geekhack
geekhack Community => Off Topic => Topic started by: wellington1869 on Thu, 06 August 2009, 15:01:08
-
Ok, so I heard from a friend that if you buy a large (say, 64gb or more) solid state drive and format it as NTFS (instead of with smaller 32gb FAT32 partitions, windows limits you to 32gb partitions with fat32 I beleive), then he claims the drive life is shortened because windows with NTFS drives writes a log to the drive constantly (and since solid state drives only last for X number of write cycles).
Is he full of it?
-
Considering that's the default, it isn't going to shorten the lifetime of the drive.
I'm still not fond of flash memory drives, because of this limited lifetime.
-
Ok, so I heard from a friend that if you buy a large (say, 64gb or more) solid state drive and format it as NTFS (instead of with smaller 32gb FAT32 partitions, windows limits you to 32gb partitions with fat32 I beleive), then he claims the drive life is shortened because windows with NTFS drives writes a log to the drive constantly (and since solid state drives only last for X number of write cycles).
Is he full of it?
I think he's slightly out of date. There's certain tweaks you can make but it should work. Bare in mind, Windows 7 will be optimized for SSDs, and it will use NTFS, probably without many changes.
I'm still not fond of flash memory drives, because of this limited lifetime.
The MTBF rating on drives like the OCZ Vertex is about 20% higher than that of regular consumer hard drives. Of course, MTBF is a bit of an iffy statistic... Do you have any concrete data on how many read/write cycles a good quality SSD can take before dying?
and since solid state drives only last for X number of write cycles
Yep, unlike them regular drives which last forever and ever!
-
Do you have any concrete data on how many read/write cycles a good quality SSD can take before dying?
Most flash technologies are around 1-2 (up to 5 for some) million erase/write cycles. But most SSD manufacturers build hardware wear leveling into the device (think of it as block re-allocation based on cell usage) that prolong the effective allowable write cycles significantly. How much more depends on how you use the drive.
Unless you constantly write to the majority of the drive, I suspect it would take a significant amount of time for a small set of often written files to wear out all the blocks on the device. Other then that, lifetime depends on how you use it.
In the case of NTFS the drive hardware would bounce the journal around the drive in an attempt to keep the cell usage balanced across all blocks. It wouldn't just keep writing to the same block until it was burned out.
-
Mhmm. I'm interested in getting a small-ish SSD to host my OS and games, and leave all my data on a Samsung 1TB drive. That said, I'm probably going to wait a year for the prices to drop, and for SSD optimization on Windows 7 and linux to mature.
-
There's certain tweaks you can make but it should work
ya i'm looking into the tweaks. I guess with winxp the tweaks should really be done? I think they basically turn off the journal or whatever.
In the case of NTFS the drive hardware would bounce the journal around the drive in an attempt to keep the cell usage balanced across all blocks
i was hoping that was the case.
I'll see if I can find some MTBF hardware testing of SSD drives on windows machines.
-
I thought that Intel still made the most reliable ones, especially concerning prolonged use and steady read/write speeds even when the "anti-wear" logic kicks in.
-
I thought that Intel still made the most reliable ones, especially concerning prolonged use and steady read/write speeds even when the "anti-wear" logic kicks in.
Nah, they just make fast and expensive ones. There's actually been some firmware issues with Intel SSDs that will destroy data. Pretty scary.
I'd say reliability depends mostly on the controller and memory.
-
Worrying about the lifetime of SSDs is kind of dumb in my opinion. I've never had a mechanical hard drive last more than about 2 years. Biggest problem with SSDs are cost to size, so they're not so good for media usage. But I'd kill to get one at work where doing greps on large project trees wastes quite a lot of time and SSDs are near instaneous.
-
Also, be wary of the cheap controllers (j micron for example). Not so much because of data corruption, but because they tend to have awful performance in some situations due to tiny or bad cache usage. As far as I know the only decent controllers are the intels and newer OCZ.
-
Even with wear leveling, I still wouldn't trust MLC flash drives much, even if they'll probably last me quite a while. If you go SSD, try springing for an SLC drive. They're magnitudes more reliable. Storage Search (http://www.storagesearch.com/) has a lot of SSD info... they've been around a lot longer than the current mainstreaming of SSDs. Here are two relevant articles - "SSD Myths and Legends (http://www.storagesearch.com/ssdmyths-endurance.html)", and "Are MLC SSDs Ever Safe in Enterprise Apps (http://www.storagesearch.com/ssd-slc-mlc-notes.html)"
-
drunken, thanks for those links. The first article there (persuasively) argues that on today's SSD's its kind of ridiculous to worry about write endurance on these drives, especially ones using SLC technology, where continuous writing to the disk still gives a theoretical endurance of over 50 years basically.
