SPCR

Thanks for the review. I believe there's a small error in labeling on the third page. The third graph, "File Copy Performance," has the Kingston drive labeled in black numbers and dark colors instead of using the scheme of light red/light blue/white numbers.

Oh a Western Digital SSD, can we be sure it has exactly zero platters?

I've been using a 60GB Corsair Force drive for a little while and the performance is indeed impressive. On the other hand I have been suffering somewhat from known stability issues related to the Sandforce-1200 controller when running a system for more than half a day at a time. It ought to be solved with the latest firmware from Sandforce but Corsair is both slow to relay firmwares and a bit oddball when it comes to communicating. That OCZ drive would have been nice right now.

The 64GB C300 is notably absent from the roundup and it'd be nice to warn readers that Intel's next generation drives should shake up the pricing scheme in the near future (early 2011).

Thanks for the review.

I notice an AMD system was used for this review.

I recently built an AMD system with a SandForce drive and found all sorts of surprises with the AMD south bridge (I am using a SB850).

A lot of people are saying that AMD's Cool'n'Quiet and C1E has a huge impact on speeds. I haven't tried disabling those features, but I do see that HDTune's read speed goes up with each CPU core I load with Prime95. Loading all four cores gives me maximum SSD performance. :O

Also, a lot of forum posts around the web say that the AMD AHCI drivers do not handle TRIM. Others say the newest version does. But the AHCI drivers I saw in Catalyst 10.10 have old dates. Microsoft's default AHCI drivers seem to work very well.

Oh, and specific to the SB850, supposedly disabling the SATA 3.0 6 Gbit/s feature improves speeds further.

josephclemente wrote:A lot of people are saying that AMD's Cool'n'Quiet and C1E has a huge impact on speeds. I haven't tried disabling those features, but I do see that HDTune's read speed goes up with each CPU core I load with Prime95. Loading all four cores gives me maximum SSD performance. :O

Oh, and specific to the SB850, supposedly disabling the SATA 3.0 6 Gbit/s feature improves speeds further.

Cripes, computers can be confusing sometimes! Thanks for this info, I'm building a rig soon with same southbridge and was planning on using an SSD (or maybe two in RAID if funds allow)

josephclemente wrote:I notice an AMD system was used for this review.

I recently built an AMD system with a SandForce drive and found all sorts of surprises with the AMD south bridge (I am using a SB850).

A lot of people are saying that AMD's Cool'n'Quiet and C1E has a huge impact on speeds. I haven't tried disabling those features, but I do see that HDTune's read speed goes up with each CPU core I load with Prime95. Loading all four cores gives me maximum SSD performance. :O

Also, a lot of forum posts around the web say that the AMD AHCI drivers do not handle TRIM. Others say the newest version does. But the AHCI drivers I saw in Catalyst 10.10 have old dates. Microsoft's default AHCI drivers seem to work very well.

Oh, and specific to the SB850, supposedly disabling the SATA 3.0 6 Gbit/s feature improves speeds further.

Is it a noticeable speed difference or just lower nr's on HDtune. If HDtune is only IO and all CPU's go to sleep at each read this is perfectly possible.

I've been considering buying a SSD drive as well...
I've been wanting to upgrade my mainboard/CPU/RAM combo but it often seems the performance hit comes from a slow and burdened hdd...
If application load and read performances are much butter with SSD, then it might be a more suitable upgrade and more noticeable across everyday use in contrast to specific tasks (like video encoding)...

Not certain how it would behave with a s939 Nforce4 sata controller though, after reading the above comments...

I have a 160GB Intel SSD. It's a mid-range one and performance is excellent. While other SSDs will out-perform it on raw transfer rates it wins by far on reliability and hassle-free use.

The problems with SSDs slowing down over time are now receding, especially thanks to Windows 7's use of the TRIM command. Even so, over time Intel SSDs maintain performance better than most of their rivals.

The only down-side with SSDs is that encryption is no longer transparent. On a HDD you would be pushed to notice any difference between an encrypted OS and a non-encrypted one, but with an SSD there is a marked performance drop. Still faster than a HDD but annoying none the less. The new AES acceleration in Intel CPUs should help but I can't find any benchmarks on SSDs with them yet.

Forget the next generation Intel SSD Chipset, look at the next Sandforce controller.

http://www.anandtech.com/show/3971/sand ... d-60k-iops

I wasant planning on getting another SSD for a while, but each generation leapfrogs the last, and are also becomming cheaper.


Andy

My discomfort with SSDs is the lifespan imposed by the technology...
It's different to say that a product has given lifespan according to use, from saying that average life-time for a product is *.
Can't easily come to terms with a product with such specifications...it will also be much more difficult to sell it and more risky to use in other/older computers...
I still have my first 20GB Maxtor IDE, some 7 years old, running fine in my uncle's computer in much less than favorable conditions...
...an hdd like that, if it presents no problem you could recommend it or use it in an old setup...but a 7 year old SSD, having had similar treatment and use, I'd probably keep it in a drawer or something...

josephclemente wrote:I notice an AMD system was used for this review.

