Log in

View Full Version : Optimal NTFS cluster size for DVD-RB?


raddygast
16th March 2005, 07:53
Here's a question I don't think has been asked here before.

I have a 160GB external drive (using USB2.0 at the moment though it can also do firewire). Doing most of my DVD processing work on that drive.

What would be the optimum NTFS cluster size? I have just reformatted it with 64KB clusters, but it strikes me that with all the tiny files that RB sets up in the Working directory, I may end up wasting enormous space. Then again I usually delete the working directory pretty quickly after a successful rebuild.

I just thought 64KB clusters would be optimal when most of the file space is taken up by ISOs of 4GB to 8GB, and large (1GB average) VOB files. Then I realized that DVD-RB does create a crapload of tiny files while it works, so what cluster size do you guys use?

mrslacker
16th March 2005, 17:29
I use the windows default. It's either 512 or 1024 depending on the volume size, etc.

DVD-RB creates some small files, like the avs files, but not that many. Although, many FLG files would be bigger than these block sizes. However, just pretend you are using a 64k block size and you mainly have files >64kb. Or say you have files that are 65k, 129k, 193k, 257k, etc., or any file with 1kb more than a multiple of the block size. Say goodbye to that 63kB. That's a wasted space perspective.

Maybe you will have better performance with a larger block size, but you will waste more space regardless of what your file sizes are. If you want to accomodate to larger file writes, try a 8k or 16k block size. You might want to search elsewhere to see how windows and NTFS get allong with larger block sizes. I suspect the drive's interface and buffer will be your limitations. USB 2.0 is way too slow to worry about tuning the fs, IMO. In reality, your enclosure probably won't top the drive's capabilities. Run HDTach from tcdlabs.com to be sure. You will need the PRO version or another program to test write speed. You might be shocked at how slow it really is.