Will SSD replace Hard Drive?

[Blain]

Originally posted by: AshPhoenix
I think SSDs will be quite affordable in 2010, of course they will be more expensive than HDDs but the benefits of SSDs are worth it.

That's totally subjective.

 

[jedisolo]

I've been using SSD's for over a year now, I think they are worth it.

 

[Idontcare]

Originally posted by: Thraxen

Originally posted by: Idontcare
So my expectation is that once the small capacity spindle drive market is consumed by cheap low capacity SSD's we will reach a point where we find the >1TB drives (or >2TB at that time) could actually start to rise in cost because of the lower volume higher manufacturing costs that will be associated with them. And as the $-volume goes, so too does the R&D investments for the next iteration (as was the case with CRT's) and eventually investments into creating new spindle drives (regardless of capacity) will simply cease.

I just hope the transition doesn't take too long. I'm not against SSDs as I firmly believe that solid state drives are the future, I've just gotten used to the luxury of enourmous storage capacities and do not want to have to backpeddle in storage capacity and pay ridiculous prices at the same time. The trade offs just aren't worth it for something like a home file server since the benefits are minimal at best.

Flash doubles in capacity (density) about every 12-16 months for equivalent cost (cost per bit declines by ~50%), it won't take to many doublings before 1TB drives cost as little as today's 80GB drives.

Originally posted by: Blain

Originally posted by: AshPhoenix
I think SSDs will be quite affordable in 2010, of course they will be more expensive than HDDs but the benefits of SSDs are worth it.

That's totally subjective.

So too the transition from CRT's to LCD's, single-core CPU's to dual-core CPU's, wired networks to wireless networks, etc.

The motivation for users to adopt changing technology lies in the subjective part of the experience. HDTV and Blu-ray are going thru this transition right now as well.

 

[ShawnD1]

Originally posted by: WaitingForNehalem
I'm still worried about SSD's reliability. I read that the cells degrade. Then there is SLC vs MLC. I read that once SLC's are filled they again degrade in performance. Spindle drives are proven and have been around a long time.

Spindles are notoriously unreliable. There's a reason we spend so much money on backup hard drives and Raid 1 or Raid 5. I've only ever had 2 hard drives fail on me since 1992, but then again I've only owned about 10 hard drives in that time. I've also owned at least 10 USB flash drives and none of them have ever failed. It should also be noted that lots of people carry flash drives to school/university and they work until you do something ridiculous to destroy them. A spindle drive isn't so resilient; you can't drop it and expect it to work perfectly afterward.

As for the idea that flash has a limited number of writes, that is absolutely ridiculous. I've had a 4gb flash drive acting as a Readyboost drive for the past 2 years in a computer that never shuts off. The performance manager says it's constantly writing more data than any other process, and yet the drive still works. If you manage to somehow reach the write limit on a SSD, your computer is the exception. Most people will never hit that limit.

Back to the topic. No SSD will not replace spindle (in the next few years). Wikipedia says USB flash drives were first sold around 2000 or so. It has been 8+ years since, and the prices are still crazy expensive. A 32gb SATA SSD costs as much as a 1000GB spindle drive. USB is a bit cheaper because it's not as good, but it's still insanely expensive compared to spindle. SSD needs to make some major breakthrough before it can be compared to spindle price/gb.

edit:
I just thought of something. USB flash drives are a lot smaller than the expensive SATA flash drives, but they still hold the same capacity. Couldn't a company just parallel a bunch of USB-type flash drives and make a giant 500gb flash drive? It couldn't be any bigger than a regular 3.5" drive.

 

[Blain]

[Edit]

 

[Idontcare]

Originally posted by: ShawnD1
As for the idea that flash has a limited number of writes, that is absolutely ridiculous. I've had a 4gb flash drive acting as a Readyboost drive for the past 2 years in a computer that never shuts off. The performance manager says it's constantly writing more data than any other process, and yet the drive still works. If you manage to somehow reach the write limit on a SSD, your computer is the exception. Most people will never hit that limit.

Your post confuses me. First you argue that the idea of MLC/SLC flash technology having limited cell life is "absolutely ridiculous"...then you go on to contradict yourself by acknowledging they do have limited life but based on your quite limited sample size and experience you are willing to conclude that it is of no practical consequence to virtually everyone else on the planet save for a handful of exceptions...

Originally posted by: ShawnD1
Back to the topic. No SSD will not replace spindle (in the next few years). Wikipedia says USB flash drives were first sold around 2000 or so. It has been 8+ years since, and the prices are still crazy expensive. A 32gb SATA SSD costs as much as a 1000GB spindle drive. USB is a bit cheaper because it's not as good, but it's still insanely expensive compared to spindle. SSD needs to make some major breakthrough before it can be compared to spindle price/gb.

