Current state of SSD's

Current state of SSD's

My niece thinks she needs a new laptop. Hers' is all of 4 years old and
she imagines its running slowly. (Or it is but could be fiixed.)

She wasts a solid state drive. Last I heard, you could only write to
them so many times before they started forgetting what they knew.

Are they still unreliable in this or some other way?

There was an interesting discussion recently
at Slashdot. I'd just been thinking that maybe
SSDs are now mature enough to be a sensible
investment. But there seemed to be a lot of
people who've had them die suddenly.

Mayayana wrote:
There was an interesting discussion recently
at Slashdot. I'd just been thinking that maybe
SSDs are now mature enough to be a sensible
investment. But there seemed to be a lot of
people who've had them die suddenly.

If you happen to come across some articles on that, please post some links
when you get a chance. It would be interesting to know more about that.

micky wrote:
Current state of SSD's

My niece thinks she needs a new laptop. Hers' is all of 4 years old and
she imagines its running slowly. (Or it is but could be fiixed.)

She wasts a solid state drive. Last I heard, you could only write to
them so many times before they started forgetting what they knew.

Are they still unreliable in this or some other way?

There are three kinds of flash chips in SSDs.

SLC - single level cell, the best, used in low capacity
enterprise drives, or as a cache in an SSHD drive.
Hardly any consumer drives would use these.

MLC - stores two bits worth of data per the same storage cell.
This is the current workhorse of the consumer market
for SSDs. These are OK.

TLC - stores three bits worth of data per storage cell.
These have exhibited degradation when data is left
"cold" on a section of drive for periods of more than
three months. I cannot recommend a drive using these
chips, to any one. Even if firmware is shipped, to
automatically refresh "cold" data, this is a poor
way to run flash memories (a wasteful operating mode).

If we were to see "slow readout syndrome" start to show
up on MLC drives too, then I would not be able to recommend
them either.

For the time being, MLC still seems safe. Check the customer
reviews to see if anyone has experienced less than satisfactory
performance.

At least one web site, has been testing SSD drives to
exhaustion. And some of them have exceeded their
expected life by a wide margin. Too bad there
wasn't a way to predict which ones would behave
like that. As any testing that has taken a few years
to complete, the drives today would use entirely
different chips. And we wouldn't be able to buy exactly
the same SSD drives any more.

At least one company was caught changing chip types, without
changing the SSD drive model number. So the SSD companies know
they'd receive some pretty bad PR feedback, if they do this
again. At the very least, a product using MLC today,
with model number 12345678, should also contain MLC
chips tomorrow. Chips with the same geometry (like
20nm gate dimensions say). If they changed from MLC to
TLC, without changing the model number to signify such
a production change, then expect them to get blasted
by the feedback.

The SMART statistics on SSDs offer a different set of
metrics than on hard drives. You will need a more modern
up-to-date utility to read SSD SMART. In particular, monitor
the "remaining drive life", a percentage figure. You see,
some manufacturers have designed bizarre behavior into
the SSD drives, at the end of their service life. At least
one company, turns off both writes *and reads* when the
percentage life hits zero. They claim this prevents people
from getting unreliable data, but for people who do not
do backups, their only copy of a file could be on
that drive. So check the end-of-life policy, or do
backups daily if you're too lazy to do the leg work.

Both hard drives and SSD drives, are at the mercy of
firmware bugs. Even hard drives have been bricked by
inappropriate behavior with respect to controlling
data structures. So in fact, "instant death syndrome"
can happen with both SSDs and HDDs.

One difference is, the HDD is more sensitive to handling,
so if dropped, that's an area which the SSD can outperform
the hard drive. It takes a drop a little bit better than
a running hard drive would, as nothing is spinning, and
there is no microscopic flying height to worry about.

Paul

On 27/05/2015 6:32 PM, micky wrote:
Current state of SSD's

My niece thinks she needs a new laptop. Hers' is all of 4 years old and
she imagines its running slowly. (Or it is but could be fiixed.)

She wasts a solid state drive. Last I heard, you could only write to
them so many times before they started forgetting what they knew.

Are they still unreliable in this or some other way?

I'd say they're safe and reliable. I got two of them, one in my desktop
and one in my laptop. In actual fact, the larger these SSD get, the more
reliable they get. The reason is that there is a lot of redundancy built
into them, and the larger the memory pool the larger the redundancy pool.

I would suggest not getting any SSD that's less than 120GB at least.

Yousuf Khan

Bill in Co wrote:
Mayayana wrote:
There was an interesting discussion recently
at Slashdot. I'd just been thinking that maybe
SSDs are now mature enough to be a sensible
investment. But there seemed to be a lot of
people who've had them die suddenly.

If you happen to come across some articles on that, please post some links
when you get a chance. It would be interesting to know more about that.

SSD brickage due to firmware issues, was an issue in the
past. But even HDD have had firmware issues of a similar
type, with similar results (sudden death syndrome). The
only difference seems to be, we cut the HDD guys more
slack when they foul up.

As long as you check the Newegg customer reviews for the
SSD you want to buy, there are enough reviews to get a
clear picture of product quality, hen you have some idea
whether it's going to brick on you within the first
month of usage.

