Wouldn't caching any mechanical drive holding primarily bulk data with an SSD be a waste? Certainly an array of mechanical drives - SSD's are WAY faster than spindles when it comes to small random access read and writes, but sequential speeds on spindles are pretty fast. Plus, doesn't Intel's Z68 caching ignore large files anyway? Seems like you'd be wasting the potential of the SSD.
I see SSD caching as a great way to take advantage of what SSD's do best: read and write small random bits of data. Let's the spindles handle all the bulk data transfer.
Curious to see some hard numbers, but I wouldn't be surprised if putting an SSD cache in front of a couple WD Black drives actually decreased performance (when used for data storage). Or at least provided very little practical benefit.
There is another thread -- I think it's on "Motherboards" -- where we had an extensive discussion on this and I don't think it's dropped to "inactive." Several people provided ATTO and CrystalDiskMark results on their "accelerated" HDD with SSD caching.
It seems that bigger gains were reaped with an SATA-II SSD and an SATA-II HDD. That is, reported in a benchmark test at a review web-site, "up to" a 400% gain in performance over HDD standalone benchies, asymptotically approaching 80 to 90% of the SSD's performance after user "software and OS" habits are "learned."
I threw in with my test of a Caviar Black SATA-III and Elm Crest SSD SATA-III, followed by a SATA-III Veloci-Raptor and the same Elm Crest. The Elm Crest on an SATA-III controller is supposed to hit 520 MB/s or thereabouts in sequential "reads" and about half that for "writes." We saw a degradation against that spec for the formatted partition of the Elm Crest under the RAID0 configuration for caching, so that the "reads" fell to between 350 and 400, with a corresponding or proportionate drop for the writes.
With the Cav Black connected to an SATa-III port, "Accelerated" (SSD-cached) "reads" were just over 200 with "writes" between 98 and 99 under "enhanced" mode. In "Maximized" mode, reads and writes are near the same level, with some slight degradation of the accelerated-drive "reads" but a sizeable improvement in writes.
With the "accelerated" Raptor and enhanced-mode, reads were around 262 with writes at about 135 to 140. [These are "sequential" reads and writes-- I must remind.] In Maximized mode, reads take a small hit dropping below 250 while writes are over 180.
Point being: this suggests that attempts to capture greater-than-SATA-II performance with this technology don't seem to yield any gains. Instead of a 400% improvement over --say -- the Raptor's 145 MB/s sustained-throughput spec, the result is less than 200% of that or less than 100% improvement.
But also. SSD's will not spell the end of HDDs. They both fit into the established model of "memory" or "storage" in a pyramidal hierarchy of a computer system. Right now, a 500GB SSD will run you about $1,000. You can get a 2 TB drive for a tenth of that. When 1TB SSD's are released, we will see 3 or 4 TB HDDs.
NOw for the type of computing that we're used to, known only to our current or past experience, a 1TB SSD would be preferred for all OS software and files, but there's the problem of expense -- a factor in that pyramidal hierarchy of trade-offs between speed, storage volume and cost.
I very much like the idea of a 600 GB drive offering 260 MB/s reads and 180/190 MB/s writes. It seems "instantaneous." But I'd really like to see that performance realized to this mythical SATA-III, 6.0 GB/s spec, so that the overall accelerated performance is around 375 or 400 MB/s for those sequential writes, and maybe 325 to 350 for the reads.
If I missed or forgot something, and someone thinks that wish is still achievable with the equipment I've described, please let me know . . .
Facebook
Twitter