Firmware Can’t Fix Everything
Despite the massive improvements seen with the new firmware, there are a few things not yet rectified.  Lets get these out of the way before proceeding further.

The MAC thing

Some MacBook users have reported their X25-M failing to cooperate with Bootcamp.  This is caused by a timing issue related to the SATA handshake process that does not play nicely with some of the newer MacBook hardware.  Low level handshake timings are only partially controlled by the user-flashable firmware.  The earlier part of the power-on process is controlled by the bootloader, which is unfortunately not end user flashable.

There is a very narrow slice of drives affected by this issue.  It really boils down to how much of an early adopter you were.  While all current (as of this writing) X25-M drives have shipped with the same 8610 firmware, the bootloader ROM *was* revised about a month into retail sales.  The 160GB model is not affected, as it went into production with the newer ROM.  This is a very rare timing issue that depends on variables other than just the bootloader.  It is entirely possible you will never be affected, even if you own a MacBook and a first run X25-M.  Intel is making things right and encourages those effected users to contact their support team for resolution.  One of our pre-production review units showed this problem, and the 8820 firmware did not correct the issue.

IOPS

This is not a new issue – It was initially masked by the fragmentation slow downs noted in my last article on the matter.  Now that write speeds are relatively constant, we can see that a ‘new’ X25 (i.e. HDDErased) performs significantly better in terms of IOPS than an X25 that has been subjected to a single sequential write pass.  Even though no fragmentation is introduced by this process, it increases overhead, as the X25 has to track every LBA that is written.  They are only tracked from the first write on, meaning that if you never write to an area, it never has to track it.

I conducted some rough testing and saw that IOPS performance of a completely written-to X25 dropped to about 60% of a freshly wiped unit.  It is not as prevalent in real world usage, as my IOPS testing focused on smaller files.  Don’t be afraid to ‘use’ your X25 on account of this.  Many solutions are on the horizon that will bring IOPS back towards ‘new’ condition.  The key is the TRIM command, which once implemented will enable an X25 to purge entries from its LBA remap table, reducing overhead and restoring IOPS to its former glory.  For those curious, the 8610 and 8820 firmwares do not yet implement TRIM.  Intel will likely hold out until the release of Windows 7.


« PreviousNext »