads land on the smooth platters, chances are very high that the drive is dead due to stiction. I don't know if any design still has textured landing zones; my suspicion is that everybody uses ramp loading these days to keep the heads off the surface.
thegodmovie.com - watch it
Your collection of gay pr0n if you're a Bible salesman in Alabama?
To have a right to do a thing is not at all the same as to be right in doing it