Paul

On 27/05/2015 23:32, micky wrote:
Current state of SSD's

My niece thinks she needs a new laptop. Hers' is all of 4 years old and
she imagines its running slowly. (Or it is but could be fiixed.)

She wasts a solid state drive. Last I heard, you could only write to
them so many times before they started forgetting what they knew.

Are they still unreliable in this or some other way?



If its running slow now, maybe it just needs all the crap cleaned off. I
have done too many laptops which are running like a snail at first, but
be time I have cleaned them out they are more or less just like new again.

Anyways an SSD can only run at the speed of what SATA speed the laptop
is now . That is if it's SATA 2 now then it will always be sata 2.

| If you happen to come across some articles on that, please post some links
| when you get a chance. It would be interesting to know more about that.
|

This was the discussion:

http://hardware.slashdot.org/story/1...e-powered-down


The topic was actually about false rumors of SSD
failure, but the posters brought up other issues.

I got the impression that Intel and Samsung made
the best SSDs. But then Samsung had some kind of
bad batch recently. I've had a hard time finding
consistent information. The specs also seem to
vary more than hard disk specs do. And they really
haven't been around long enough to support claims
of long life.
My own hesitation is twofold:

1) Given the relatively high cost, I really don't care
about booting my computer in 5 seconds vs 30
seconds.

2) If there's any risk of seemingly arbitrary, instant
failure that's a deal-breaker for me, since I don't
really need the speed. I've never actually had a hard
disk fail. When they do there's generally some warning.
I help friends with their computers and have a couple
I'm currently trying to give away. In most cases those
are XP PCs with hard disks more than 10 years old.

Mayayana wrote:
| If you happen to come across some articles on that, please post some links
| when you get a chance. It would be interesting to know more about that.
|

This was the discussion:

http://hardware.slashdot.org/story/1...e-powered-down


The topic was actually about false rumors of SSD
failure, but the posters brought up other issues.

I got the impression that Intel and Samsung made
the best SSDs. But then Samsung had some kind of
bad batch recently. I've had a hard time finding
consistent information. The specs also seem to
vary more than hard disk specs do. And they really
haven't been around long enough to support claims
of long life.
My own hesitation is twofold:

1) Given the relatively high cost, I really don't care
about booting my computer in 5 seconds vs 30
seconds.

2) If there's any risk of seemingly arbitrary, instant
failure that's a deal-breaker for me, since I don't
really need the speed. I've never actually had a hard
disk fail. When they do there's generally some warning.
I help friends with their computers and have a couple
I'm currently trying to give away. In most cases those
are XP PCs with hard disks more than 10 years old.

It turned out the false alarm on data life, was
for drives that had already hit the "0% life left"
limit. Then, the drives were stored at elevated
temperature, and then the data lasted for a week.
On some drives, the drive ceases to communicate
when the SMART statistic says "0% life remaining",
so you cannot even read the data and discover it is
corrupted. I didn't catch that detail, when
reading the original bad looking documentation.

If your drive has life left, it should not
perform that poorly. And nobody should be running
an SSD with valuable data, all the way down to
0%, just to find out what happens (without having
a backup).

Moral of that story, is research the prospective
drive design, to see what the end-of-life policy
is (read-only or brick, are the two policies).

If you own a single SSD, then the improvement is
the zero seek time parameter.

If you own multiple SSD, then the enhanced transfer
rate is a boon. But only one poster in the newsgroup,
runs an all-SSD operation. The rest of us are too poor
for that sort of thing. So the chances of you doing
SSD-to-SSD transfers all day long, are pretty limited.
Sure, you can go out right now, buy two of them
and make a liar out of me, but practical people
normally only buy one, and don't do fancy stuff
with it. So they're doing SSD to HDD or HDD to SSD
transfers, with the HDD limiting the transfer.

So seek time is the improvement. When your AV decides
to scan the entire Windows folder, that will go a lot
faster due to reduced seek time. And if you do Windows
Search, without generating an index first, the SSD
makes it possible to do such searches in real time.

Your Windows OS is a bottleneck to the improvements.
While the SSD can have as high as 100,000 IO per
second, the OS refuses to deliver them that fast.
A much much lower performance number is the result.
For example, Windows has interrupt mitigation, which
may cap the response from a subsystem. But I have a
feeling some other file system code, chops even more
performance off the thing. Not sure what that is.

I've done tests with RAMDisks here, and spotted these
issues. Desktop Windows gets in the way of high I/O, no
matter what your best intentions might have been.
So if you're disappointed with your new purchase,
that's why.

Paul

| It turned out the false alarm on data life, was
| for drives that had already hit the "0% life left"
| limit. Then, the drives were stored at elevated
| temperature, and then the data lasted for a week.
| On some drives, the drive ceases to communicate
| when the SMART statistic says "0% life remaining",
| so you cannot even read the data and discover it is
| corrupted. I didn't catch that detail, when
| reading the original bad looking documentation.
|
| If your drive has life left, it should not
| perform that poorly.

Yes. I did mention that it was about a false
alarm. What I thought was interesting, though,
was the discussion it sparked. Though there seem
to be a lot of strong opinions on both sides of
the issue.