>-----Original Message-----
>From: Peter Humphrey <pe...@prh.myzen.co.uk>
>Sent: Thursday, January 4, 2024 5:01 AM
>To: gentoo-user@lists.gentoo.org
>Subject: Re: [gentoo-user] LiveGUI USB Image
>
>On Thursday, 4 January 2024 02:48:13 GMT Adam Carter wrote:
>> > > dd if=/path/to/iso-image of=/dev/sd? bs=4M status=progress
>> > >
>> > > Replace the obvious bits.
>> >
>> > I've tried a few values of block size over the years, but so far I
>> > haven't noticed any difference. I haven't run any proper tests though.
>>
>> I think it's just that the default blocksize is (or was) very small
>> (512
>> bytes?) so setting it to anything non-small helps a lot.
>>
>> eg one example (from
>> https://superuser.com/questions/234199/good-block-size-for-disk-clonin
>> g-with
>> -diskdump-dd#234204) seems to show that most gains are in by around 16k.
>> There's probably a lot of testing noise in these results.
>>
>> $ ./dd_obs_test.sh
>> block size : transfer rate
>> 512 : 11.3 MB/s
>> 1024 : 22.1 MB/s
>> 2048 : 42.3 MB/s
>> 4096 : 75.2 MB/s
>> 8192 : 90.7 MB/s
>> 16384 : 101 MB/s
>> 32768 : 104 MB/s
>> 65536 : 108 MB/s
>> 131072 : 113 MB/s
>> 262144 : 112 MB/s
>> 524288 : 133 MB/s
>> 1048576 : 125 MB/s
>> 2097152 : 113 MB/s
>> 4194304 : 106 MB/s
>> 8388608 : 107 MB/s
>> 16777216 : 110 MB/s
>> 33554432 : 119 MB/s
>> 67108864 : 134 MB/s
>
>Interesting. I think I'll stick to my usual 64MB block size.
>
>--
>Regards,
>Peter.
So, basically the default is 512, and if you're copying to something where the
hardware block size is larger than that (4096 is common on a lot of things
these days) then every block on the physical device has to be rewritten
multiple times. (Eight times in the example case.)
Needless to say, that's... rather slow. Also rather hard on your device.
Once you get larger than the physical block size, as long as you have an even
multiple of the physical block size, then any additional gains come solely from
the IO scheduler being able to make slightly more intelligent choices with
larger chunks.
LMP