USB thumbdrives started as floppy-drive replacements at floppy-disk capacities (1-8 MB) and have grown since then over 1000x to tens of GB's. And costs have decreased.

Why is this? Moore's law. It's not just for CPU's.

Flash density doubles every 12-16 months (or faster, it's far more aggressive doubling rate than logic IC's), costs decrease ~50% per bit every 12-16months, and performance increases at about half that rate.

And what you see today in terms of $/GB pricing of SSD's is the very early phase of a new product segment gaining critical mass while competing with an entrenched existing production capacity. As SSD's gain mass (market share) the production investments into existing spindle-drive facilities will dwindle (as it did for CRT's as LCD's took over) and before you know it that 1000GB drive won't be dropping in price any longer as the volumes decline and the technology investments stop being invested.

It really is merely a matter of a few years before the transition is in full-effect.

Originally posted by: ShawnD1
edit:
I just thought of something. USB flash drives are a lot smaller than the expensive SATA flash drives, but they still hold the same capacity. Couldn't a company just parallel a bunch of USB-type flash drives and make a giant 500gb flash drive? It couldn't be any bigger than a regular 3.5" drive.

That's basically what the first generation of really ultra cheap SSD products were like, albeit in a consumer acceptable package format.

http://www.newegg.com/Product/...egory=636&N=2003240636

Nowadays you'd be hard-pressed to bundle a bunch of thumbdrives and still come out cheaper $/GB than the really cheap and cruddy (but still better than USB) SSD's on newegg. That window has closed, but a hobbyist would have fun with the project nonetheless.

 

[Lonyo]

It'll take a while, like 10 years, for SSD's to phase out mechanical drives completely I expect, but in the meantime they will certainly start to erode more market share than they already have.
You're still looking at a minimum 5 years at a 50% reduction/yr for SSD prices to match current mechanical drive prices, but then SSD's will kill other markets first.

Initially 1.8" drives will die out (they are already very expensive/GB and getting replaced by flash), then 2.5" drives will get replaced (again more expensive/GB than 3.5" drives, and SSD's offer more speed improvement over them than mechanicals, and capacity issues aren't so pressing usually), then after 1.8" and 2.5" drives get phased out, 3.5" drives will be already nearing the end and probably get wiped out completely.
SSD's have more early market applications outside desktop computers though, replacing high cost/GB drives found elsewhere (which they are already slowly doing), and they will certainly be going within 3 years, but I think mechanical 3.5" drives will last longer unless SSD's make some kind of breakthrough to reduce costs faster than they are doing.
3.5" drives are GB's/$ while SSD's are still $'s/GB, so we have a long way to go to replace mechanicals in cheap machines, even with SSD speed advantages.

 

[cbn]

I think Windows 7 with its "trim" feature and "automatic alignment" will be a major reason for mass adoption of SSD. In fact, hassling with alignment and lack of "trim" is a major reason I am not tempted to consider one of those OCZ "vertex" 30GB drives for $130.

 

[taltamir]

Originally posted by: WaitingForNehalem
I'm still worried about SSD's reliability. I read that the cells degrade. Then there is SLC vs MLC. I read that once SLC's are filled they again degrade in performance. Spindle drives are proven and have been around a long time.

Pure FUD. don't beleive any of this shit.
The problems with SSD is that they are expensive, and that some crap models are really slow in random writes making them stutter, thats it really. (aside from the occasional firmware bug, but those exist in spindle drives too)

 

[taltamir]

Originally posted by: Thraxen

Originally posted by: Idontcare
So my expectation is that once the small capacity spindle drive market is consumed by cheap low capacity SSD's we will reach a point where we find the >1TB drives (or >2TB at that time) could actually start to rise in cost because of the lower volume higher manufacturing costs that will be associated with them. And as the $-volume goes, so too does the R&D investments for the next iteration (as was the case with CRT's) and eventually investments into creating new spindle drives (regardless of capacity) will simply cease.

I just hope the transition doesn't take too long. I'm not against SSDs as I firmly believe that solid state drives are the future, I've just gotten used to the luxury of enourmous storage capacities and do not want to have to backpeddle in storage capacity and pay ridiculous prices at the same time. The trade offs just aren't worth it for something like a home file server since the benefits are minimal at best.

thats when you use an SSD for your OS and Games, and 2TB spindle drives for your other junk.

 

[ShawnD1]

Originally posted by: Idontcare

Originally posted by: ShawnD1
As for the idea that flash has a limited number of writes, that is absolutely ridiculous. I've had a 4gb flash drive acting as a Readyboost drive for the past 2 years in a computer that never shuts off. The performance manager says it's constantly writing more data than any other process, and yet the drive still works. If you manage to somehow reach the write limit on a SSD, your computer is the exception. Most people will never hit that limit.

