Personal computing discussed
Moderators: renee, morphine, Steel
ClickClick5 wrote:I put an OCZ Agility 4 64GB (FW: 1.5.2) into my brothers PS3 this past saturday. So far so good. GT5 loads about 65% faster and GTA4 loads 4% faster. But after all saturday playing GT5, no issues yet.
MadManOriginal wrote:It didn't die within hours, that's something!
Forge wrote:If anybody following this thread wants to really understand why your SSD is unlikely to live beyond 5-6 years on average, go read about write amplification. Basically, any time you rewrite or write non-optimally, your disk is going to do garbage collection and TRIM rewrites, behind the scene. Also, wear levelling sometimes triggers cells to be rewritten in a way that's non-optimal for flash lifespan, because it's good for performance.
There are a lot of moving parts.
accord1999 wrote:XtremeSystems has a thread where people continuously write to various SSDs and see how much they can write before the SSD dies.
http://www.xtremesystems.org/forums/sho ... nm-Vs-34nm
Their experience with small SF-1200 based drives suggest a 120GB drive would be good to 500 terabytes of writes or more. There's also a 256GB Samsung 830 that's still alive after nearly 5 petabytes of writes.
just brew it! wrote:Unless they are also testing to see how long the drive is capable of retaining written data after being abused this way, the results are meaningless. As the insulators of the floating gates in the flash cells sustain damage from repeated writes, the flash cells' capability to store data long-term decreases. The drive may still be perfectly capable of retaining data for a few days or weeks, but will gradually corrupt the data over time. Unless you're only using the drive for short-term scratch storage or as a swap partition, retention time matters!
morphine wrote:Forge wrote:If anybody following this thread wants to really understand why your SSD is unlikely to live beyond 5-6 years on average, go read about write amplification. Basically, any time you rewrite or write non-optimally, your disk is going to do garbage collection and TRIM rewrites, behind the scene. Also, wear levelling sometimes triggers cells to be rewritten in a way that's non-optimal for flash lifespan, because it's good for performance.
There are a lot of moving parts.
I presume you meant, "if you do really heavy writing"? Because with a regular-Joe usage pattern, I see them lasting 10 years on more, controller/power problems notwithstanding.
Forge wrote:In fact, Average Joe's usage pattern is going to be much worse than any of ours, since he has no concept of, nor does he care about overprovisioning, disabling indexing, TRIM, and more. He'll buy that SSD, partition the full visible space as one big NTFS partition, and jam it full of files, till it's 99% full. Then he's going to gripe about how it's getting slower, and if he's really Joe Average, he'll probably download a disk defragmenter and run it multiple times, willfully ignoring all the warnings each time.
flip-mode wrote:You really disable TRIM?
Microsoft wrote:Indexing a new item can cause updates to thousands of indexes—one for every word found in the item. Updating all of the existing indexes would take a long time. Windows Search deals with this by creating indexes in memory and flushing them to disk as a single index when they are big enough.
Microsoft wrote:9) Each token is added to an in-memory word list. When enough in-memory changes are accumulated, they are written to a shadow index on disk, and the property store writes are also flushed to disk. This is done to minimize disk writes. Eventually the merger merges the multiple shadow indexes.
Ryu Connor wrote:I'm not sure why people make it out to be such a monster.