Defrag SSD?

[HFS+]

Is it worth defragging SSDs?

 

[Coup27]

No. SSDs should not be defragged and it will harm them if you do.

 

[Mark R]

Defragging SSDs is of very limited value. It won't cause any "harm", other than shortening the life of the drive by a negligible margin.

There are some occasions where it can be handy. E.g. if you want to optimize a bunch of slow PCs, it's often quite handy to clone the hard drive onto an SSD. Defrag and virus scan the SSD, then clone it back to the HDD. Much, much faster than doing the same job direct from HDD.

 

[Coup27]

Defragging SSDs is of very limited value. It won't cause any "harm", other than shortening the life of the drive by a negligible margin.

There are some occasions where it can be handy. E.g. if you want to optimize a bunch of slow PCs, it's often quite handy to clone the hard drive onto an SSD. Defrag and virus scan the SSD, then clone it back to the HDD. Much, much faster than doing the same job direct from HDD.

Defragging an SSD is of no benefit at all. SSDs have garbage collection which place all of the data in a sequential and neat pattern inside of the SSD. As access time is 0.1ms regardless where the data is located, defragging it will not improve this at all, but will add a significant number of power and erase cycles to the NAND. One defrag may not cause a lot of harm, but if the user thinks defragging is a good idea and does it weekly, that will soon add up.

Your second suggestion also adds a massive amount of writes to an SSD for the sake of saving a bit of time. I would also question how much time would be saved when adding up the entire process of cloning, swapping disks, running said operations, and cloning back.

 

[bryanW1995]

Never defrag an ssd.

 

[Magic Carpet]

Is it worth defragging SSDs?

You can try and bench it. Personal experience >.

 

[exdeath]

Defrag on SSD is a waste of time. Even when the data appears sequential to the OS, it's still logically mapped in random places internally by the SSD due to wear leveling, write amplification, etc. You're just moving sectors from one random location to another random location, regardless how "defragged" it looks. Even when it looks linear in Windows after a defrag, internally behind the SSD NAND controller its still just as physically fragmented in the flash chips as it ever was.

 

[taltamir]

exdeath is correct.

SSDs wear leveling means that the actual location of data does not match what it shows the OS, there is a lookup table that matches the two in the SSD controller.

Defragging the SSD will shuffle data on the lookup table to make it appear defragmented... which causes the drive controller to shuffle the actual data randomly which will decrease the drive's lifespan. It will have no effect at all on your speed.

What you need to do for an SSD is give it TRIM info, which allows the controller to do maintenance stuff which improves performance similarly to defragging.

 

[martixy]

In short:
Please, for the love of all that is holy in technology, DO NOT defrag SSD's.

 

[Mark R]

For interest's sake, I did do some performance tests with defragging an SSD on my Intel X2-M.

There is some performance benefit, but it is small - and for general use would not be noticable, or probably even benchmarkable. However, there are some circumstances, where the performance increase may be detectable.

The issue comes from how the OS accesses data on the drive. If you have a block of data that is stored "sequentially", then the OS will request the data in the largest chunks supported by the drive (usually 512 kB per each SATA request). The result is that the drive will internally parallelize the individual flash lookups, and consolidate them into a single response.

Where a file is highly fragmented, the OS will request the data in the size of the fragmented chunks - these could, in extreme cases, be 4 or 8 kB chunks. Each request needs a SATA transaction, takes up a space in the request queue, and has overhead both in the drive controller and host SATA controller. This is why the "4kB random read" benchmark scores are so much lower than the "sequential read" scores on SSDs.

In effect, reading a highly fragmented file has an element of "random read" in it, which reduces performance compared to a sequential access.

Now, this is not really much to get excited over. But, I set up a contrived case, whereby I filled up the drive with 4K files, and then deleted some of them at random to free-up space, and then filled it up with a single large file, that would end up being massively fragmented.

Streaming that file was horribly slow (about 30 MB/s). Following a defrag, streaming that same file was over 200 MB/s.

In terms of the amount of writes that the defrag put on the drive, it was on the region of 40 GB. In terms of how much wear that put on the drive, I'd estimate the cost of that defrag at about $0.02. (i.e. it cost less than the electricity used by the PC during the time the defrag was running).

So, would I recommend defragging an SSD? No. It's unlikely to be of any real value, except in highly exceptional circumstances. Does it cause temporary performance degradation? It may do. A freshly defragged and unTRIMmed drive, may well have slow writes due to exhaustion of free pages. TRIM support in the OS, or the use of a TRIM tool (like the intel toolbox) will restore performance. Is there significant damage? Not for a one off, or even a weekly defrag. Even doing it daily is unlikely to cause a relevant amount of wear.

 

[taltamir]

There is some performance benefit, but it is small

No there isn't. You are randomly shuffling data. That CAN result in a slight increase OR a slight decrease in performance. But overall it makes no difference.

Also, turn on AHCI for NCQ

Also Also, what is the margin of error on your tests and what was the difference between defragged and non defragged within that margin of error?

 

[Lipoly]

I know this is a thread resurrection, but was interested in this topic. My SSD is sitting @ 75% fragmentation w/some files being in 40 thousand non-contiguous parts as reported by defraggler (2.6GB World of Warcraft file).

Either I am totally missing something, or there is some misinformation in this thread. It is my understanding that people suggest against defragging SSDs b/c the access time is so fast, moderate fragmentation impacts performance very little and therefore puts undue wear on the SSD.

I totally agree for the majority of cases.

However, at some point it seems it WOULD be beneficial to defrag b/c .1ms adds up if a file is in 40k pieces (40k * .1ms = 4 seconds overhead). If this file were contiguous there would be one access penalty, the rest would blaze through @ full read/write speed.

What am I missing?

Defragging an SSD is of no benefit at all. SSDs have garbage collection which place all of the data in a sequential and neat pattern inside of the SSD.

Where did you get this information? If the data were sequential, I would expect it to be reported as such in defrag tools. Garbage collection simply involves moving pages out of a block of storage so the stale pages can be freed (by deleting the entire block)...there is no guarantee they will be sequential after this.

No there isn't. You are randomly shuffling data.

Not true, defrag does not randomly shuffle data. Same as above...how did you determine this?

My real question is does anyone use a tool like defraggler to defrag specific files what are massively fragmented? I would never defrag the entire disk, but w/a proper tool like defraggler, it seems like it may be reasonable to defrag certain files.

 

[readymix]

i have defragged. i admit that. owning 8 ssd's the total number of defraggings over the last 2+ years would be appaling, should i divulge. it's not that i think it helps but like the office or shop i just have to tiddy this mess up before i do anything else the death of my ssd's from defragging seems to be greatly exagerrated.

 

[Ferzerp]

..

 

[Lipoly]

..

Thanks ferz, appreciate the input.

Readymix: I am also skeptical of the "never, under any circumstances defrag an SSD" mentality that seems to permeate the community...I'm just unsure how effective it can be and under what circumstances people who are not anti-defrag zealots do it.

 

[Shmee]

This thread continuing does more harm then good. The OP was probably just trolling anyways. If you have a specific question about SSD's or fragmentation, or whatever, go ahead and make a new thread. Closed.