(01-15-2020, 04:14 PM)wdt Wrote: OK, I am glad to say I was wrong, the uSD slot is 200? mHZwow, 900MB/s, what's the mode of operation it is? looks like the freshly invented SDe.
For your edification,, this with iozone on PBP (debian iozone does not want -l, BTW)
This, I think, is the fastest uSD I have ever tested, most will have 4k <10000,,,
crap cards 1000 or less (numbers are K/s)
Samsung EVO Select 64G
Oops, damn forum SW messed up my table formating, when I try to edit looks fine, sorryCode:random random
kB reclen write rewrite read reread read write
102400 4 17313 17328 649082 704138 651892 17466
102400 16 25964 17453 974273 810781 712337 18438
102400 512 20166 30826 921817 916929 1070827 18496
102400 1024 16714 29684 900426 891031 836840 30842
you probably forgot to assign the flag for the direct operation to the storage. and I forgot it too, and cannot check now. anyway, Rockchip makes the input frequency for the SDMMC controllers at 150MHz, so when possible, even SDR104 will be SDR75 only. 
which is not that bad.and honestly, while I don't insist, I can't get the hype about A1/A2 classes. that often are promoted almost like panacea. first, it's just markings without too much of explanation on what's behind it (the whole IOPS magic is pretty obscure) and second, flash based storages don't have such a drastic difference between sequential and random IO (tests kind of prove that). it's random inside anyway. also, for boot/start up time speed up, high sequential read speed is important too. it's advantageous not only for "taking selfies and filming stupid instastories", as A1/A2 believers often tell. but of course, if one has the the will to spend a little bit more, than the more expensive the better. if not fake.
ANT - my hobby OS for x86 and ARM.