Seagate To Introduce Faster SATA Standard
Re: Seagate To Introduce Faster SATA Standard
they could use it with this
http://i.gizmodo.com/5166798/24-solid-s ... -5-seconds
(prepare to be amazed, yet annoyed at the same time)
http://i.gizmodo.com/5166798/24-solid-s ... -5-seconds
(prepare to be amazed, yet annoyed at the same time)
Re: Seagate To Introduce Faster SATA Standard
400mb sec..pah the other guys got 2 GB/sec
Re: Seagate To Introduce Faster SATA Standard
But at what cost?
Re: Seagate To Introduce Faster SATA Standard
From what I can tell, SSD's have 3 issues right now. Well, really 2 main ones: wear-levelling & write speed penalties. Writing on all consumer types requires an erase pass right now (afaik) and there's 2 approaches (hence 3 issues).
Wear levelling is the practice by which writes are spread out across the drive (by the drive's low-level logic) so that it 'evens out' the write patterns, avoiding the case where your swapfile would wear out its portion of the disk (for example). It's not as big an issue as it was in 2006, since there have been advances in flash-memory storage but still present in a machine where the drive may be in use for many years.
Write penalties are also being avoided by doing either deferred writes (for smaller writes, something that RAID chipsets already support) or the 'remap on write' as used in the new Intel X-25 SSD's. There are also flash memory types on the horizon to address the writing performance penalties, but to my knowledge these haven't shown up in the market yet. And the high-end server SSD's that use actual RAM chips (with an onboard BBU) make the consumer SSD's seem cheap by comparison.
Deferred writes only work for so long, and eventually memory fills up so this causes a the delayed write to actually fare worse in heavy use or highly multitasked situations. Hence the well publicized problems with OCZ drives that used jmicron controllers, up to 1 second delay on writes really hurts when several writes stack up. Some drives add additional on-drive memory to help buffer the writes, but at some point you still need to actually access the drive.
Remap on write works to both help with wear-levelling (as it's part of the algorithm) and to speed up write speeds as the controller will 'erase' the 'freed up' blocks where the write was supposed to occur to during idle cycles. However there's a noticeable slowdown over time with these drives and there's been a handful of articles recently speculating on why, the most obvious case brought to light is that the drive tends to wind up fragmenting the heck out of the filesystem (in a way that's been dubbed 'micro-fragmenting'). Also it's worth pointing out that current defragmenting techniques are completely useless with the drives. Though Intel may address the situation once they actually acknowledge that they can repeat the 'fragmentation issue, they're currently denying being able to 'replicate the conditions that lead to fragmentation' (which is a way of denying it without actually denying it).
So it seems to me that the drives are useful right now for machines that won't be heavily used (netbooks) or for situations where you'll be doing mostly reading back of data once written (media playback). Otherwise you'll get stellar read speeds with write speeds that are on par (or slightly better than) current desktop drives, but degrade over time or have other penalties. And of course older magnetic drives will be so dirt-cheap that they'll still be used for large storage.
Wear levelling is the practice by which writes are spread out across the drive (by the drive's low-level logic) so that it 'evens out' the write patterns, avoiding the case where your swapfile would wear out its portion of the disk (for example). It's not as big an issue as it was in 2006, since there have been advances in flash-memory storage but still present in a machine where the drive may be in use for many years.
Write penalties are also being avoided by doing either deferred writes (for smaller writes, something that RAID chipsets already support) or the 'remap on write' as used in the new Intel X-25 SSD's. There are also flash memory types on the horizon to address the writing performance penalties, but to my knowledge these haven't shown up in the market yet. And the high-end server SSD's that use actual RAM chips (with an onboard BBU) make the consumer SSD's seem cheap by comparison.
Deferred writes only work for so long, and eventually memory fills up so this causes a the delayed write to actually fare worse in heavy use or highly multitasked situations. Hence the well publicized problems with OCZ drives that used jmicron controllers, up to 1 second delay on writes really hurts when several writes stack up. Some drives add additional on-drive memory to help buffer the writes, but at some point you still need to actually access the drive.
Remap on write works to both help with wear-levelling (as it's part of the algorithm) and to speed up write speeds as the controller will 'erase' the 'freed up' blocks where the write was supposed to occur to during idle cycles. However there's a noticeable slowdown over time with these drives and there's been a handful of articles recently speculating on why, the most obvious case brought to light is that the drive tends to wind up fragmenting the heck out of the filesystem (in a way that's been dubbed 'micro-fragmenting'). Also it's worth pointing out that current defragmenting techniques are completely useless with the drives. Though Intel may address the situation once they actually acknowledge that they can repeat the 'fragmentation issue, they're currently denying being able to 'replicate the conditions that lead to fragmentation' (which is a way of denying it without actually denying it).
So it seems to me that the drives are useful right now for machines that won't be heavily used (netbooks) or for situations where you'll be doing mostly reading back of data once written (media playback). Otherwise you'll get stellar read speeds with write speeds that are on par (or slightly better than) current desktop drives, but degrade over time or have other penalties. And of course older magnetic drives will be so dirt-cheap that they'll still be used for large storage.
Re: Seagate To Introduce Faster SATA Standard
SSD's would be a huge boon for romplers & sample libraries (which are installed & left mostly alone) once the price drops a bit more...(hint hint braincell).
But otherwise you're correct stardust, SSD's would be an *especially* bad choice as a recording disk.
But otherwise you're correct stardust, SSD's would be an *especially* bad choice as a recording disk.