SSD "restore" alternative while waiting for true TRIM: zero-out?

[Loa]

Hello,

Just found this on diglloyd's site: his solution to the SSD slowdown with usage (because of our lack of TRIM support) is to:

1) clone your SSD (back-up)
2) copy a single 100%-size file on it (like copy a 73GB file onto a 80GB SSD)
3) erase that file
4) clone it back in

Is it really that simple???

http://macperformanceguide.com/Storage-SSD-Reconditioning.html

Loa

 

[300D]

Well, its a great way to wear out the drive.

 

[nanofrog]

Are you actually having issues with the drive slowing down ATM?

If not, wait awhile, and see if TRIM support becomes available. Otherwise, the extra writes will reduce the lifespan if done on a regular basis. Remember, those high ratings are a statistical analysis of 90% of the cells. The remaining 10% have a lower write cycle lifespan, and may only be as good as the flash chips are without any assistance from wear leveling.

If the drive is getting full, the wear will be amplified, as the unused space has decreased significantly (needed for wear leveling).

 

[Loa]

Are you actually having issues with the drive slowing down ATM?

Ha! I don't even have a SSD right now! But that lack of TRIM support is holding me back from buying one so I'm keeping an eye out for solutions.

Also, if one full write is reason to worry about wearing out the drive, then the drives are worthless. My feeling is that we're way too pessimistic about SSD wear, and that in the real world wear won't be a real issue.

But only time will tell. SSDs slowing down (in write performance) is a very real issue though.

Loa

 

[nanofrog]

Ha! I don't even have a SSD right now! But that lack of TRIM support is holding me back from buying one so I'm keeping an eye out for solutions.

Also, if one full write is reason to worry about wearing out the drive, then the drives are worthless. My feeling is that we're way too pessimistic about SSD wear, and that in the real world wear won't be a real issue.

But only time will tell. SSDs slowing down (in write performance) is a very real issue though.

Loa

I'm thinking in terms of frequency, and used capacity. Wear leveling needs unused space, and I can't help but think it's going to be limited rather quickly on 80Gb and smaller drives. So on nearly full drives, dead cells are going to be much more of an issue than those that have say 20%+ capacity unused.

If such a method (what you linked of digilloyd's site) where to be used, and no TRIM support was available, it would add up. Each person would do this at their own discression, and it's hard to say how much that would affect the lifespan, especially with absolutely no idea of the write frequency otherwise.

So for a boot drive, it probably won't matter, but if it's used for movie production for example (one disk for the current project, then moved off to slower/ large/ cheap drives for archival purposes), it very well may.

 

[Loa]

Well, its a great way to wear out the drive.

Unless I misunderstood what the TRIM function is: with an active TRIM function, every time you delete a file you end up writing back every un-deleted pages in a block. So while TRIM let's you keep your original write performance (provided that you don't overwrite too many files, as the TRIM doesn't help then), it forces the SSD to write every time you delete.

So which one do you prefer: writing a huge file to your SSD every month, or having the SSD write every time you delete a file?

Again, IMO, we're making a huge fuss over an issue (wear) that won't be a big deal.

Loa

 

[nanofrog]

Unless I misunderstood what the TRIM function is: with an active TRIM function, every time you delete a file you end up writing back every un-deleted pages in a block. So while TRIM let's you keep your original write performance (provided that you don't overwrite too many files, as the TRIM doesn't help then), it forces the SSD to write every time you delete.

So which one do you prefer: writing a huge file to your SSD every month, or having the SSD write every time you delete a file?

Again, IMO, we're making a huge fuss over an issue (wear) that won't be a big deal.

Loa

TRIM just changes the block information, so the cells don't have to be erased before writing new data. Helps with speed, and could actually help with reducing the writes as well.

