Question of FIO benchmark

[learner.study]

Hello:

I'm using FIO benchmark and would like to know if buffered IO is write
thing to do for sequentail-write kind of test.

If I specify "direct=1", I see seq-write performance as 60MBps whereas
with direct=0, I get 500MBps.

I'm using the following options of FIO:
./fio --name=global --runtime=60 --rw=read --bs=32768 --direct=0 --
ioengine=aio --iodepth=8 --name=job1 --filename=/dev/sdc1

Can someone let me know if doing write=0 is the correct thing to do
for seq-write experiment and why the difference in performance is 10
times!

Thanks!