SSD hard drives??
Sep 26, 2007 at 5:07 AM Post #16 of 24
The order of magnitude is in the 100s of thousands, but even so for things like your swap file, it wouldn't be that hard to amass that many writes to a block over the lifetime of a computer. Modern flash is getting better by the day in that respect though, so it's quite possible that it's no longer a major issue. There's also always wear leveling which, for a standard workstation type use-case probably increases lifetime several fold easily.

They're still only really suitable for very specific cases though. The biggest advantages being reliability and power usage. I definitely wouldn't be adopting the tech myself for at least a few generations. When SSD is available at a similar price point (with less capacity of course) it might be a viable option for me, since I don't tend to need huge capacity all the time (would be fine to attach an external disk for that). Only if battery life is of utmost importance to you can I see it as a serious consideration as the only disk in a system.
 
Sep 26, 2007 at 6:13 PM Post #19 of 24
Quote:

Originally Posted by error401 /img/forum/go_quote.gif
The order of magnitude is in the 100s of thousands, but even so for things like your swap file, it wouldn't be that hard to amass that many writes to a block over the lifetime of a computer. Modern flash is getting better by the day in that respect though, so it's quite possible that it's no longer a major issue.


Write cycles are in the millions to tens of millions range for high quality NAND flash. For NOR flash, write cycles are in the hundred thousand range. SSDs use NAND flash, and with wear leveling, properly implemented SSDs can rival hard drive MTBFs now.

Also, the swap file case isn't valid for a SSD drive. It's a technique designed to reduce seek times on hard drives by physically relocating often used data close to each other. That does not translate to SSDs.

Quote:

Originally Posted by error401 /img/forum/go_quote.gif
They're still only really suitable for very specific cases though. The biggest advantages being reliability and power usage.


Swap reliability with durability. SSDs vs HDs reliability is a wash, but SSDs can survive lots of abuse that HDs cannot.
 
Sep 26, 2007 at 7:31 PM Post #20 of 24
Quote:

Originally Posted by marvin /img/forum/go_quote.gif
Write cycles are in the millions to tens of millions range for high quality NAND flash. For NOR flash, write cycles are in the hundred thousand range. SSDs use NAND flash, and with wear leveling, properly implemented SSDs can rival hard drive MTBFs now.

Also, the swap file case isn't valid for a SSD drive. It's a technique designed to reduce seek times on hard drives by physically relocating often used data close to each other. That does not translate to SSDs.


Do you have a source for that reliability figure? Just did a quick Google and it seems a modern Hynix 16Gb module (2GB) quotes 100K write cycles (an Intel device is the same). I didn't think we were an order of magnitude beyond where we were at in December yet...

Most users aren't going to know they need to disable their swap file, and for many that's not really an option anyway. Compound that by the fact that Windows' (XP anyway) memory management is awful and it will use gigs of swap when there's absolutely no reason for it to, and you have a recipe for premature failure. Not to mention that files like the registry are written many, many times without user intervention. Wear leveling will take care of the majority of this sort of thing since most users don't, on average, write to every block 100,000 times over the disk's useful lifetime. It's still a serious consideration when the MTBF of a spinning disk doesn't, from the studies I've read, seem to be related at all to its activity level. For the average user I think you're right and that it's probably a wash, but definitely something to be aware of and to consider. I don't think I'd trust anything important to a computer with just SSD just yet...but then I don't trust anything less than full redundancy anyway.

Quote:

Swap reliability with durability. SSDs vs HDs reliability is a wash, but SSDs can survive lots of abuse that HDs cannot.


Exactly. Oops.
 
Sep 29, 2007 at 6:28 PM Post #24 of 24
Quote:

Originally Posted by error401 /img/forum/go_quote.gif
Do you have a source for that reliability figure? Just did a quick Google and it seems a modern Hynix 16Gb module (2GB) quotes 100K write cycles (an Intel device is the same). I didn't think we were an order of magnitude beyond where we were at in December yet...


The 100k writes number is a lot older than December of last year. It's been the industry standard quoted lifespan for NAND flash for a very long time. Well, very long in computer time at least.

The reason the 100k number is still around is because there are two types of NAND flash. SLC, one bit per cell, and MLC, 2 bits per cell. MLC is newer and has a significantly shorter lifespan than SLC. As there are 2 bits (4 states) per cell, instead of 1 bit (2 states) per cell, threshold degradation, the primary failure mechanism, affects MLC much sooner than it does in SLC. In return, it has ~ 4x the density. MLC hit the 100k lifespan mark fairly recently and the Hynix part you linked to is an MLC part.

SLC lasts significantly longer than that, even if it isn't stated on the datasheet, and is what is used in SSDs. Many commercial vendors will guarantee performance significantly beyond the 100k number, and Toshiba's design guide quotes a 1M cycle endurance for NAND. Also tests to destruction, like the one shown here, shows that lifespan can significantly beat the 100k number.

Course, production errors can have significant impact on lifespan. Samsung dumped a whole lot of chips not too long ago because the tested parts had an MTBF of ~ 5k writes.
 
You must log in or register to reply here.

Users who are viewing this thread

Total: 2 (members: 0, guests: 2)
Back
Top