Hi, I’ve perform some performance test on my v1.2 version rockpi. And in some tests I see that SSD connected with USB has higher writing speed:
/media/2tb - mounted device on USB
/media/clouddrive -SSD connected with extended board
Same manufacturing of SSD, but one has 1tb, another 2tb.
Interesting if someone perform tests on sata extender as well?
This is not the normal speed, please confirm whether the power supply is enough.
@setq
More than enough. I’m using original cable and adapter. When no loading it takes 9v and 0.280A of power. During this test it rise to 0.400A and some times to 0.800A.
I’ve also tried with enabling PCIE 2gen from https://wiki.radxa.com/Rockpi4/dev/libmraa#Install_essential_packages . But still no results.
Next probably I will try reconnect SSD.
Is iozone telling you the same?
Real world comparisons:
@igorp with iozone it looks near same results. But it far from expected values. In many parameters there similar results.
'**SSD via nvme** Command line used: iozone -a -s 1G -r 1024k /media/clouddrive/mnt/`
' random random bkwd record stride
' kB reclen write rewrite read reread read write read rewrite read fwrite frewrite fread freread
' 1048576 1024 37098 39140 1210040 1528877 1516556 49675 1503168 43507 1451226 44287 39881 1432341 1531885
'
' **SSD via USB** Command line used: iozone -a -s 1G -r 1024k /media/2tb/mnt/
' random random bkwd record stride
' kB reclen write rewrite read reread read write read rewrite read fwrite frewrite fread freread
' 1048576 1024 35029 43690 1392640 1479077 1537154 43400 1309000 41388 1078389 42734 43170 1385460 1544194
And results really not same with SSD test on my PC.
Afte adding oflag=direct to dd command, I got higher result then original :
'**SSD via nvme** dd if=/dev/zero of=/media/clouddrive/tmp bs=100M count=100 oflag=direct; rm -f /media/clouddrive/tmp
'100+0 records in
'100+0 records out
'10485760000 bytes (10 GB, 9.8 GiB) copied, 37.7723 s, 278 MB/s
'**SSD via USB** dd if=/dev/zero of=/media/2tb/tmp bs=100M count=100 oflag=direct; rm -f /media/2tb/tmp
'100+0 records in
'100+0 records out
'10485760000 bytes (10 GB, 9.8 GiB) copied, 52.7008 s, 199 MB/sDon’t have NvME addon for Rockpi, which is why I tried on most similar board I have, Nanopi M4 which only uses 2 PCI lines:
random random bkwd record stride
kB reclen write rewrite read reread read write read rewrite read fwrite frewrite fread freread
1048576 1024 497044 890079 1589469 1664055 1657840 959355 1597012 1118415 1668736 758941 739922 1610518 1671279
I would say expected numbers for Samsung SM961/960 PRO 256Gb http://ix.io/28BD This is also the only spare drive I have.
Below is my x86 workstation running preview of Ubuntu desktop 20.04 with the same NvMe just 512GB in size. Does this looks normal? 
random random bkwd record stride
kB reclen write rewrite read reread read write read rewrite read fwrite frewrite fread freread
1048576 1024 3009473 5714058 13608162 11424007 14298600 5703742 13671657 9139751 14379382 4917607 4911407 13871346 14319922@igorp I think in my rockpi there present power problems, as problem with freezing, which I mentioned here USB connected SSD leads to feeeze still valid.
Probably I will try to power-up with differ way, throw GPIO pins. (I will research if it’s possible)
If you don’t have a proper USB-C PSU - I am using this one https://amzn.to/2uCDbLP Its an overkill for this but I also charge or power my x86 Laptop with it. Alternatively, you can use such adaptors:
https://twitter.com/armbian/status/1179762024357740546 and some regular 5V PSU witch can secure 3+ amps.
Also eliminate software and try with Armbian, both kernels.
@igorp I have original fast charger 9v/3A, Also tried with differ one 9v/2.5A.
So I expect problem not related ti adapter itself, but to board.
Software support is most critical component. Did you tried the same on Armbian?
@igorp finally Ive test with few OS. But looks problem really HW.
So lets wait for response @radxa team.
@jack maybe you can support as well?
Up!
Can someone test SSD speed on v1.2 board?