Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
OWC Rant
- Thread starter brianric
- Start date
- Sort by reaction score
The most problematic Sandforce controller is the 2281 which was in the Vertex 3 and a bunch of other SATA3 drives, including most current OWC SSD products like the Accelsior.
You're correct that Intel managed to offer a fairly reliable product based on this controller, but unlike other OEM's they spent a year validating their own firmware rather than using Sandforce/LSI firmware.
just ran a speed test on my pcie ssd in my sig .. hoping to see 600 / 800 .. on the blackmagic disk test .. i was getting 530/650 .. just a tad higher than an owc 6g ssd i had in a mbp awhile ago ..
There are a number of issues with this product... first of all, Sandforce controllers can only achieve their advertised transfer rates with highly compressible data. With regular image and video file content their write performance is not in the same ballpark and worse than most other controllers. Then there's the fact that the card only uses a x2 bus interface which will bottleneck a couple of SATA3 SSDs in RAID0. And finally, one member here did some proper benchmarking with QuickBench and found the small block random I/O far worse than a pair of regular SATA3 SSDs. Add to that the ridiculous price they charge for this, and I'm amazed anyone buys it.
yea I am kind of tempted to sell it and pick up 2 samsung 840 pros and raid them instead
I think before a SSD is given a rating it needs proper analysis which are technical too. I am not sure if benchmark results can be fully trustable to a extent of helping in SSD purchase! To me, it is just another tool & results are just an approximation. I have used sandforce based SSDs on both kinds of data & the speed results are no match to the benchmark results which clearly indicates there is a gap!! Just dont listen to someone's exp,benchmark results before concluding on a SSD. You will surely agree on me if you actually did the same tests on sandforce based SSDs
You're exactly right, the problem is that most SSD vendors only advertise the maximum large block sequential transfer rates for their products and SandForce OEMs go one step further in advertising that when using highly compressible data (like all 1s or 0s). Thanks to exhaustive reviews like those at AnandTech we get the full picture, and this is where we learn that SandForce drives are not all they are cracked up to be.
However, this thread is not so much about performance as it is about reliability and the number of reports of premature failure of SandForce 2xxx based drives speak for themselves. Anyone that values reliability should be aware of their notorious failure rate.
My anecdotal experience:
I own four Intel 330 Sandforce-based SATA3 drives, no troubles for one solid year of use in a recording studio environment.
Sandforce bad for uncompressible video, but I do audio, using 44.1k & 48k 24 bit WAV files, which are compressible, and the drives have performed very well. Fine for boot, too; my understanding OS and Apps are compressible files. (Please correct me if I am wrong here) And I also understand reads on uncompressible data are fine, in my case, sample libraries, just slow writes when installing the libraries.
I do wish there was a definitive answer on Trim with Sandforce-based drives. I haven't searched for a couple of months, but when I did, found little but disagreement on whether Trim is desirable for these SSDs. So I am running sans Trim at this point.
PS: EVERYTHING is backed up to multiple other drives, direct backup/clone and Time Machine. On and off site.
I own four Intel 330 Sandforce-based SATA3 drives, no troubles for one solid year of use in a recording studio environment.
Sandforce bad for uncompressible video, but I do audio, using 44.1k & 48k 24 bit WAV files, which are compressible, and the drives have performed very well. Fine for boot, too; my understanding OS and Apps are compressible files. (Please correct me if I am wrong here) And I also understand reads on uncompressible data are fine, in my case, sample libraries, just slow writes when installing the libraries.
I do wish there was a definitive answer on Trim with Sandforce-based drives. I haven't searched for a couple of months, but when I did, found little but disagreement on whether Trim is desirable for these SSDs. So I am running sans Trim at this point.
PS: EVERYTHING is backed up to multiple other drives, direct backup/clone and Time Machine. On and off site.
^^^^With their Toshiba drive's Apple uses Sandforce controllers.
Look here:
https://forums.macrumors.com/threads/1633123/
Lou
Look here:
https://forums.macrumors.com/threads/1633123/
Lou
Thanks. This was even more clear...
http://www.anandtech.com/show/6063/macbook-air-13inch-mid-2012-review/4
Anyway, it seems like those Toshiba rebranded SandForce drives have had their fair share of premature failures as well.
I can just say you got to be unfortunate if you had failure again, which is hard to believe! I am sure there are many friends like DPUser who agree upon my point that sandforce SSDs are reliable & there are many false stories which are aimed to mislead people! Hope you make a wise choice.
I think there are some subset of those stories where the Sandforce SSDs are the wrong tool for the wrong job which increasing the failure rate.
Huge applicability mismatches are:
i. extremely large amount of incompressible data. (e.g., a working scratch job for highly compressed video production). Uncompressed data is not only slower through the Sandforce drive it likely also drives up metadata overhead. [ Sandforce controllers are far more tuned for more mainstream balance of some compressed media and other data. Not just one uniform set of compressed. ]
ii. TRIM for TRIM sake. Again the non atomic aspects aspects of TRIM force Sandforce to flush alot of the multithreaded aspects of of handling the metadata. It screws with normal context their drives were optimized for. Coupled with i. it is yet another deviation from nominal metadata movement.
iii. Mismatched NAND and under-provisioned configurations. Pricing pressures pushed some vendors into configs that Sandforce really wasn't tested or provisioned for to the same degree as others. [ this shouldn't matter to the consumer of the drive but indicative that the controller isn't root cause; it is parts selection/combination defect. ]
Things wouldn't be surprised at contributing issues:
a. drives stuck in SSD drive bays with mismtached cooling solutions coupled with "I know better than Apple about thermal and power distribution" configurations.
b. funky power. Sandforce makes a big deal about getting by without RAM so doesn't need back-up power capacitors, but frankly they have to multiple things in flight at any one time and doing a clean shutdown saving of the metadata in any circumstance would eliminate any chance of metadata corruption in those context. [ Again SSDs primarily designed for laptops, with batteries, don't have these kinds of problems at the same rate as a Mac Pro would. So a bit out-of-context application of part. ]
Just RAID 1 or 10 everything. It costs twice as much but it's multiple times more secure.
RAID 10 would cost 4 times as much, not twice as much.
I just wish SandForce would issue an official guidance on TRIM. I have been aware Apple engages TRIM on its SandForce-based laptops, as noted earlier here by Lou, but have not noted a consensus among pundits or users, or an official position.
This stated, my Intel 330s have been purrfect without TRIM enabled.
Another thing preventing me from trying TRIM enabler is the reported issue when multiple SSDs are running. I recall TRIM will only be enabled on one drive...
I have three SSDs running, trim enabler in enabled on two of them and disabled on the OWC PCI Express SSD with a SandForce controller.
I too havent had any issues with my OWC Mercury Electra SSD. It has been highly reliable experience!