Do SSD's have a real world noticeable benefit?

[aamsel]

Originally posted by: gersson
When Windows 7 comes along, things will be better because of the trim command. Not to mention trim needs to be supported by the RAID controller...

Am preparing to setup two SSD's in RAID 0 on an ICH10R, fully aware of possible data loss concerns and the other standard RAID 0 pros and cons.

If TRIM (when implemented into Windows 7) requires RAID controller support, would it likely take longer for standard on-board RAID controllers (such as ICH10R) to get TRIM support than standard SATA controllers?

I have had bad luck over the years with RAID 0 and am doing what I can to talk myself out of using that configuration. :laugh:

These would be X25-M drives.

 

[taltamir]

would it likely take longer for standard on-board RAID controllers (such as ICH10R) to get TRIM support than standard SATA controllers?

It is not necessirily the CONTROLLER as it is the MODE OF OPERATION.

If your mobo is set to SATA/IDE mode, it will just forward the TRIM command as a generic unknown command to the drive. the SATA controller doesn't recognize or interfere with any command.
If your mobo has an old and or crappy SATA controller it will mess up the TRIM address and corrupt data.
If your mobo is set to RAID mode, every command must be recognized and translated into something fitting for a raid array (from something meant for a single drive).
It will require special modifications to the controller to make it support raid, I would expect FEW if any older boards to get this, instead I would expect it as a feature of some future motherboards. But in a non raid mode any non crap controller should work with it right now.

I am curious about AHCI though, how does AHCI affect it? anyone know?

 

[aamsel]

As for future TRIM, support, who knows (of course), but...
I would think that ICH10R would be new enough to be supported.
But, in that it is an integrated chip controller, and not dedicated, maybe not.
But then, since it is an INTEL chip, more likely, perhaps?

Other than possible data loss, up to loss of the complete array, possible
slower access times (although faster throughput), and the inability to transport
drives to any other machine, and other reasons to NOT RAID 0 X25-M's on a
basic controller (ICH10R)?

Most of the reviews I have read say good things about RAID 0 and X25-M, and
have read several real-world improvement reports.

I had understood that, in regular SATA operation AHCI is required to get the best
out of an X25-M. Correct?

 

[taltamir]

As for future TRIM, support, who knows (of course), but...
I would think that ICH10R would be new enough to be supported.
But, in that it is an integrated chip controller, and not dedicated, maybe not.
But then, since it is an INTEL chip, more likely, perhaps?

what does this even mean? can you please clarify what you are saying here?

 

[aamsel]

I was speculating as to the likelihood of simple on-board RAID controllers, such as the ICH10R, being supported for the SSD TRIM functions by Windows 7, realizing that it would not be before at least the first service pack.

This is all well into the future, so not a big deal.

 

[taltamir]

it has NOTHING to do with WINDOWS supporting the CONTROLLERS, it has to do with the controllers being upgraded via a firmware upgrade to include TRIM forwarding capability...
the chances of that a slim to none. Expect trim support in future raid controllers, not existing ones.

there is also nothing SIMPLE about the ICH10R, in fact it is the most complex of all the mobo raid controllers, with its hybrid raid capability.

 

[Red Squirrel]

SSDs are only good for faster loading, and DB servers. They are also disposable as they have limited writes. You're better off using a raid 0 with 4 sata drives and just have backups. (not that I'd recommend this).

 

[taltamir]

They are also disposable as they have limited writes

Bullshit. SSDs are more reliable then spindle drives and for the vast majority of users will LAST longer too... a motor breaks quicker than you will run out of writes, with a typical home users easily going for HUNDREDS of years before running out.

 

[aamsel]

Would a RAID 0 of (2) TWO X25-M 80GB provide a noticeable real-world
difference over a single 160GB X25-M?

Which would be the preferable choice if either could be chosen?

 

[aamsel]

oops

 

[Idontcare]

Originally posted by: taltamir

They are also disposable as they have limited writes

Bullshit. SSDs are more reliable then spindle drives and for the vast majority of users will LAST longer too... a motor breaks quicker than you will run out of writes, with a typical home users easily going for HUNDREDS of years before running out.

It's not bullshit, they really do have limited writes and they really are disposable.

It may be bullshit to say the limited writes have any practical consequence, but it is an indisputable consequence of the physics that govern the function of the devices. They have limited writes.

