Seems like the pace of articles is ramping up!

What worries me is the steadily reducing amount of spare capacity on SSDs. I have an older Intel SSD that is rated for a maximum of 14TB of writes. It ran out of spare capacity after about 18 months in my main computer (used mostly for web and software development). Had it replaced under warranty and the new one is well on its way to hitting that 14TB limit too.

In another few months it will be returned to Intel, just before the warranty period is over. Then I'll put it in my laptop for some lighter use. Now I have a Sandforce based Crucial drive in my main machine, but I keep it backed up because I have a feeling it won't last much longer than the Intel drive did.

You wrote 25.6 Gigabytes of data to your SSD per day for 18-months, that is insane I am not surprised that you get through drives so quickly if you are writing that amount of data to them.


Andy

26GB is nothing. Also don't forget that writes are amplified. My machine has 16GB of RAM so it isn't even page file thrashing.

If andy's figure of 25.6 GB per day is correct, that is a crapton for the average user. I would positively argue the majority won't even download that much in a month, never mind introduce it as daily writes to the SSD even if it was their only drive. The dev work must be using some really heavy loads - guess this is where the 5-year warranty of the 520 would come in handy.

I personally don't mind paying a reasonable extra for longer warranties, especially a five year one. Longevity is good for everyone's sake: it's stability for me, value for later owners, spares natural resources and gives manufacturers both a chance for healthy profits and an incentive to actually QC their products. I say "healthy profits" as opposed to lean profits from disposable crap that only keeps the manufacturer afloat while there's constant high volume sales going on - lean profit/high volume is fine for toilet paper, but SSDs (and other PC hardware for that matter) aren't biodegradeable or made from renewable resources AFAIK.

Some programs(*) write a lot needlessly to the system drive so it's worth going over ones system once with this in mind. I wouldn't be too surprised to find some office workstations getting around those numbers. Fortunately with todays prices a 120GB drive today is unnecessarily small for a system where this factor might be of importance. 240GB drives are around what the current 120GB models cost when they were introduced.

(*) If working everyday with Photoshop and the system has RAM to spare, try a RAMdisk for scratch area.

I feel the exact same way.

Personally i feel that best ssds to get are Samsung 830 as the best in the market, and Crucial M4 if you want a good value while still very competitive in performance. I avoid as much as possible sanforce drives.

I dont know what this means.! If you are referring to the SSD automatically moving around the writable blocks in the form of "wear leveling", then that is not part of the 14TB of writes, they only list "user writes". Also out of interest could you point me to to where you saw that figure.



That's why I was so shocked about that amount of data being written so regularly.

FYI the maths were rough and in brackets. 18 months (365x1.5=547), 1TB = 1024GB (14,000 / 547 = 25.6GB per day).

Another FYI, the larger the SSD capacity the more space there is for wear leveling meaning that the larger the SSD to more writes it can do safely.


Andy

SSDs are arranged in pages, with a typical page being many kilobytes in size (4-8kB being typical I think). When data is written to a page, the whole page must be erased and rewritten, regardless of how much data is actually being changed. So for example a write of a single 512-byte filesystem block might require a whole 4-8kB page to be erased and rewritten. Hence amplification - a 512-byte write at the filesystem is amplified to an effective 4-8kB write at the flash memory level, and data that is unchanged in the page is rewritten - needlessly from the filesystem point-of-view but necessarily because of the way flash works.

SSD controllers will do their best to coalesce writes to a given page, so for example if you sequentially write a page-size-worth of data in 512-byte blocks, this should become a single erase-write cycle. But there's only so much they can do, particularly if the stream of writes is very random.

I just checked my laptop. The Intel tool reports 1.46TB of writes in around 5 months of moderate use.

I was surprised by how much data gets written at first. I used TrueCrypt on an older machine and it also keeps those stats, which is when I first noticed.

According to the spec sheet for the Intel 510 series of SSD's.

