Western Digital 'advanced format' promises slight boost in usable space
By Donald Melanson 
posted December 11th 2009 3:43PM
posted December 11th 2009 3:43PM
Sure, you may have been able to double your hard drive's storage space with DriveSpace back in the day, but it looks like Western Digital has now come up with a new, more foolproof way to eek out a few more gigabytes from its drives. Dubbed 'advanced format,' the company's new partitioning method promises to do away with each individual sector having its own Sync/DAM header and ECC, and instead move to larger, 4k physical sectors. As you can see in the helpful illustration above, that shift would also actually increase the size of the ECC, but WD says the net gain in usable storage space should still be between seven and eleven percent (which would certainly be welcome when we're talking about 1TB or 2TB drives). Unfortunately, you won't simply be able to update the firmware and reformat your current drive, but WD will apparently be moving its Caviar Green line to the new format in the "next few weeks," and eventually shift the rest of its drives over as well.
Is it going to be clear that these new drives have this increased space because of this sectoring scheme? Seems to me like they could just manufacture a drive with less density, format it with this structure to give it the same size, and then sell it for the same cost. Manufacturing cost would be less and WD could pocket the difference.
I see it unlikely that they'd advertise the drives like laundry soap as "Now having 11% more space"...
This is a step in getting closer to 750GB/platter. Expect 3 TB drives from WD in a few months
@Dudeman
Expect to curl up in a ball and cry when 3TB of pr0n crashes. RAID FTW.
There won't be any 3TB drives until they fix the stunning failure rates of the 2TBs
Some people don't seem to be clear on exactly what's going on here, so let me give a shot at explaining.
Let's imagine you're the drive controller and you've been given an instruction to write a block of data to a give location on the drive. We're going to skip over the head seeking to the correct track/cylinder and assume that that has been done already. Now the controller has to wait for the correct block to rotate under the head.
In order for a the drive to write the block at the correct location on the platter, it first has to have the disk head in read mode as it searches for the start of the block. Each block really contains several parts: a sync header, the data itself, the ECC (error correction code), and a gap. As the sync header goes under the head, the drive controller identifies which block it is. Once the correct block has been identified, the drive switches over to write mode, then writes out the data and the ECC.
Because of several sources of variability (motor spin speed, thermal expansion, etc.), the actual physical location where the block was written to may move around slightly each time that block is written. Because of that, the gap area is necessary so that when one block is written, there is no chance that it will overwrite the sync header for the next one.
By making the blocks bigger, they reduce the number of sync headers and gaps that are needed, allowing more space for actual useful data.
The drive controller CPU can make this change transparent to the operating system. Reads are unaffected, since the drive always reads extra data into its cache anyway. It's writes that may require changes.
If only a few 512-byte blocks within a 4K block require writing to, then the drive must first read the whole 4K block into its buffer, change the 512-byte blocks in the buffer, then write out the whole 4K block again. This is doing a read-modify-write cycle rather that just a pure write, and it can slow things down. That's why it's better if the OS also uses the larger block sizes.
As far as changing existing drives to use the larger block size, I imagine that it's possible, but it would require the manufacturer to develop very-low-level reformat software that would run on the drive controller CPU. This kind of low-level stuff is usually just done at the factory. They'd probably much rather it be done under the carefully controlled conditions at the factory rather than worry about all the possible ways it might go wrong at user's home.