OCZ Solid 30GB SSD Review

The OCZ Solid Series SSD is a low end product based on the JMicron JMF602B controller. I have a look at performance of this low-end SSD solution.

The SSD comes packaged in a typical OCZ box, in an anti static bag and with some manuals. There are no SATA cables provided nor an USB mini cable to interface with the USB port present.

The manual shows some specifications of the device, although these aren't related to performance but operating conditions and estimated lifetime of the device.

The back of the device mentions power consumption figures, which don't exceed 1.75 watts. Compared with an WD Scorpio Blue 320GB 2.5" HDD, it draws 0.75 watts less at peak. Drive capacity, as mentioned in the title is of 30GB.

The drive supports the SATA II interface and also brings an extra USB port so that it can be plugged to any computer. I have tried it USB style on the OLPC XO 1.5 and it worked perfectly, although limited at around 20MB/s.

The insides is where it gets interesting. The JMF602 rev. B chip controlling the MLC flash is driving 16 NAND chips. These chips are made by Samsung and are the K9GAG08U0M model.

These chips have the following characteristics:

• MLC NAND
• Data retention: 10 years
• Endurance: to be disclosed at publishing time, somewhere around 5k cycles to 10k cycles, although it should be closer to 5k which was what the most recent revision used.
• Each chip has 2.0625 GiB of capacity, with 64MiB serving as spare blocks
• Page size: 4KiB
• Block erase size: 512KiB
• Page program time: 800us
• Block erase time: 1.5ms

Drive manufacturers tend to miss providing specifications of page size and erase block size that might be otherwise useful when setting up the filesystem for optimal performance and endurance. Only NAND datasheet analysis can provide these numbers but they do so at the cost of a voided warranty, which was the case here.

The16 NAND chips have an aggregate of one gigabyte of spare blocks, which isn't that much these days when manufacturers are starting to allocate in excess of 25% of the capacity for wear levelling. Formatted capacity of the device is 29.84GiB, more than the announced 30GB(27.94GiB), but below the specifications that call for 32GiB(plus the hidden 1 GiB). This small discrepancy suggests that OCZ is allocating spare blocks to be managed by the JMF602 chip. This may bode well for long term reliability(the JMF602 provides both dynamic and static wear levelling, although OCZ doesn't mention which one is used), but do still try to use Windows 7 to keep the partitions aligned to 4KiB boundaries, which will increase performance by eliminating "read-modify-write" write amplification and assure a good lifespan of the drive.

The backside

Benchmarks

HD Tach read/write tests show a strange pattern in a drive that had a partition freshly allocated before the tests. The drive itself doesn't support TRIM and it's not handling well some past activity, meaning that sequential writes can go as low as 20MB/s. Seems the only choice might have been a secure erase, in which case it would've provided the 90MB/s we see in certain areas.
Average reads are good at 126MB/s, especially for a device of this class.

This IOMeter test runs 50% random reads and 50% random writes, over 30 minutes, with a queuing depth of 100 commands, somewhat higher than you can probably stress your system to do.

This is the same methodology as used in the 1TB Caviar Green review, which delivered the following performance figures:

• 87 IOs/s
• Average latency: 11.4ms
• Maximum latency: 57ms
• Max CPU usage: 1.3%
• Average write speed: 0.17MB/s

As for the OCZ Solid Series, Iometer shows us worse results in random write tests:

• 9.1 IOs/s
• Average latency: 109ms
• Maximum latency: 454ms
• Max CPU usage: 100%
• Average write speed: 0.02MB/s

These bad results in the random write test are caused by the lack of onboard cache, so much it is missing that during heavy random write scenarios the drive can stall for periods of up to 0.5s, 10 times more than a typical hard disk drive. Since Iometer only performs the test with RAW data on the drive, no partition alignment problems are influencing these results, which can only get worse if you don't do a proper alignment and "read-modify-write" patterns occur when writing data. 

Final words

It's not difficult to be disappointed with the OCZ Solid SSD series by looking at random write performance but there is a brighter side to this device. The drive is one of the cheapest ones available and does decrease the boot time of the OS in a big way since read speeds are above 100MB/s and read latency is flash type low. The only noticeable downside comes from heavy background disk activity - like windows indexing service - because otherwise the system does feel faster than with a traditional hard disk drive. It's also faster than available SD cards or USB pen drives even though the random write numbers tend to describe another scenario. A quick comparison of the XO OS booting on the USB connected Solid SSD vs a quality 4GB SD card showed marked improvements in browser startup time and webpage loading, which exercised random writes to disk.
Booting the Ubuntu Linux OS also showed a very good improvement in boot speed but OS updates were somewhat painful to endure.

The OCZ Solid Series SSD is available at Newegg for $125 and might be a first introduction to SSDs if you manage to find one at a lower price. With the Indilinx based OCZ Solid 2 Series starting at $119, the Solid 2 is a better proposal since it's free of random write speed problems and is available and currently selling for less.

If you have any experience with cheap SSDs to share with me, do drop by the forums or leave me a comment below.