For 99.99999% of end-users this is of no practical consequence because the absolute magnitude of the number of writes is so high that we will not sufficiently exhaust them before disposing of the device as part of a standard upgrade cycle. But limited writes they do have nonetheless.

 

[Idontcare]

Originally posted by: aamsel
Would a RAID 0 of (2) TWO X25-M 80GB provide a noticeable real-world
difference over a single 160GB X25-M?

Yes it will if your real-world usage of the array is transferring large files to memory or from memory, or to another disk array which has even higher bandwidth.

Originally posted by: aamsel
Which would be the preferable choice if either could be chosen?

I would go with raid-0, personally.

With this specific SSD comparison your reliability is actually nearly identical, the 160GB drive has 2x the number of flash chips over each 80GB drive, so raid-0'ing two 80GB drives doesn't increase the likelihood of failure at the physical media any more so than having all the chips contained in a single package.

There are other points of failure (controller IC's) that do elevate the risk of failure for the raid-0 array, but that topic is really a fringe concern over the technicality of an elevated risk of failure and virtually never has anything to do with contemplating the actual magnitude of the real-world probability of failure.

If raid-0 failure rate (call it X) is to be viewed as excessively high and intolerable then I submit that not going raid-0 and at best reducing that failure rate to X/2 should also be viewed as still being an unacceptable failure rate.

For me, regardless whether the fail rate is X or X/2, I do backups and live life on the hairy edge of probably having a raid-0 failure at some point in the next 50 yrs, if that.

 

[taltamir]

Originally posted by: Idontcare

Originally posted by: taltamir

They are also disposable as they have limited writes

Bullshit. SSDs are more reliable then spindle drives and for the vast majority of users will LAST longer too... a motor breaks quicker than you will run out of writes, with a typical home users easily going for HUNDREDS of years before running out.

It's not bullshit, they really do have limited writes and they really are disposable.

It may be bullshit to say the limited writes have any practical consequence, but it is an indisputable consequence of the physics that govern the function of the devices. They have limited writes.

For 99.99999% of end-users this is of no practical consequence because the absolute magnitude of the number of writes is so high that we will not sufficiently exhaust them before disposing of the device as part of a standard upgrade cycle. But limited writes they do have nonetheless.

what he was saying, and correct me if i am wrong. is that SSDs are disposeable compared to spindle drives which are not disposable... of course they have an operational limit, everything we make does, its called entropy. I was not arguing that they don't have limited writes, I was arguing the claim that they are "disposable" because of it, and even further, that spindle drives are somehow LESS disposable (aka will last longer) because of it.

You are not going to see many if any people throwing away their SSD because it ran out of writes. They will be completely obsolete before that happens for current gen SSDs. (some of the really early ones is a different matter though)

 

[alcoholbob]

I don't believe write degredation is as bad as data retention, which is a problem that plagues flash memory whereas even with a mechanical error on a HDD the data is recoverable.

 

[Idontcare]

Originally posted by: Astrallite
I don't believe write degredation is as bad as data retention, which is a problem that plagues flash memory whereas even with a mechanical error on a HDD the data is recoverable.

At worse, you have 1 year to remove your data from a flash-based SSD once the cells can no longer be written to.

It's also possible for a flash cell to lose its charge over time (albeit a very long time). Intel adheres to the JEDEC spec on how long your data is supposed to last on its SSDs. The spec states that if you've only used 10% of the lifespan of your device (cycles or GB written), then your data needs to remain intact for 10 years. If you've used 100% of available cycles, then your data needs to remain intact for 1 year. Intel certifies its drives in accordance with the JEDEC specs from 0 - 70C; at optimal temperatures your data will last even longer (these SSDs should operate at below 40C in normal conditions).

http://www.anandtech.com/stora...howdoc.aspx?i=3403&p=4

 

[taltamir]

Originally posted by: Astrallite
I don't believe write degredation is as bad as data retention, which is a problem that plagues flash memory whereas even with a mechanical error on a HDD the data is recoverable.

another utterly wrong statement.
Data retention does not "plague" SSDs, as Idontcare explained. and the assumption that it lasts forever on a spindle drive is wrong, spindle drives lose data via bitrot and bit flips by background radiation.

Also a mechanical error on a spindle drive is not in any way shape or form more recoverable than an SSD cell failure. either case the data is irrevocably gone, and the location is rerouted to a spare.

