Featured News: Page (1) of 1 - 12/02/14 Email this story to a friend. email article Print this page (Article printing at MyDmn.com).print page facebook

Some SSD questions answered

Talking about TRIM with OWC's Lawrence O'Connor By Ko Maruyama
As we update our Macs to Yosemite and better SSD drives, the community of Mac users have been talking about what to do, not to do, upgrading, and general confusion - especially when it comes to looking at SSDs.
I had a chance to ask Larry O'Connor to dispel some of the myths surrounding SSDs for our creative workstations. 

Here are his answers:

#1 Drives operate just fine without TRIM
The vast majority of SSDs have a significant dependence or, at least, a benefit from OS side TRIM. These drives operate better when the OS is managing where the free space is on the NAND as opposed to being designed to excel doing such management independently.

 This is pretty much the case with nearly every SSD on the market today with the exception of those with LSI/Sandforce controllers. While LSI/Sandforce based drives are happy to take TRIM commands from the OS - we have done extension real world testing and on going real world studies (drives in various operational use throughout all departments going on 5 years now) that support our statement that our drives not only operate just fine without the use of TRIM, but also without any declination of performance and without an increase in NAND wear/write amplification.





When we got into the SSD space - it was because here was a controller solution that negated the need for OS side TRIM... and note that this was at a time when there was absolutely no TRIM support under OS of any kind... nothing to hack on, nothing even for Apple drives. And back in 2008, 2009 - SSDs of that time quickly slowed down and expressed other issues under normal work loads that made them amazing in the short run - but slower than a hard drive after a period of months if not weeks depending on how the system was used.
TRIM is a kludgy crutch and Sandforce designed, from the beginning, their processor to be be independent of TRIM.
In hind sight - it's a good thing Apple took so long to support TRIM even for their own drives. It was a defining reason for our selection of Sandforce and one we'd do again -in hindsight - even if TRIM was not a factor. While we're a little frustrated with the next generation processor time line from LSI/Sandforce, the current processor is still a true real-world leader.


#2 - Benchmarks are all telling - not.
Tweaktown recently did some additional coverage on the subject of benchmarks and SSD performance highlighting an Intel SSD. Like us, Intel widely uses the Sandforce processor in their own drives. What was perplexing to them (never to us) was that while LSI/Sandforce drives like ours really don't excel in typical benchmarks - even so called real world tests - in real world use and even during testing - they 'feel' faster and appear to get work done faster. Well - they are faster... while most other drives have more of a 'bath tub' benchmark result across a broader workload,

LSI/Sandforce based drives maintain a substantially more consistent performance across with overall higher real performance. If your system only did certain things and/or only did one thing at a time - we'd lose... but that's not how computers work and where the heavy lifting is done - where real work flow goes on - our drives kick butt and take names - as do Intel's.

The easiest analogy - there are plenty of cars that are run away from the rest on a straight away... but roads have curves and typically more curves out there than straightaways. Our LSI/Sandforce based SSDs- are built right (and we totally respect Intel's product and glad to compete with them) to hold tight and speed away through the curves and still plenty of straight away speed.



#3 - Same processor inside, same drive, same performance - not.
There is a choice of NAND configuration, in addition to the processor, that makes a huge difference. A competitor got angry at a review recently because they claimed their drive was exactly like ours. Well - building with the same processor doesn't make that so. Most manufacturers build each capacity in a way that makes them most economical to build. Typically means even using the same NAND chips across all the drives - just more for each high capacity. We build each capacity based on the config that delivers a desired performance profile. It's more work and a little higher cost - but it's worth it and the difference shows up. Tweaks in the firmware parameters also play a part. Two drives can look identical spec wise, but behave very differently.

BONUS CLARIFICATIONS:

Drives that need TRIM just operate slower without it?


There is more to it than that... not only do they operate more slowly - especially in some load situations - they also are operating with reduced wear leveling efficiency. This can mean they are experiencing much higher write amplification (how many times the data has to be rewritten to complete a single write) or even doing more wear on a specific NAND location as opposed to effective wear leveling. Minimizing Write Amplification and maintaining good wear leveling critical to the life of an SSD. Sandforce/LSI SSDs - with TRIM utilization at all - are able to keep write amplification under all others by a wide margin.

TRIM works on drives in hardware raids. No

Drives in a hardware raid receive no trim commands. Hack or no hack. Mac or PC. Got no TRIM available when a hardware RAID is used with SSDs. Makes LSI/Sandforce drives heads and shoulders above for such use.

Apple supported 3rd party drive TRIM enablement and then abruptly reversed that support with 10.10 Yosemite - False

Apple never supported TRIM for 3rd party drives and applications that have enabled this TRIM for non Apple drives have effectively been hacking an Apple KEXT, a driver in the OS. What Apple did with 10.10 is add a security check to installed KEXTs. This prevents malicious hacks from impacting a user via driver modification - but also results in a modified driver as being deemed compromised and in the case of the TRIM hack, results in the system to halt and not continue to boot. There is a work around for this - but it requires this security to be disabled and this also is a PRAM stored parameter that resets when PRAM is reset and then puts you back where you were.

Again - while we, being honest, get some benefit from what Apple has done in 10.10 and with making a HACK the only way to enable TRIM in the first place on OS X for non Apple drives, I disagree in general with the closing of various avenues within the OS that create barriers to a customer's choice in solutions as well as impediments in general to other potential enhancements beyond just this SSD impact. That being said - I do understand Apple's position and their goals. These are the rules we must work within and that's what we all have to do.

------------------------------
Here are some things that I knew before talking to Larry:

All OWC Mercury SSDs have a SandForce Processor installed. SandForce processors improve an SSD performance and reliability. A key feature of SandForce processors is the DuraWrite technology. DuraWrite is a technology that helps reduce write amplification. As discussed earlier, write amplification is the act of data on the SSD being moved from block to block. Write amplification is expressed as multiples, based on how many times the data has been rewritten since it was passed to the SSD. This equation is calculated by taking the total amount of data written, then dividing it by the actual size of file. For example, a 2.0MB file is written to the SSD. Through garbage collection, it had parts of the file rewritten around the SSD totaling 5.0MB of data moved. The write amplification would be expressed as 2.5x.

If you have more questions, be sure to check out the blog entries about TRIM at OWC blogs.  There's a bunch of great information there, but sometimes we just want to be able to buy a good drive that we know will work.  The URL for that is easy:

Find your studio rigs and find out which is right for you http://eshop.macsales.com/shop/SSD/OWC

http://blog.macsales.com/?s=TRIM


Page: 1


Ko Maruyama is a freelance animator in Los Angeles.  In addition to working on film and broadcast animations, Ko teaches at Pasadena's Art Center College of Design - focusing on motion design.  When working, writing or testing software allows, you can find him lending a hand in the After Effects board and lurking among the Cinema4D, Visual Effects and Photoshop posts within the DMNForums.
Related Keywords:apple, SSD, OWC, macsales, Sandforce, TRIM

[ServletException in:/common/ads/links.jsp] The absolute uri: http://java.sun.com/jstl/core cannot be resolved in either web.xml or the jar files deployed with this application'

Our Privacy Policy --- @ Copyright, 2015 Digital Media Online, All Rights Reserved