Are solid-state HDDs much faster than mechanicals ones?

K

Ken Blake

Are there any downsides (other than the cost)?

Much faster? Yes.

Downsides? They are reputed to wear out faster. But in my view, they
have been available for too short a time, and nobody knows that for
sure. Time will tell whether or not it's true.
 
P

Peter Jason

Ken Blake said:
Much faster? Yes.

Downsides? They are reputed to wear out faster. But in my
view, they
have been available for too short a time, and nobody knows
that for
sure. Time will tell whether or not it's true.
Thanks, I'll probably buy a small one just to test is out.
Do they need defragging occasionally?
 
P

Paul

Peter said:
Are there any downsides (other than the cost)?
They have limited write cycles, which means you
can't be entirely careless with them.

The device has an internal controller, and one of it's
jobs is "wear leveling". There is a level of indirection,
between the external sector address, and which block of
NAND cells responds to that address. By moving data around,
on each write operation, the controller keeps track of how
many times each block has been written, so that they all
"wear out at the same time". Then, the lifetime of the
device is measured in "gigabytes of writes", where you
multiply the device capacity, by the number of writes
allows. (Say 128GB times 3000 writes max, or 300TB of
writes. SLC NAND may have ten times the write cycles of
MLC NAND, but is much more expensive. The manufacturer
will smooth over this statement by saying "you can write
1GB per day for the next 8 years" or some such, to make it
sound more reasonable.)

Windows 7 is one of the first OSes, where the designers
provide some level of support for SSDs. For example, if
you plugged an SSD into WinXP, the OS would let you run
the defragmenter. In Windows 7, the OS should refuse to
do that (because it doesn't really help). The SATA SSD
has a sub 100 microsecond access time, so it doesn't
really matter if it is moderately fragmented. There is
next to no penalty for "moving the heads". So rather
than the lukewarm job Windows 7 does on a regular hard
drive, it shouldn't do anything to an SSD. That helps
extend the useful life of the SSD.

Windows 7, in the built-in AHCI driver ("MSAHCI"), has
support for TRIM. That is part of maintaining the SSD,
with respect to files that have just been deleted. In
other situations, you may need to run a utility program,
to achieve the same ends. (Some devices don't need TRIM,
because they have internal functions that emulate it.)

http://en.wikipedia.org/wiki/TRIM

On previous OSes, there are innocent sounding functions,
such as updating the "last accessed" field for files,
each time you examine the file. That is entirely unnecessary,
in terms of maintaining your personal information, and
was only done on regular hard drives, because there is
virtually no penalty associated with writing over and
over again. With the low limit on total write cycles per
NAND block, it pays to revisit design decisions like that,
and disable any features that don't add value. There
are two choices for a designer, with respect to "last
accessed". Either update all values, just as the OS
is shutting down (so you only waste effort on it, once
a day, or once a session). Or eliminate the function altogether.
Useful fields, such as "creation date" and "last modified",
continue to be supported, but stupid ones such as
"last accessed" can be turned off.

You can find acres and acres of advice on setting up
and maintaining SSDs. Much of this will be taken care
of, by Windows 7, so not everything you read in such articles
should be acted upon. But there are still some interesting
things you should know about. This is a forum maintained by
one manufacturer, with helpful hints for extending the
usable lifetime of the SSD.

http://www.ocztechnologyforum.com/forum/showthread.php?43460-Making-XP-pro-SSD-friendly

http://www.ocztechnologyforum.com/f...h-stuttering-and-increases-drive-working-life.

Since I don't own one, I don't have any more bookmarks
than that :)

Are they much faster ? Absolutely.

http://www.newegg.com/Product/Product.aspx?Item=N82E16820148363

Interface Type SATA III (6gbit/sec max)
Sequential Access - Read up to 355MB/s (3X a hard drive)
Sequential Access - Write up to 215MB/s (2X a hard drive)
Random 4K Read/Write: 60K / 45K IOPS (way better...)

It's the random I/O that makes them really shine. And
especially on reading or loading stuff into memory. I
mean, if you wanted hard drive to have higher sequential
access, you could always put a bunch of them in RAID0,
but there is no way for the hard drive to catch up on
"IOPS". If you do random 4K reads on a hard drive,
even if you use a big RAM cache, eventually the
heads are going to need to move - and then the
hard drive loses out.

HTH,
Paul
 
S

Sunny Bard

Peter said:
Thanks, I'll probably buy a small one just to test is out.
Do they need defragging occasionally?
No point when the seek time of all sectors is the same, and the
pointless re-writing would only accelerate the "wearing out" process.
 
L

LouB

Ken said:
Much faster? Yes.

Downsides? They are reputed to wear out faster. But in my view, they
have been available for too short a time, and nobody knows that for
sure. Time will tell whether or not it's true.
What is there to wear out if it is solid state and there are no moving
mechanical parts?? Do electrons cause friction and heat up?
 
