What kind of sustained transfer speeds are you getting?

#1

I have 4 1TB SATA drives on my SATA HAT. Doing an SMB transfer over GBE, I find that the transfer speed bounces anywhere between 56MB/s down to really low single digits or less when copying large files to the Pi.

ATOP shows two of the drives getting super busy but the other two gets barely anything. (see https://imgur.com/a/CaD7FSu)

I’m running Raspbian Buster, created the array with mdadm, running an instance of Nextcloud, although for this testing, I’m connecting with smb to do the copy.

Any ideas how to troubleshoot this?

#2

The SATA HAT is connected via USB and even without the SATA HAT, my Raspberry 4 has a very uneven performance on the USB 3.0. A connected SSD fluctuates between 10 and 68MB/s in SMB transfers. On average I get 19MB/s. If I connect the same SSD to my RockPi4 and do the identical test, I can manage max 117MB/s with an average of 98MB/s.
Maybe there are optimizations that you can set in Raspbian to make the USB more stable and / or faster. I use UbuntuMATE Groovy on the RasPi and the newest armbian on the RockPi. It feels like worlds lie between the two devices. I’m really annoyed to have bought the Raspberry, a second RockPi4 would have been better.

#3

Search raspberry forum for quirks setting and disabling UASP

#4

Thanks, that helped a little. But the USB hub does not come close to the performance of the RockPi 4.

#5

What’s your RPi4 Spec? (RAM, …)
Temperature of the RPi4? They tend to become hot and go into temp throttle mode
Power supply your using?

I 'm currently working on a custom casng based on the RPi4 / 4GB but with heatsink + fan on the CPU; I haven’t come into any issues regarding speed or whatsoever unless the brief period I was using my fixed 12V output from my benchtop labo power supply (only 500mA current) Everything worked untill I started doing (large) transfers on the disks, up to complete craches.

#6

I’m using the 4GB model. I have the Radxa kit that included a top board and case. Throttling is certainly plausible, since the air intake is so choked. I want to take a drill to the base and make the holes bigger at the very least.

I’ll run some tests to measure how hot it’s getting.

Update: Ran a test, monitoring the CPU temp with vcgencmd in a script. Temps were between 45C with occasional spikes of 50.6C, so it doesn’t appear to be thermal throttling.

#7

if you can, check the drives alone. Maybe the issue is the drive. If one is not OK than it can be that the next in the row makes troubble too

#8

Alright, I’ll check each drive individually.

#9

I don´t think that you have a thermal issue.

I use a passive case for a overclocked Raspi4b 4gb and 60°C is max i get under real stresstests.

Bildschirmfoto%20zu%202020-12-05%2019-12-35
Bildschirmfoto zu 2020-12-05 19-12-35.jpg1920×1080 726 KB