Unless you mean TOTAL failure of mechanical drive, aka, motor failure. in which case disassembly of the drives to read the platters can be done, but such repairs are also doable on SSDs.

 

[Idontcare]

Remember, we like Astrallite so please lets avoid the "another utterly wrong blah bla blah bla" type posts and just help each other learn things we might not have previously known about while fostering an environment that keeps minds open to new information, including our own.

 

[Rubycon]

Originally posted by: taltamir

They are also disposable as they have limited writes

Bullshit. SSDs are more reliable then spindle drives and for the vast majority of users will LAST longer too... a motor breaks quicker than you will run out of writes, with a typical home users easily going for HUNDREDS of years before running out.

Who lives that long?

Speaking of reliability how does SMART work with these drives if cells are bad? Is this data reported/logged as with magnetic media with a defect map which grows, etc.? If so how would a program like Spin Rite work? (besides really fast! :laugh: )

 

[Emulex]

bell curve fit man. a barracuda/cheetah may last 10 years; it may croak in 2 months.

ssd is no different.

however SLC is rated 10X longer than MLC.

really you are going to see the intel units lasting alot longer than the others because they use the ram for write leveling (purely) than just cache.

 

[taltamir]

Originally posted by: Rubycon

Originally posted by: taltamir

They are also disposable as they have limited writes

Bullshit. SSDs are more reliable then spindle drives and for the vast majority of users will LAST longer too... a motor breaks quicker than you will run out of writes, with a typical home users easily going for HUNDREDS of years before running out.

Who lives that long?

Speaking of reliability how does SMART work with these drives if cells are bad? Is this data reported/logged as with magnetic media with a defect map which grows, etc.? If so how would a program like Spin Rite work? (besides really fast! :laugh: )

smart actually has a value that starts at 100 and decreases slightly on the intels AFAIK, it has a threshold of 0.
A guy came in and said one of his "custom values" in smart for the intel X25 showed a value of 96 on a 6 months old drive. He started burning it in and after 3 days of non stop burn in it dropped another point... (must have been pretty close to that drop anyways)

There was a second value with a threshold of 10 that decreases much slower, I don't know what that is.

I calculated that at the rate he was going it would last him 12.5 years from date of purchase.
Which is certainly not hundreds of years, but he seems to be doing a lot more writing than I do on my main drive.

 

[Idontcare]

Originally posted by: taltamir
I calculated that at the rate he was going it would last him 12.5 years from date of purchase.

Can you imagine what it would take for you to find yourself in the situation in life where you felt compelled to continue using an 80GB SSD from 2009 in the year 2021? :shocked:

Think back 12 yrs to the hard drive storage system you had back then and then imagine trying to use that storage system today...I think you'd sell body fluids for $30 and go buy a 2yr old drive in the F/S forums before you'd waste your time and patience attempting to use your harddrives from 12yrs ago.

So yeah, for all practical purposes the lifetime of an SSD is like the lifetime of a CPU, both exceed their practical replacement rate by a wide wide margin.

 

[taltamir]

funny you should say that. last year i fixed a factory floor computer that was used to upload specs into a piece of machinery... the computer was from 1997... it failed because the HDD had so many bad sectors and so much corrupt data that practically nothing was working anymore on its OS (windowsNT4). they had older computers in storage that weren't working anymore.

 

[Rubycon]

Originally posted by: taltamir
funny you should say that. last year i fixed a factory floor computer that was used to upload specs into a piece of machinery... the computer was from 1997... it failed because the HDD had so many bad sectors and so much corrupt data that practically nothing was working anymore on its OS (windowsNT4). they had older computers in storage that weren't working anymore.

There are S100 based CCS300's still in use for controlling lights and stage in old theaters. These have Z80 cpus, 64K (K not M!) RAM and a pair of 8" floppies! The spindle motor on those floppy drives is induction like a window fan and equally powerful! :Q

 

[taltamir]

thats quite amazing, hardly believable... this reminds me a story about a server... which was running for 20 years straight... and it was sitting INSIDE a wall

 

[alcoholbob]

Hey my 1.44" disks are still working from 1996! I had to use 6 floppy discs for Microsoft Close Combat. Man that was such an awesome game. I still have them when I need a DOS 6.1 boot disk for those golden oldies.