Your post confuses me. First you argue that the idea of MLC/SLC flash technology having limited cell life is "absolutely ridiculous"...then you go on to contradict yourself by acknowledging they do have limited life but based on your quite limited sample size and experience you are willing to conclude that it is of no practical consequence to virtually everyone else on the planet save for a handful of exceptions...

Have you ever met someone who has seen a flash drive fail? Has anyone ever reported that readyboost suddenly stops working because a drive hits the write limit? Estimates on these things are in the ballpark of 100k writes. For something like a 32gb flash device, that's roughly 320TB of writes. If you were writing to that drive at a constant 30mb/s (the write speed of many USB drives), it would take 106666666 seconds to hit that. That's about 3.5 years of 24/7 nonstop writing to disk. If you happen to be the one person in the world who really does write to disk at high speeds 24/7 (not including servers), then maybe a Raptor drive is a better solution. Even then, I wouldn't expect a standard hard drive to do any better when operating under those conditions.

Originally posted by: ShawnD1
Back to the topic. No SSD will not replace spindle (in the next few years). Wikipedia says USB flash drives were first sold around 2000 or so. It has been 8+ years since, and the prices are still crazy expensive. A 32gb SATA SSD costs as much as a 1000GB spindle drive. USB is a bit cheaper because it's not as good, but it's still insanely expensive compared to spindle. SSD needs to make some major breakthrough before it can be compared to spindle price/gb.

USB thumbdrives started as floppy-drive replacements at floppy-disk capacities (1-8 MB) and have grown since then over 1000x to tens of GB's. And costs have decreased.

Why is this? Moore's law. It's not just for CPU's.

Flash density doubles every 12-16 months (or faster, it's far more aggressive doubling rate than logic IC's), costs decrease ~50% per bit every 12-16months, and performance increases at about half that rate.

And what you see today in terms of $/GB pricing of SSD's is the very early phase of a new product segment gaining critical mass while competing with an entrenched existing production capacity. As SSD's gain mass (market share) the production investments into existing spindle-drive facilities will dwindle (as it did for CRT's as LCD's took over) and before you know it that 1000GB drive won't be dropping in price any longer as the volumes decline and the technology investments stop being invested.

I should clarify that when I said USB flash was sold in 2000, I was using that as a benchmark for when flash was a practical storage alternative to floppy disks. Flash as a whole is nothing new. My Nintendo used flash cartridges.

Moore's Law also applies to normal hard drives. The 1TB external hard drive I bought on newegg 2 days ago cost less than the 400gb drive I bought in January 2008. Flash will keep getting cheaper, but so will standard hard drives. Some people have expressed concerns about magnetic drives reaching some kind of theoretical limit when it comes to the physical size of each bit, but why would anyone think that limit doesn't apply to flash as well? When Moore's Law stops working on normal hard drives, it will also stop working on flash drives.

I believe you are wrong when you say/imply SSD is in the early phase of basically making spindle drives obsolete. Flash has existed since 1980 but it has never managed to replace spindle drives. Sure all of my Nintendo and Sega Genesis games used flash and all of them are of such high quality that they still work today, but flash is rarely a feasible solution for storing large amounts of data. Flash's main job has always been storing relatively small amounts of data while consuming very little power and being very reliable. While I certainly do think flash will gain market share, I don't believe for a second that we will abandon spindle drives. I've heard people say the same basic thing about high performance Raptor drives taking over because they too are a lot faster and better than normal drives, but that never happened, nor will it happen. There's always a certain point where we say "this is fast enough" and we prefer size and price over speed or reliability.

Thats when you use an SSD for your OS and Games, and 2TB spindle drives for your other junk.

This is exactly what I'm thinking. I've already moved most of my \Program Files\ to flash at a reasonable price since Program Files is a very small portion of the computer's hard drive space. The size of software doesn't seem to get bigger at the same rate as everything else, so SSD is quickly becoming a practical solution for OS and installed applications.

 

[taltamir]

there is a difference between "limited lifespan" and "degrading performance"... the idea that they DEGRADE and become worse and slower over time is simply false, its FUD.
But they will run out of writes eventually and the drive becomes read only drive. Which will take an obscene amount of writes, and is really quite harmless.

 

[ShawnD1]

Originally posted by: taltamir
there is a difference between "limited lifespan" and "degrading performance"... the idea that they DEGRADE and become worse and slower over time is simply false, its FUD.
But they will run out of writes eventually and the drive becomes read only drive. Which will take an obscene amount of writes, and is really quite harmless.

Actually the degrading performance thing is real. Anand covered it in one of the more recent articles under the Storage heading.

http://www.anandtech.com/stora...howdoc.aspx?i=3531&p=8

Now let?s think about what?s just happened. As far as the OS is concerned we needed to write 12KB of data and it got written. Our SSD controller knows what really transpired however. In order to write that 12KB of data we had to first read 12KB then write an entire block, or 20KB.

