Intel 520 Series SSD Full Review - SandForce on Steroids?
Introduction, Specifications, and Packaging
Today we take a look at Intel's newest 6Gb/sec SATA SSD - the 520 Series. This is the second non-Intel controller to appear in one of their products. The first was the Marvell controller, which appeared in the 510 Series last March. This time around, Intel has gone with SandForce. This should leave at least one SATA 6Gb/sec model to be released. Taylorsville is the code name for the next SATA 6Gb/sec native-Intel controller, which has been on their roadmap since mid-2010 but has yet to actually materialize. While Taylorsville development continues, Intel has stop-gapped the 6Gb/sec slot with the 510 and now the 520 Series. Intel seemingly worked wonders with the stock Marvell firmware, and while the Marvell controller was much improved over stock, it still lagged far behind other higher performing SATA 6Gb/sec solutions. The SandForce was one of the much more capable controllers eating the 510's lunch, but how much further could Intel improve upon the SandForce firmware?
I guess a good question to answer up front is - What took them so long?!?! The answer is a bit complicated. Intel has actually been working on getting the 520 out the door for over a year now. They had to start with the same base SandForce firmware but accomplish two things for their version to be successful:
- Optimize to perform better than other equivalent SandForce models (from competitors).
- Pass Intel's stringent validation testing.
They didn't say so directly, but I can only imagine Intel's process was plagued by multiple 'back to the drawing board' moments. Trying to one-up competition like OCZ can't be easy as they've been tweaking SandForce firmware since the very beginning. There's also those nasty bugs that would cause random BSOD's or even permanently brick the drive. Such failures have no place in an Intel SSD. Intel's upper limit for each SSD line is a 0.75% annual failure rate, and we've seen SandForce SSD's failing at a higher rate than that this past year.
With each tweak made, Intel would have to once again pass their drives through another round of full validation testing. This is no small task for Intel. As an example: It took Intel just a couple of weeks to recreate and correct the long-term performance issue I discovered back in 2009, but despite mountingpressure, they could not release the updated firmware until it had successfully passed their validation a full three months later. Intel takes this testing very seriously, and that's what leads people to trust their reliability.
Intel clearly takes their performance specs seriously. Their spec sheet for the 520 Series contains two full pages of IOPS and throughput figure tables. Instead of cutting and pasting a huge swath of data here, I'll link you to the 'Intel SSD 520 Product Specification' PDF over at the Intel site, and bullet some of the highlights below:
- Capacities: 60/120/180/240/480 GB
- 7mm and 9.5mm thickness available (480GB capacity only available in 9.5mm)
- Power: 850mW active / 600mW idle
- AES-256 encryption - can be enhanced by use of ATA drive password
- 550MB/sec sequential reads / 520MB/sec sequential writes
- 4KB IOPS: 50,000 read / 80,000 write *"out-of-box SSD"* ('used' figures in the 50,000 range)
- SSD 520's are *not* validated for data center usage (though I don't think it will stop anyone)
I'd like to congratulate Intel for being so forthcoming in their specs. They go out of their way to provide figures for both compressible and incompressible data. Such data is vital to those needing it - as SandForce controllers are among very few that can compress data on-the-fly, and they perform very different with each data type. The Intel specs further break down the incompressible figures, providing figures for both Iometer and AS-SSD benchmarks. This is a far cry from the *one* IOPS figure you see listed on most competitors spec sheets.
I should note that as with other 480GB SandForce-driven SSD's, *any* SandForce 2281 SSD with a capacity greater than 256GB will see a dip in 4K IOPS performance. This is due to the way the SF controller handles the mapping of LBA's. To double capacity from 240GB to 480GB, the SandForce controller's finite number of allocations must be reconfigured to utilize 8KB blocks (up from the standard 4KB - intentionally matched to the NTFS 4KB cluster size). This negatively impacts IOPS performance as a 4KB random write translates to the equivalent of an 8KB random write once the added overhead is taken into account.
This is also clear in the Intel specs, as incompressible 4KB writes see a near 50% cut when comparing the 480GB to the 240GB capacity points.