Now as far as the write cycles being an issue, it will depend on the specific usage pattern. For now, SSD is aimed at enthusiast users, who typically don't do a lot of writes (compared to say a database that requires updates constantly, such as a banking system). But it may be an issue for workstation useage as well, and certainly is for enterprise useage. Obtaining a setup (single drive or array) where the capacity is higher than is needed will also solve this, it may be difficult given the current capacities available, and of course cost. Always the specifics....

 

[Loa]

TRIM just changes the block information, so the cells don't have to be erased before writing new data.

As far as I know, you can't "change" anything on a drive (SSD or not) without writing.

To change a block's info when deleting some pages from it, you have to read it completely (not a problem), and then write back all the pages that have not been changed (wear issue).

And since you can only delete in blocks of 512K on SSDs (per Wikipedia), every time you have to erase / overwrite / update small files, you end up rewriting a big chunk of that 512K block.

Now, my OS drive has 800 000 + files. What does that tell us: that the OS use a frigging lot of very small files. And every time it has to erase or modify one of the those very small files, it will write nearly the whole block.

So the TRIM command seems to me to be a write intensive command which, for a OS drive, might end up doing more writes than a "single huge file write" would do once a month.

Loa

 

[Thiol]

I personally wouldn't let the lack of TRIM stop anyone from getting an SSD. An average user will not notice the degradation. Back when this issue first appeared with the original Intel X25M, I couldn't observe a difference with a fresh drive, before the firmware update, and after the fix. The truth is that an SSD is so fast with random access compared to a hard drive that any reduction in performance is not noticeable with day to day applications.

 

[Loa]

The truth is that an SSD is so fast with random access compared to a hard drive that any reduction in performance is not noticeable with day to day applications.

Most probably. But still, diglloyd's found a 75% degradation in write performance after 6 months of intense use. (with the updated Intel firmware) That's quite significant!

(I'm also waiting for my year-long contract begins before buying one! haha)

Loa

 

[orpheus1120]

Use HDDERASE.

It's a DOS program. Do not use v4.0, instead use v3.3.

Google "HDDERASE", "Ultimate Boot Disk" and "MagicISO".

 

[Thiol]

Most probably. But still, diglloyd's found a 75% degradation in write performance after 6 months of intense use. (with the updated Intel firmware) That's quite significant!

(I'm also waiting for my year-long contract begins before buying one! haha)

Loa

Where did you get the 75% number? I couldn't find it in your link.

Tom's Hardware finds less dramatic results:
http://www.tomshardware.com/reviews/ssd-performance-power,2279.html

Personally, I don't understand how big of a "problem" this is. All I can say is that I've thrilled with my current SSD.

 

[FieryFurnace]

I am using a Vertex since March (in my MacBook), basically daily for a couple of hours. And I didn't and still don't feel any decrease in speed.

 

[Loa]

Where did you get the 75% number? I couldn't find it in your link.

He states 1/4 performance in the introduction, then specifies with 28% with data to back that average under: "Highly variable write times, very slow average".

Tom found less radical results for 2 reasons: he didn't use the drives for 6 months, nor did he use them as intensely as Diglloyd. Lloyd did say that his usage was extreme and didn't reflect a typical user. On the other hand, the degradation was also extreme.

...as I said: cash is the main reason I'm not buying one right now. I'm just looking for solutions to that slowdown.

Loa

 

[Loa]

I am using a Vertex since March (in my MacBook), basically daily for a couple of hours. And I didn't and still don't feel any decrease in speed.

Good for you, but did you test this out? If you got any slowdown it would be a very very gradual process and would be hard to detect by "feel".

Loa

 

[FieryFurnace]

Good for you, but did you test this out? If you got any slowdown it would be a very very gradual process and would be hard to detect by "feel".

Loa

No, I didn't really test the speed.
All I did was counting seconds for start up and how long it takes certain programs to launch, but I didn't run any speed test programs.

Still, if something would have slowed down a lot, I would have noticed it.
And when the day comes where I definitely feel that the systems gets slower, I will look into resetting my SSD and/or updating my firmware.