What's the benefit of higher allocation unit size for NTFS?

Less fragmentation. Set it to 64k.

Higher sizes give slightly higher read/write speeds at the cost of hard disk space.

For instance if you use 8KB block sizes then even if you only need 4KB to store a file the disk will still reserve 8KB so you wasted an additional 4KB to store that 4KB.

File: Untitled.png (9KB, 1305x673px) Image search: [Google]

9KB, 1305x673px

>>60680174
this

>>60679240
It lets you access files quicker but you need more ram and a bigger cpu to do it.

>>60680414
Are the requirements slightly higher or is it something to be concerned about?

>>60680693
there are a lot of things to be concerned about if you actually use NTFS

>>60680414
>>60680755
troll

Just use the default.

>>60680831
I work with lots of video files, tens of gigabytes in size. I figured the 64k allocation size would be better for this. No?

Does this shit even make a noticeable difference on 2017 hardware? I could see this meaning something over ten years ago but not today.

>>60680908
In your specific case 64K would be ideal. The amount of data wasted (up to 64KB per file) is nothing.

>>60680989
don't do this on the drive with the OS

>>60681064
Oh yeah definitely, this is strictly for storing data. Operating systems have tons of files and the amount of waste would be huge.

>>60680971
It does somewhat. Performance-wise I don't really know but larger volumes usually require larger block sizes. In certain situations smaller block sizes are more ideal for making use of your storage drive. Larger block sizes really only work well with a relatively small number of files.

>>60681328
Or if you have a file server that serves primarily large files.

>>60681096
There is the other issue that while the speed and efficiency of reads and writes goes up as allocation block size increases, Windows 7 and below cannot natively read anything @ 4096 or higher without issues.

AF format drives use hybrid allocation sizes, the OS interperates it as one unit, the drive actually utilizes another block size, and windows for a long time was not coded to allow for this. Drive firmware cannot make up totally for the way the OS checks file integrity, OR the way it stores location data based on sector/cylinder.

Running 4096 (AF drive Default, which is most drives over 160GB) or larger sectors will totally fuck up the boot loader, and the OS ability to run block level OS file verification. It breaks Windows Update, and it breaks dr.watson/SFC abilities.

To get around this, it has to have a hand off boot partition (which it makes as 100MB)

To get around this hand off partition, and to get around some of the issues with the way it self verifies Core OS components, you need to run 512 sectors.

If they are running windows 8.1 or newer then i am pretty sure the OS has finally be updated to account for these issues, but i am not sure.

>>60679240
Depends what you plan to store;
4k (default) for system drive and swap file
16k for shadow copy partition
64k for databases and virtual drives

File servers should be benched to determine optimum size though I usually default to 8k. 64k and a 1byte file will eat 64k of disk. So know what you plan to store.

If you set it to 64k, will that even make a noticble difference with modern CPUs and ssds?

>>60682349
You technically want the file system and underlying RAID if you have one to be optimized for whatever load the filesystem will be under. For example, if you typically pull 256k blocks of data in a read or write, it makes sense to have 64k blocks and 4 disks with a 64k stripe underneath as when the application does its typical read or write of 256k, the block will be broken up into 4 x 64k and engage all 4 RAID disks thus increasing IO and throughput. Plus the filesystem will write that data in 64k blocks which is the best outcome in our scenario.

Writing an 8k block of data results results in only 1 disk of the array being used, which could be advantageous too if there's lots of requests but inefficient for the file system as everytime it reads or writes it processes data in 64k chunks even when there's only 8k.

SSDs are rated at 4k typically, but same rules apply.

You should leave your system and swap file at 4k if using Windows else you can't do the RAM dump or write a swap file elsewise.
Major application software like SQL and Exchange recommend 64k and I use 64k for any huge file accesses like backups too. Else it's benchmarking and monitoring your disks to see what kind of IO you are driving.