M

Mike Paff

What is there to wear out if it is solid state and there are no moving
mechanical parts?? Do electrons cause friction and heat up?
Writing and erasing Flash sectors requires forcing electrons through
a thin insulator to or from a floating gate. That act of forcing the
electrons through slowly damages the insulator, limiting the number
of write/erase cycles that a sector can tolerate.
 
P

Peter Jason

Thanks, this is most informative.
Peter



Paul said:
They have limited write cycles, which means you
can't be entirely careless with them.

The device has an internal controller, and one of it's
jobs is "wear leveling". There is a level of indirection,
between the external sector address, and which block of
NAND cells responds to that address. By moving data
around,
on each write operation, the controller keeps track of how
many times each block has been written, so that they all
"wear out at the same time". Then, the lifetime of the
device is measured in "gigabytes of writes", where you
multiply the device capacity, by the number of writes
allows. (Say 128GB times 3000 writes max, or 300TB of
writes. SLC NAND may have ten times the write cycles of
MLC NAND, but is much more expensive. The manufacturer
will smooth over this statement by saying "you can write
1GB per day for the next 8 years" or some such, to make it
sound more reasonable.)

Windows 7 is one of the first OSes, where the designers
provide some level of support for SSDs. For example, if
you plugged an SSD into WinXP, the OS would let you run
the defragmenter. In Windows 7, the OS should refuse to
do that (because it doesn't really help). The SATA SSD
has a sub 100 microsecond access time, so it doesn't
really matter if it is moderately fragmented. There is
next to no penalty for "moving the heads". So rather
than the lukewarm job Windows 7 does on a regular hard
drive, it shouldn't do anything to an SSD. That helps
extend the useful life of the SSD.

Windows 7, in the built-in AHCI driver ("MSAHCI"), has
support for TRIM. That is part of maintaining the SSD,
with respect to files that have just been deleted. In
other situations, you may need to run a utility program,
to achieve the same ends. (Some devices don't need TRIM,
because they have internal functions that emulate it.)

http://en.wikipedia.org/wiki/TRIM

On previous OSes, there are innocent sounding functions,
such as updating the "last accessed" field for files,
each time you examine the file. That is entirely
unnecessary,
in terms of maintaining your personal information, and
was only done on regular hard drives, because there is
virtually no penalty associated with writing over and
over again. With the low limit on total write cycles per
NAND block, it pays to revisit design decisions like that,
and disable any features that don't add value. There
are two choices for a designer, with respect to "last
accessed". Either update all values, just as the OS
is shutting down (so you only waste effort on it, once
a day, or once a session). Or eliminate the function
altogether.
Useful fields, such as "creation date" and "last
modified",
continue to be supported, but stupid ones such as
"last accessed" can be turned off.

You can find acres and acres of advice on setting up
and maintaining SSDs. Much of this will be taken care
of, by Windows 7, so not everything you read in such
articles
should be acted upon. But there are still some interesting
things you should know about. This is a forum maintained
by
one manufacturer, with helpful hints for extending the
usable lifetime of the SSD.

http://www.ocztechnologyforum.com/forum/showthread.php?43460-Making-XP-pro-SSD-friendly

http://www.ocztechnologyforum.com/f...h-stuttering-and-increases-drive-working-life.

Since I don't own one, I don't have any more bookmarks
than that :)

Are they much faster ? Absolutely.

http://www.newegg.com/Product/Product.aspx?Item=N82E16820148363

Interface Type SATA III (6gbit/sec
max)
Sequential Access - Read up to 355MB/s (3X a
hard drive)
Sequential Access - Write up to 215MB/s (2X a
hard drive)
Random 4K Read/Write: 60K / 45K IOPS (way
better...)

It's the random I/O that makes them really shine. And
especially on reading or loading stuff into memory. I
mean, if you wanted hard drive to have higher sequential
access, you could always put a bunch of them in RAID0,
but there is no way for the hard drive to catch up on
"IOPS". If you do random 4K reads on a hard drive,
even if you use a big RAM cache, eventually the
heads are going to need to move - and then the
hard drive loses out.

HTH,
Paul
 
K

Ken Blake

What is there to wear out if it is solid state and there are no moving
mechanical parts?? Do electrons cause friction and heat up?

"Wear out" was probably a poor choice of words. Change it to "fail"
But I can't explain what the failure is.
 
L

LouB

Ken said:
"Wear out" was probably a poor choice of words. Change it to "fail"
But I can't explain what the failure is.
I think Peter Jason did(?)
 

Ask a Question

Want to reply to this thread or ask your own question?

You'll need to choose a username for the site, which only take a couple of moments. After that, you can post your question and our members will help you out.

Ask a Question

Top