OCZ SSD Caching? Good or Bad?

[nanaki333]

http://www.ocztechnology.com/ocz-synapse-cache-sata-iii-2-5-ssd.html

SATA III Solid State Drive / Caching Solutions / 80,000 IOPS / 64GB-128GB

The OCZ Synapse Series is optimized for caching applications to dynamically manage the Synapse SSD in conjunction with standard hard disk drives (HDDs), to provide users with SSD-level performance across the entire capacity of the HDD. The OCZ Synapse Series leverages the latest and most advanced hardware and software technology to deliver superior storage performance without sacrificing HDD capacity. For users craving the performance advantages of SSDs combined with the high capacities associated with HDDs, OCZ offers the first no-compromise approach to enabling high-performance and high-capacity storage for virtually any PC platform.

The OCZ Synapse Series integrates Dataplex™ caching software to dynamically manage the use of both SSD and HDD for superior overall storage performance. This combination creates an environment where the most frequently used "hot" data stays on the ultra-fast SSD, while the "cold" data remains on the larger capacity HDD (sold separately). Advanced caching algorithms learn user behavior and adapt storage policies to ensure optimal performance for each individual user, maximizing productivity for the most demanded programs and applications:

 

[Diogenes2]

According to OCZ, it looks pretty good.

I would look for some independent reviews, to get a better picture..

 

[nanaki333]

i'm more concerned with real world performance. all of ocz's drives looked great on paper, but real world users have a lot of issues with them. not exactly plug 'n play. if they make this idiot proof like intel rst + intel 311, then they'll have a winner. if they have to keep releasing new firmwares and software updates every few weeks, i can't see this catching on.

 

[sub.mesa]

Have a look at the datasheet:
http://www.ocztechnology.com/res/manuals/OCZ_Synapse_2%205in_Product_sheet.pdf

It lists the incompressible performance of this product:
Sequential read: 200MB/s
Sequential write: 70MB/s or 130MB/s (64GB versus 128GB)
Random read: 50MB/s (13.000 IOps) or 90MB/s (24.000) (64GB versus 128GB)
Random write: 70MB/s (18.000 IOps) or 130MB/s (33.000) (64GB versus 128GB)

While some advantage does come from compression, the official specs including those listing 80.000 are based on writing zeros which of course is complete bullshit. Also note that the listed capacities 64GB or 128GB do not include 50% overprovisioning if I understand the datasheet correctly; meaning that the actual visible size is 32GB or 64GB. This would be in line with the SRT limitation of maximum 64GB size.

For Intel Z68 SRT caching I would consider Crucial M4 64GB or 128GB if you want to have both OS and cache on the same SSD. The reason is not really difference in performance, which is negligible, but rather the fact that the Marvell NAND controller in Crucial M4 is much more reliable than the Sandforce controllers in OCZ' products.

 

[groberts101]

but rather the fact that the Marvell NAND controller in Crucial M4 is much more reliable than the Sandforce controllers in OCZ' products.

OCZ's products?.. you mean Sandforce, right? especially since despite the firmware or difference's in PCB layout and nand used.. they are all dealing with the same issues?

It will be nice if the new 2.13 firmware revision(and other mfgrs versions which mimic's of all these "superior controllers" disabling of partial slumber mode) in their firmware's will actually get to the root of the problems with this controllers dropped drives/bsod's(regardless of who slaps it into an SSD.

And then who's fault would it really boil down to if it does fix the drive drop-outs/bsod's? Sandforce for not disabling the sata 3.0 feature?.. or the others involved for not making the host support the feature properly?

Hmm?.. let me guess. Gotta be Sandforce, right?.. or NO!.. we all know it must be OCZ themselves who pushed the entire industry to circumvent the actual sata 3.0 spec's! Yep.. that sounds about right. :whiste:

Here's the little linky-link to learn about the power mgmt functions that sata3 was supposed to bring along with it(and which Intel pushed so hard to implement). Damn that Sandforce(and OCZ of course) for trying to comply to it! They should have just bypassed it altogether like the others did. lol
http://www.sata-io.org/documents/SATA-6-Gbs-The-Path-from-3gbs-to-6gbs.pdf

 

[sub.mesa]

Indeed Sandforce, but since virtually all of OCZ' current products are Sandforce, and of all the manufacturers shipping SSDs based on Sandforce OCZ is the biggest manufacturer, the term "Sandforce controllers in OCZ' products" is quite accurate I believe. Note that this may change, since OCZ acquired Indilinx some time ago and is working on their own controllers to replace the more expensive (and buggy) Sandforce controllers.

The problems with link power management is just one thing. The real problem is that Sandforce controller does not store the data you send to it, it only stores references to your data. If you store the same JPEG file 5 times, it will only be stored once on SSDs with Sandforce controller, using deduplication (and compression) techniques. This added complexity is probably at the heart of all the bugs that cause Sandforce SSDs to suddenly die and corrupt very easily. Especially since they are not shipped with a supercapacitor array like the Intel 320 does, which even with the recent firmware bug is still the most reliable consumer SSD available.

There is nothing wrong with DIPM or Link Power Management features themselves; those are very well-designed features that reduce power consumption of SSDs by up to a factor of 10, going from 0,6W to 0,065W.

 

[nanaki333]

Indeed Sandforce, but since virtually all of OCZ' current products are Sandforce, and of all the manufacturers shipping SSDs based on Sandforce OCZ is the biggest manufacturer, the term "Sandforce controllers in OCZ' products" is quite accurate I believe. Note that this may change, since OCZ acquired Indilinx some time ago and is working on their own controllers to replace the more expensive (and buggy) Sandforce controllers.

they already started with their new controller in the vertex plus. unless there's something else in the works they're not telling us.

 

[sub.mesa]

According to this dated review from Anandtech, the Vertex plus is nothing more than a pimped Indilinx Barefoot:
http://www.anandtech.com/show/4020/ocz-vertex-plus-preview-introducing-the-indilinx-martini

The real wait is for a competitive controller branded as OCZ after the acquirement of Indilinx by OCZ. The Everest controller has been named, though I've not yet seen it in the wild:
http://www.xbitlabs.com/news/storag..._Next_Generation_Indilinx_SSD_Controller.html

Regardless of future products, virtually all current OCZ SSDs are Sandforce based. So if you want to avoid Sandforce due to their high failure rate, do not buy OCZ products. Crucial M4 and Intel 320 should be decent alternatives.

 

[groberts101]

while I would agree with some of that.. I still see people buying the P outta' those barefoot drives. So I would guess it's probably 80/20 for OCZ's SF/Indy sales proportions.. but.. probably close enough to validate your remark, I guess.

The fact that APM/PTS being disabled is another matter altogether. Surely there's a reason behind Intel's decision to disable it on their drives?

Then there's the question of.. why did SF take so long to disable it for their firmware?(regardless of vendors using thier chips).

Like I already said more than a few times.. and just you wait and see.. Intel is just getting all their ducks in a row for the Cherryville release. Things are magically being sorted leading up to that release and it's far too much of a coincidence when a company like that is involved. Conspiracy theories.. or not.. if the shoe fits.