From page 9 of the PDF in the link below.

http://www.intel.com/content/dam/www/pu ... cation.pdf

"Minimum Useful Life/Endurance Rating 5 years

The SSD will have a minimum of five years of useful life under typical client workloads with up to 20 GB of host writes per day."

I did ask earlier where this figure of 14TB of written data in the SSD's lifetime came from, well wherever than number came from it was not from Intel regarding this model of SSD.

The number is 36,400 GB of data written to the SSD should not be a problem across a 5-year period. That is 36.4 TB of data. not 14 TB.

And as far as write amplification is concerned, for someone that is using Photoshop this wont be an issue as anything that gets written to the SSD is going to be in the magnitude of Megabytes rather than 4/8KB or smaller blocks so there will be little in the way of waste. I would assume that you dont have your internet browser save data to the SSD as you browse.! That would cause the data amplification to be a problem as there will be many tiny files being written constantly.


Andy

The Intel SSD in question is an X25. I can't remember where I found the 14TB figure now.

Also keep in mind that Intel are using an SSD controller that does compression now, where as my older drive does not. They don't state what average level of compression they assume.

I use SSDs for my main system drive, and that means all temporary files go on it including web browser caches. The whole point of having an SSD is to increase the speed of the system and fast access to caches and temporary files is an essential part of that. Furthermore my laptop, like most, only has an SSD and no hard drive since there is only one drive bay.

Before the data is written to the drive it is compressed, so that would not affect the quantity of writes.



I didn't notice any difference when I moved from having Firefox keep cached data on my old HDD to when I got my first SSD and totally disabled any drive caching in Firefox, it only uses RAM to cache data which means that as soon as you close all open Firefox windows all cache is wiped.

This would be an interesting test for you to try. Set Firefox (or whatever your preferred browser is) to "Private Browsing" mode and use it for a while to spot any performance differences. FYI when I stopped using drive caching for Firefox I was on Virgin 10Mb cable, now I am on 30Mb and have not even considered turning caching back on. I did this to reduce the quantity of data being written to my SSD as this was a major concern for potential SSD buyers a couple of years ago.


Andy

Of course it does. If the data is compressed 50% before it is written then potentially only 50% as many sectors need to be written. In the past storage manufacturers always assumed 50% compression ratio for things like tapes, but Sandforce don't make any specific claims. Whatever their lifetime write limit is they are assuming a certain level of compression and thus reduced number of writes.

One interesting thing to note is that encryption will completely break compression on SSDs. Encrypted data is basically uncompressable, so if you use BitLocker, Truecrypt or whatever the Linux/Mac equivalents are then your SSDs lifetime will be reduced. Both Intel and Sandforce claim that their drives encrypt data but are too sketchy on details for it to be of any real use. One would hope that by setting an ATA password it is hashed and used to encrypt the drive key, but neither Intel nor Sandforce have been able to confirm that when asked. Desktop BIOSes almost never let you set an ATA password anyway, and laptop BIOSes usually combine it with the BIOS password which logically must be stored in plaintext somewhere in CMOS memory, making it essentially useless.



I find it makes a big difference, but then again I often have 30+ tabs open. I have 16GB of RAM but Firefox doesn't seem to want to use more than about 1GB for memory cache no matter what settings I use. I have Virgin 125Mb but that doesn't make slow web sites any faster. I browse a lot of Japanese sites too and the network latency to the other side of the planet is significant.

Anyway, my point stands, SSDs are IMHO not a simple drop-in replacement for HDDs and particularly on the cheaper models I question the lifetime write limits. My Intel drive had a 3 year warranty and hit the limit after under half that time, so I'm sceptical about Intel's expected durability. My solution is to back up regularly and if it hasn't failed before the end of the warranty I will write to it until it does because otherwise it is going to fail shortly afterwards.

Could the 1 GB be Firefox's automated memory cache limit? I think that's the figure (thereabouts) for disk cache at least.