Our [hypothetical] SSD is quite slow, it can only write at 1KB/s and read at 2KB/s. Writing 12KB should have taken 12 seconds but since we had to read 12KB and then write 20KB the whole operation now took 26 seconds.

To the end user it would look like our write speed dropped from 1KB/s to 0.46KB/s, since it took us 26 seconds to write 12KB.

Are things starting to make sense now? This is why the Intel X25-M and other SSDs get slower the more you use them, and it?s also why the write speeds drop the most while the read speeds stay about the same. When writing to an empty page the SSD can write very quickly, but when writing to a page that already has data in it there?s additional overhead that must be dealt with thus reducing the write speeds

edit: the performance "degrades" to a speed that is still very fast. The drive doesn't turn into something unusable.

 

[taltamir]

that is incorrect, that is NOT degrading cell performance, that is STEADY STATE, there is a ONE TIME reduction from "clean" to steady state level, TRIM COMPLETELY ELIMINATES IT ("restoring" the "degraded" drive into "clean" state), it is not an exact value though as it will fluctuate going up and down depending on your writes... all it is really is fragmentation... the CELLS however are not degraded in any way shape or form from steady state.

It is clearly and carefully explained in the article should you read it in full instead of misquoting snippets out of context and than draw erroneous conclusions from them.

so let me revise my statement:

there is a difference between "limited lifespan", "fragmentation related slowdown" and "degrading performance"... the idea that they DEGRADE and become worse and slower over time is simply false, its FUD.

But fragmentation will cause a limited slowdown (it will not get worse "indefinitely) until the TRIM command allows the controller to completely eliminate the issue (regular defragmenting software is incompatible with SSD and will WORSEN the situation!)

AND they will run out of writes eventually and the drive becomes read only drive. Which will take an obscene amount of writes, and is really quite harmless; especially compared to the motor eventually failing on a regular drive, making it unwritable AND unreadable.

 

[ShawnD1]

Originally posted by: taltamir
It is clearly and carefully explained in the article should you read it in full instead of misquoting snippets out of context and than draw erroneous conclusions from them.

Are you absolutely certain you read that article?

http://www.anandtech.com/stora...owdoc.aspx?i=3531&p=10

While the TRIM command will alleviate the problem, it won?t eliminate it. The TRIM command can?t be invoked when you?re simply overwriting a file, for example when you save changes to a document. In those situations you?ll still have to pay the performance penalty.

Every controller manufacturer I?ve talked to intends on supporting TRIM whenever there?s an OS that takes advantage of it. The big unknown is whether or not current drives will be firmware-upgradeable to supporting TRIM as no manufacturer has a clear firmware upgrade strategy at this point.

I expect that whenever Windows 7 supports TRIM we?ll see a new generation of drives with support for the command. Whether or not existing drives will be upgraded remains to be seen, but I?d highly encourage it.

To the manufacturers making these drives: your customers buying them today at exorbitant prices deserve your utmost support. If it?s possible to enable TRIM on existing hardware, you owe it to them to offer the upgrade. Their gratitude would most likely be expressed by continuing to purchase SSDs and encouraging others to do so as well. Upset them, and you?ll simply be delaying the migration to solid state storage.

So you're telling this guy he should ignore the performance degradation because an unreleased OS that will cost at least $200 will support a command that is not yet supported by any hard drive's firmware? Seriously. Come on now.

 

[taltamir]

Again with the misquoting and misunderstanding.

there is no "problem" for the end users because every worthwhile benchmarking site, like anandtech and pcper benchmarks the drives in steady state condition! There is a "problem" form the ENGINEER's standpoint that you COULD get better speed if you fixed it, but its not a "problem" for end users, their drives are not going to "degrade" beyond 1. the first few weeks. 2. below the performance promised by manufacturers and reviewers.

Secondly, this is not a "degrading cells", i am sorry it just isn't. Calling this degrading cells is like saying "Mama say alligators are so ornery cause they got all them teeth and no toothbrush"

thrid, companies like OCZ already have a beta tool that will enable trim functionality in winXP and vista... its working, but i wouldn't use it yet because it is beta and i wouldn't risk my data.

And finally, cell write speed will not degrade, so fragmentation will fluctuate around a set amount getting worse and then getting BETTER, without much overall change. (after the initial "degradation" in first few weeks), so there is no "cells lose their speeds getting progressively slower until they don't operate at all"

 

[magreen]

@ShawnD1: taltamir is right. You're misunderstanding the article that you're quoting. The part about trim not fixing the overwriting issue, and you still having to pay a performance penalty in that case, is not a degradation issue. It's just the nature of SSDs. It has nothing to do with them degrading over time. And it's also just one small case.

And as taltamir quite ably explained, even without the trim command, the SSDs currently on the market do not degrade to worse than their steady state performance. And AT tested them in their steady state and found that they were still outrageously fast.