I recently built an AMD system with a SandForce drive and found all sorts of surprises with the AMD south bridge (I am using a SB850).

A lot of people are saying that AMD's Cool'n'Quiet and C1E has a huge impact on speeds. I haven't tried disabling those features, but I do see that HDTune's read speed goes up with each CPU core I load with Prime95. Loading all four cores gives me maximum SSD performance. :O

Also, a lot of forum posts around the web say that the AMD AHCI drivers do not handle TRIM. Others say the newest version does. But the AHCI drivers I saw in Catalyst 10.10 have old dates. Microsoft's default AHCI drivers seem to work very well.

Oh, and specific to the SB850, supposedly disabling the SATA 3.0 6 Gbit/s feature improves speeds further.

FWIW, C&Q/C1E had no effect on our results and the MS AHCI driver was used.

Lifespan of any half way decent SSD is not an issue any more. In fact for most desktop users it is more likely that a HDD will fail during the lifetime of the PC than an SSD will. HDDs are a lot more fragile and sensitive to heat. Laptop drives especially so. I fix PCs for a living and about half the ones which come in have failed HDDs.

Lawrence Lee wrote: FWIW, C&Q/C1E had no effect on our results and the MS AHCI driver was used.

Thanks

andyb wrote:Forget the next generation Intel SSD Chipset, look at the next Sandforce controller.

http://www.anandtech.com/show/3971/sand ... d-60k-iops

I wasant planning on getting another SSD for a while, but each generation leapfrogs the last, and are also becomming cheaper.


Andy

The sf2000 is an enterprise grade controller, will probably be a while before it shows up in consumer ssd's.

Pierre wrote:My discomfort with SSDs is the lifespan imposed by the technology...
It's different to say that a product has given lifespan according to use, from saying that average life-time for a product is *.
Can't easily come to terms with a product with such specifications...it will also be much more difficult to sell it and more risky to use in other/older computers...
I still have my first 20GB Maxtor IDE, some 7 years old, running fine in my uncle's computer in much less than favorable conditions...
...an hdd like that, if it presents no problem you could recommend it or use it in an old setup...but a 7 year old SSD, having had similar treatment and use, I'd probably keep it in a drawer or something...

so you are saying is that HDD are more reliable that SSD ? that is wrong on so many levels it's laughable.

No, I am saying that I don't feel comfortable with a product whose technology itself limits it to a certain lifespan...

I am not saying that between any two drives, mechanical and flash based, the former is going to "live" longer...but that I would feel more reluctant to use a really old SSD, rather than an old HDD, in a person's working pc...

Also, even when an HDD produces errors, most of the time you'll have the opportunity to save your data to another drive even after the faults arise, whereas if a SSD proves dysfunctional, then - from what I read around - you may already have lost data...

So, as I said, if I had an old hdd which gives no errors and a ssd of the same age, if I were to put one in a working desktop, I would probably choose the former...
...ssds have not been around that long to test their endurance in time and how they "behave" as they significantly age...

it's different to think about "what can go wrong" concerning reliability of hdds without having any practical definitive barrier, and "how much time this ssd has got left" counting towards zero hour by every second of operation...
the second makes more uneasy...that's what I said...
If you find it laughable, by all means laugh away...

Pierre wrote:No, I am saying that I don't feel comfortable with a product whose technology itself limits it to a certain lifespan...

HDD technology also limits their lifespan.

A lot of people don't realise this but HDDs are designed with redundant storage because it is inevitable they will have bad blocks. When a read or write fails the block is remapped to one of the spares. You can see how many times this has happened by looking at the SMART data for the drive. It is 99% certain that the HDD in the machine you wrote that post on has remapped at least one bad block in its lifetime.

The number of available spare blocks limits the lifespan of the drive. Once it runs out you start getting hard read/write errors. Don't forget that the drive motor has a limited lifespan too, as some Linux users discovered when Western Digital started making their HDDs power down after 10 seconds and Linux powered them back up a few moments later. They managed to get through an entire lifetime's worth of motor start/stop cycles in a few months that way. Take a look at the MTBF ratings for drives.

At least with SSDs they are unlikely to become totally unreadable. If the motor or the head in a HDD fails you loose everything. If an SSD wears out and cannot accept any more writes you can still at least read your data off it.

nice review. simpler than most, which is good sometimes.

i'm surprised the kingston fared so well since it uses a 'inferior' jmicron controller, i guess they've sorted out the old problems.
i believe the wd uses the same controller.

