There is a somewhat contradictory article on tomsharware.com. It says that compressed NTFS writes are leading to write amplification increase. However, there was no raw data to support this statement.
Is the NTFS compression really works this way?
With information being compressed on the fly, you're consuming more of an SSD's available write cycles than if you were writing the files uncompressed. This could have negative implications on the drive's endurance.
Yes it will, but like most things it isn't clear cut.
The newer Intel SSDs (520/330) compress data themselves but they do not give any increase space for the user to use, instead the drive can use the extra space for wear levelling. It also means less writes, as less data is written.
If you use NTFS compression, you get the lesser writes of course as less data is written due to the compression, however the drive can not compress the data any further, so there is no reserved space for better wear levelling.
Or put it another way, assuming the ideal 2:1 compression ratio for the data you are writing and a 60GByte SSD.
Without NTFS compression, you fill the drive with 60GByte of data, the drive compresses this but doesn't give you the space to use but tells you the drive is full, but in reality it has 30GByte spare for wear levelling.
With NTFS compression, you can fill the drive with 120GByte of data because NTFS compresses it to 60GByte, the SSD can't compress the data any further, so now your drive is full with very little space for wear levelling.
Essentially because you are storing twice the amount of data using NTFS compression and defeating the drives ability to compress and keep some space free, it will wear out twice as quick, assuming you get such a good compression ratio.
Actually, write amplification (WA), and an increased amount of writing (using PE cycles) to a SSD due to NTFS compression (if that really happens) or anything else, are two different things. WA does not simply mean something is causing more writes to occur on a SSD. But it is true that the more writes performed on a SSD, the more likely an increase in write amplification will occur.
Write amplification (WA) is a phenomenon that occurs due to the nature of NAND storage, and how a SSDs firmware deals with the requirements of NAND storage. In general it means that a SSD will actually write more than the amount of bytes it has been told to write, which does use more PE cycles. WA is not the fault of anything that causes a SSD to write data, but may be a side affect of writing to a SSD, depending upon its state ( such as amount of free space and usage history.)
I agree that the article did not demonstrate an increased amount of writes occur when using NTFS compression, or at least did not do a good job of doing so.
The general consensus now is the concern over using up a SSD's PE cycles is not warranted, and is not an issue the user must deal with. NAND storage has proved to be much more robust than some thought, and the number of specified PE cycles is very conservative. I have not seen anything written about "worn out" SSDs, except in special endurance testing done by some amateur SSD enthusiasts. That only occurred after over 100 Terabytes were written to a SSD, and usually several hundred Terabytes were written before the SSD began to show signs of wear. If your SSD fails within the warranty period, it will be replaced, quite likely with a new model, so IMO wearing out a SSD is a non-issue.
I believe the amplification referred to in the article occurs as follows:
First, the data is written in original (uncompressed) form.
Then, Windows compresses the written data in the background.
This creates two write cycles - one for the full data, and the other for the compressed data (albeit, smaller).
I've been able to realize triple space savings with NTFS using Drive Press (http://www.magicrar.com/drive-press.html). I think this also improves overall SSD performance, because there is more spare area available on the drive.
My average space savings are 20%-25% (but this is only after using Drive Press, which compresses more than Windows itself somehow).
TRIM is not an issue within Windows itself, but I employ the best practice of keeping the root of the drive clear so the temp file Intel's SSD Toolbox creates during manual TRIMs is not affected by drive compression.
So I have two 600 GB Intel SSDs on my system - one of them for my source codes, the other for my games. VMs are hosted on both SSDs (there's no RAID involved). The raw capacity of the SSDs is 558 GB. Both drives have 100 GB free space. The source code disk has 650 GB actual usage, the games disk has 550 GB actual usage.
For regular applications, there is no perceived slow-down at all. When I am running 4 VMs simultaneously though on VMware, and doing some heavy-lifting, there is a bit of a noticeable slow-down.
In terms of write amplification it will be worse using NTFS compression if you use the extra space saved for more data. This is because the drive already compresses data before it is written but keeps it for itself. Some simple examples for a 240GByte hard drive.
No NTFS compression:
1) 240 GByte of data saved to SSD -> SSD compresses before writing -> Only 120GByte actually written to the SSD after compression. This leaves half the drive for wear levelling and other house keeping tasks.
With NTFS compression:
1) 240 Gbyte of data saved, compressed by PC CPU to 120GByte -> SSD tries to compress but can't compress much further, so 120 GByte written. This leaves 120GByte of usable free space. If the space isn't filled then you have the same wear levelling space as no NTFS compression, however if you then fill the 120GByte with data, you have no extra free space for wear levelling or other house keeping tasks.
So you have a trade off really, steel the space saved by compressing with NTFS and use the extra space for more storage, or let the SSD compress the data behind the scenes for the SSD's own benefit.
Overall depending on your usage the SSD will probably last as long as you need it to even using NTFS compression.