LSI Releases Code To Manufacturers – New Increased Capacity ‘SandForce Driven’ SSDs Hit The Streets

Up until this morning, the most recognizable characteristic of a ‘SandForce Driven’ SSD was the fact that it’s capacities were always 60, 120, 240 or 480GB, unlike the standard of 64, 128, 256 and 512GB.

This was a strong clue which supported our belief that LSI SandForce and Intel had teamed up for manufacture of their enthusiast level 520 Series SSD way back in  April of 2011. To put it lightly, our publishing of that article met with some very strong views and opinions.

The key to ‘SandForce Driven’ SSDs was the fact that, regardless of the manufacturer utilizing the SandForce processor in their SSDs, the capacities were always set simply because a specific amount of NAND flash memory had to be set aside for firmware needs and over provisioning.  This was typically 7% of the total for client SSDs or 28% for enterprise.

That was then and this is now.  This morning, ADATA has released the first ‘SandForce Driven’  SSD that will provide the consumer with 7% added capacity, before formatting, now equaling the capacity points that we typically see in SSDs that do not use SF processors in their storage.  These include Crucial/Micron, Intel previous to the 520 Series, Samsung, Toshiba and limited SSDs using JMicron and Phison controllers.

The question with this change in LSI SandForce capacity now becomes one of whether there would be any change in lifespan or performance to the drive as the over provisioning seems to have been eliminated.  In addition, SandForce has not introduced external cache memory as the others do which makes this even more confusing.  How do they do it?

OUR CONVERSATION WITH LSI

Given the complexity of this new ‘capacity ideal’, we contacted LSI and posed our questions to Kent Smith, Sr. Director of Product Management Flash Components Division at LSI Corporation, Kent unquestionably being one of the most knowledgeable in the field today.

DRIVE CAPACITY

Kent immediately reminded us that SSDs traditionally have a stated user capacity that matches the same physical flash capacity of 32, 64, 128, 256, 512GB, etc. If, however, the SSD processor can support three or six channels (as LSI SandForce processors can), other configuration options such as 24, 48, 96, 192 and 384GB can be supported as well.

OVER PROVISIONING

Our next question spoke to the lack of over provisioning of the new drive and Kent asked me to grab a calculator.  I knew what was coming.  While on the surface, the industry calls this 0% over provisioning it is impossible because, in reality, all SSDs need some level of over provisioning to map out bad blocks and to move data during garbage collection and wear leveling. What is really happening behind the scenes of all SSDs is the difference between Billion Bytes (10^9 = 1,000,000,000) and Gigabytes (2^30 = 1,073,741,824). That delta is really 7.37% of the physical capacity.

So when you see a drive at 120GB of 128GB of physical flash you are thinking it is 7% OP, but in reality it is more like 7% + 7.4% or 14.4%. Because the translation of Billion bytes and Gigabytes was made by HDD manufacturers some time ago, most users dont think about that difference.

Until the SF-2000 series processor family, ‘SandForce Driven SSDs could not support an OP less than 7% (120 of 128 GB for user) so you never saw the typical 128GB ‘SandForce Driven’ SSD or any of the other binary capacity points. New firmware for the SF-2000 processor family now permits manufacturers to create SSDs at these full binary capacity points with what is commonly known as 0% OP.

SANDFORCE PERFORMANCE WITH NO ADDITIONAL MEMORY

There has to be a trick here right?  Well…not really.  No extra flash is required to support these new capacities, but it is important to note that, given all other conditions are the same between two drives, one with more OP will generally perform a bit higher in random write speed than an SSD with less OP. With a SandForce Driven SSD, you would see this more with high entropy data or a system without TRIM support. If you have the more common lower entropy data or TRIM support this 0% OP has less impact on random write performance.

NO NEED FOR EXTERNAL DRAM CACHE

Further, there has been no addition of DRAM to make up for the lack of OP and SandForce does not simply include a DRAM module inside their chip. Rather, the processor uses a unique method to hold the map information (the purpose of the DRAM on other SSDs). Actually,  by not having an external DRAM module, SandForce reduces the total space requirement for its components, allowing further density reductions, and resulting in less total power draw by not powering external DRAM.

FINAL THOUGHTS

First and foremost, we appreciate LSI and Kent taking the time to help us along with our questions and providing such thorough reply.  All appears to be business as usual with SandForce processors, with or without the extra over provisioning, and this seems to be a great release by LSI, as most understand just how valuable available SSD capacity truly is.

With respect to performance, AData has published their own results within their specifications.  Based on their own data, there has been no performance loss when comparing these to other ‘SandForce Driven’ SSDs with additional over provisioning.  We hope to get a sample in our hands soon enough for an update to this release.

2 comments

  1. blank

    You should probably clarify that the new firmware is NOT backwards compatible with existing Sandforce SSDs. New models only………. too bad

    • blank

      The article states that this applies to new releases and, although the title speaks of code being released, I don’t think this fits into a simple firmware upgrade or downgrade at all.

      It definitely does not apply to a firmware fix for present SF SSD owners.

      Thanks ahead.

Leave a Reply

Your email address will not be published. Required fields are marked *