The second article says even with newer MLC technology (larger capacity, cheaper, and intended for notebook/laptop/cellphone environment -- but much less write endurance) should give a lifespan for the drive on par with current regular hard drives when used in a notebook computer environment (tho he doesnt recommend MLC technology for servers).
-
Worrying about the lifetime of SSDs is kind of dumb in my opinion. I've never had a mechanical hard drive last more than about 2 years. Biggest problem with SSDs are cost to size, so they're not so good for media usage. But I'd kill to get one at work where doing greps on large project trees wastes quite a lot of time and SSDs are near instaneous.
Not to say it doesn't happen, because, believe me, I have seen it plenty of times, but I have never had a hard drive fail. I have always used Maxtor (before they were part of Seagate) and now I have a few Hitachis, all of them are at least 3 years old and still going strong.
-
Not to say it doesn't happen, because, believe me, I have seen it plenty of times, but I have never had a hard drive fail. I have always used Maxtor (before they were part of Seagate) and now I have a few Hitachis, all of them are at least 3 years old and still going strong.
ditto for me by the way. I've seen hard drives fail for others, but i've been lucky, I've had laptops for up to eight years, and the original drive on it never failed.
-
I've seen HDs fail but they typically last quite a few years. For me, the average is quite a lot longer than 2 years but I haven't been keeping track of all this stuff. I don't worry about it.
At the office, I had a 4-drive raid array in which I had to replace 2-3 drives (not all at once!) due to problems within the first year - they simply went bad, even the smart report showed a bunch of replaced sectors, etc. and it developed the dreaded clicking sounds, with associated error conditions. All these drives were purchased at the same time so might be from a problematic batch.
For most seek / bad sector types of problems (when I can't just RMA the drive), I keep a copy of Spinrite at the office. It's a life saver when you need it, and it's been keeping our laptops running in tip top shape (maintenance - run it every few months and you should be good).
-
Ok, so I heard from a friend that if you buy a large (say, 64gb or more) solid state drive and format it as NTFS (instead of with smaller 32gb FAT32 partitions, windows limits you to 32gb partitions with fat32 I beleive), then he claims the drive life is shortened because windows with NTFS drives writes a log to the drive constantly (and since solid state drives only last for X number of write cycles).
Is he full of it?
FAT32 is not limited to 32GB, it can be as big as 8TB when using 32k clusters. The Windows 2000/NT/Vista/W7 format.com is what's limited to 32GB. If you use a 3rd party program you can format it however big you like. However, you will still be limited to using files that are 4GB or smaller (2^32 - 1 bytes).
on today's SSD's its kind of ridiculous to worry about write endurance on these drives, especially ones using SLC technology, where continuous writing to the disk still gives a theoretical endurance of over 50 years basically.
The second article says even with newer MLC technology (larger capacity, cheaper, and intended for notebook/laptop/cellphone environment -- but much less write endurance) should give a lifespan for the drive on par with current regular hard drives when used in a notebook computer environment (tho he doesnt recommend MLC technology for servers).
I didn't quite read the article, but this is correct.
Current SLC drives have a lifespan that outlasts its usefulness. Last I checked the consumer drives like Intel's X25-E last 20 years and only the industrial SLC drives last 50-100, but still that's more than long enough. By the time it dies you'll have gotten a new one that's twice as fast, twice as large, and at half the price, and have probably already thrown it out or stuck it in your kid's PC.
Current MLC drives are hit and miss. The good ones (OCZ Vertex, Intel X25-M, etc) will most likely outlast your hard drive. The cheap ones, like what Asus throws in it's netbooks, would only last for maybe 2 years under normal desktop use.
-
For most seek / bad sector types of problems (when I can't just RMA the drive), I keep a copy of Spinrite at the office. It's a life saver when you need it, and it's been keeping our laptops running in tip top shape (maintenance - run it every few months and you should be good).
Spinrite is great, were it not that it would FUBAR for my new 500GB+ SATA disks. No damage done, but it just dies with a fatal math error at a certain point.
-
Spinrite is great, were it not that it would FUBAR for my new 500GB+ SATA disks. No damage done, but it just dies with a fatal math error at a certain point.
Doh! I haven't had a chance to run it on newer 1TB+ drives. They seem to be fine on the 500GB drives that I have around. It may also depend on the machine / BIOS. If you write to support they may be able to help you.