A lot of people don't realise this but HDDs are designed with redundant storage because it is inevitable they will have bad blocks. When a read or write fails the block is remapped to one of the spares. You can see how many times this has happened by looking at the SMART data for the drive. It is 99% certain that the HDD in the machine you wrote that post on has remapped at least one bad block in its lifetime.

Well I have had 6-10 storage hdds for quite some time (since 2004) now on my machine running 24/7 and have gone through many capacity drives, from 120GB to 2TB, some hdds I've kept long, some I've sold after two years approximately...
Some have had it rough also, being in University high bandwidth network working full time for P2P networks like Dc++, torrents etc...
The only time I've had this issue was with the Seagate's ST31500341ASs which had all sorts of trouble, and with a WD20EARS which presented some 35 reallocated sectors and a couple of bad sectors (irrecoverable errors) in a few hours time (returned the next day for replacement)
From the number of drives I've had (also of different connectivity: IDE, SATA) and their power-on hours and the physical movements they have had to endure (since I've moved around a lot the last 6 years), I'd say that 99% number is highly "optimistic" (pessimistic really)...maybe the other way around would be more accurate...
This is personal experience of course, but we are talking about many drives of different brands and in sustained use, not of a 2-3 drives sample. So I doubt the representative value of that percentage.

The number of available spare blocks limits the lifespan of the drive. Once it runs out you start getting hard read/write errors. Don't forget that the drive motor has a limited lifespan too, as some Linux users discovered when Western Digital started making their HDDs power down after 10 seconds and Linux powered them back up a few moments later. They managed to get through an entire lifetime's worth of motor start/stop cycles in a few months that way. Take a look at the MTBF ratings for drives.

I used to think so too, about the WD drives, there is a long post in this forum as well, but it seems this is questionable...
The thing is that despite the general, theoretical reliability hindrances of hdds, they actually do very well...they have been around for a long time and issues seem to arise when some newer features (and new models) are first introduced...

I have no gain in hdds anyway and because I keep a lot of drives - which can produce noticeable noise - I would love to have SSDs reach high capacity, low cost and become standard...
But writing off hdds on reliability criteria I think is kind of premature...

MoJo wrote: At least with SSDs they are unlikely to become totally unreadable. If the motor or the head in a HDD fails you loose everything. If an SSD wears out and cannot accept any more writes you can still at least read your data off it.

Not true.
Unreadable SSDs are one of the SSD failure modes. Usually it is seen by the SSD reporting memory size in MBs in place of GBs. This means that all the SSD can now see is the controller memory, not the flash memory.
I've had 2 SSDs fail like this from a vendor that is now defunct (hoohrah).

IMO consumer SSDs will still be used as "accelerator" drives for boot and/or gaming oriented RAID apps. As main storage ? NO.
Much better to put main storage on a Gbps connected RAID NAS with "old" technology HDDs.

Why not a single word about TRIM? Why not discuss performance impact or lack thereof?
OTOH, I'm pretty sure that most SPCR readers hunting for a good SSD have already read Anand's reviews and are in the know.

Pierre wrote:The only time I've had this issue was with the Seagate's ST31500341ASs which had all sorts of trouble, and with a WD20EARS which presented some 35 reallocated sectors and a couple of bad sectors (irrecoverable errors) in a few hours time (returned the next day for replacement)
From the number of drives I've had (also of different connectivity: IDE, SATA) and their power-on hours and the physical movements they have had to endure (since I've moved around a lot the last 6 years), I'd say that 99% number is highly "optimistic" (pessimistic really)...maybe the other way around would be more accurate...

You've been lucky. I've had many failed hard drives and many hard drives reporting reallocated sectors... but nowhere near 99%. That's a preposterous figure. If 99% of drives reallocate sectors they're not telling (which wouldn't surprise me). If you want hard data, check the famous Google paper.
In my experience, hard drives have gotten more reliable over the years if you take the long view. And I agree SDDs are unproven for long-term reliability under intensive use.
In a mobile device that might be dropped to the floor or something, yeah I figure you'd want a SSD. I've seen enough dead 2.5'' drives in laptops and portable USB enclosures...

Micron C300 64GB is missing from this review. With SATA 3 its currently the leader in performance (if you have SATA3). Its also priced relativly low.

In a couple of months we should be seeing some next gen SSDs that are x2+ faster. Mentioning this couldn't hurt this review either.

Pierre wrote:Well I have had 6-10 storage hdds for quite some time (since 2004) now on my machine running 24/7 and have gone through many capacity drives, from 120GB to 2TB, some hdds I've kept long, some I've sold after two years approximately...

I have had lots of drives over the years and never once had a single failure. Maybe I'm lucky. Of course at work I only ever see broken PCs so it's no wonder so many of them have failed HDDs.

For some statistics rather than anecdotal evidence I suggest reading the study Google did. They have hundreds of thousands of drives in use 24/7. The conclusions are very interesting, the bottom line being that after a couple of years the failure rate goes up exponentially. As I pointed out the motors/servos have a limited lifespan and the drives eventually run out of spare blocks.

I have no gain in hdds anyway and because I keep a lot of drives - which can produce noticeable noise - I would love to have SSDs reach high capacity, low cost and become standard...
But writing off hdds on reliability criteria I think is kind of premature...

For mass storage they can't be beat, but if you want a small and reliable drive then SSDs are the way to go. One area they have been popular for years in is embedded systems like routers. Most commercial routers are just Linux based system-on-chip designs with small flash memory drives. We also built quite a few pfSense based routers for clients and the ones with SSDs (actually just Compact Flash cards in IDE adapters) never give us any trouble, while the ones with HDDs often stuffer HDD failure after a couple of years.

For some statistics rather than anecdotal evidence I suggest reading the study Google did. They have hundreds of thousands of drives in use 24/7. The conclusions are very interesting, the bottom line being that after a couple of years the failure rate goes up exponentially.

It has been some two full years since I read that report (mainly interested about the impact of temperature on failure rate), but I remember results being rather different...I took another look and it seems I remembered the results quite well.
The findings actually showed that 2 year and 3 year old drives have higher chances of failure than 4 year and 5 year old drives...

Also, the document suggested no strong correlations between intensity of use, which would be related wear and tear of mechanical parts, or between relevant SMART data and higher failure rates...
...disk surface defects were actually found to be more relevant to failure...

So I would suggest giving it another read.

Pierre wrote:The findings actually showed that 2 year and 3 year old drives have higher chances of failure than 4 year and 5 year old drives...

Yes. I don't see how that contradicts what I said.

Also, the document suggested no strong correlations between intensity of use, which would be related wear and tear of mechanical parts, or between relevant SMART data and higher failure rates...
...disk surface defects were actually found to be more relevant to failure...

I don't think you understand how hard drives work or what SMART does.

When the disk is on it is rotating all the time. The motor is running constantly. It doesn't matter if the head is moving or reading/writing data, the drive still spins. In Google's case the drives never stop so start/stop cycles were not included in the data they used. Equally surface defects are not affected by activity, beyond being accessed from time to time which for a drive that is nearly 100% full of randomly accessed data (like Google's) definitely fits that category.

SMART keeps track of how many surface defects it sees. What constitutes a SMART failure is defined by the vendor and tends to be somewhat conservative, but if you look at the raw numbers for things like remapped blocks the pattern is fairly clear. The reason SMART typically fails to warn you before a drive fails is simply because its counters do not reach the limits set by the manufacturer. At work we use PC Check which is a bit more realistic with the numbers and which also runs the SMART long test. Most manufacturer's tools can run the long test and it often fails due to scanning the entire surface of the drive and picking up new defects which then push the remapped block count over the limit.

Keep in mind that drives remap blocks even when they managed to read the data back but had to recover it with error correction. The drive can tell when the block is failing so even if you don't get a read or write error it doesn't mean that remapping didn't happen. The only way to know is by looking at SMART counters.

Just a couple of points regarding HDDs.

1) Keeping a HDD spinning constantly probably causes less wear and tear than turning it on/off once or more daily, especially in PCs that are never moved. With FDB bearings, the wear on the bearing is minute. Mechanical devices undergo thermal changes when turned off/on -- heating and cooling, and the expansion/contraction that causes; over time, this is more likely to cause problems than 24/7 operation.

2) I think the absence of the above skews the google study, making it applicable mostly to HDDs that are left on 24/7. AFAIK, no one has yet done a similarly large-scale, long term study of HDDs in PCs that are turned on/off.

MoJo, the bit of analysis Pierre fetched very much contradicts a naive reading what you said! Exponential usually refers to something like failures rates of 2% the first year, 4% the second, 8% the thrid, 16% the fourth and so on.

The reason SMART usually fails to warn is that most failures can not be predicted, whatever thresholds you use. Long tests have never revealed a problem with any of my drives. It only failed on drives which had obvious signs of failure. I mainly use them to confirm problems.
I have a drive which reports about 45K hours of usage and no errors whatsoever. But it could fail anytime.

The Google study says that only 9% of their drives report any reallocations which goes to show the 99% figure is indeed on the high side...

Pierre wrote:Also, the document suggested no strong correlations between ... relevant SMART data and higher failure rates...
...disk surface defects were actually found to be more relevant to failure...

I don't know what you call relevant but I went back to the paper and I see that remembered right: they've found strong correlations between SMART data and failures. The only SMART data which is unambiguously about the surface as far as I know is the reallocation count which had a fairly strong correlation with failures. But others parameters had as good (if not better) correlations.