Jump to content

Quick review of Solidrun's Clearfog

Igor
 Share

Recommended Posts

Igor

Igor

Posted
Posted

r1_and_clearfog.jpg

IMG_5942sm.jpg

 

Clearfog PRO with R1 in the back

 

After a day of playing, I am running full featured Armbian, Debian Wheezy, built with Armbian script from sources. We had to fix few things in U-boot and upgraded kernel to .82 but generally it was easy to boot the board. Of course SolidRun also provide some basic images.

 

Three(3) phys, USB and PCI is detected in kernel 3.10.82 - I don't have much hardware to plug in so I can't do much tests at this point. There is no CPU governor (yet?), so I assume it's running full speed (1.6Ghz) all the time. There are dip switches on the board to set cpu / dram speed ...

  ____ _                  __             
 / ___| | ___  __ _ _ __ / _| ___   __ _ 
| |   | |/ _ \/ _` | '__| |_ / _ \ / _` |
| |___| |  __/ (_| | |  |  _| (_) | (_| |
 \____|_|\___|\__,_|_|  |_|  \___/ \__, |
                                   |___/ 
Welcome to ARMBIAN (Debian wheezy 3.10.94-marvell)

Last login: Sat Jan  9 10:11:28 2016 from desktop

Load: 0.00, 0.01, 0.05 - Board: 64.6°C - Memory: 973Mb

root@armada:~#

Waiting for:

I decided to drop idea to put an AC card into the board since it's simply too problematic and expensive (66USD). This card doesn't have support in legacy kernel, in 4.4 have no idea if everything else already works + I have only one AC device in the lab. The Wireless card I choose is not fresh new but it's cheap (14USD) and reported working on Linux: Atheros AR5BXB112 AR9380 450Mbps. If I get stability and around real 300Mbps is good enough.

 

M2 drive. I'll put one 128G INSSD128GM.26M2242 TRANSCEND MTS400 128GB SSD SATA3 M.2 2242 TS128GMTS400 which should be big enough for most cases.

 

Mpci2sata for one ordinary SATA. 

 

SFP module. Haven't found the proper one yet.

 

With Armbian tools it's possible to build (I collect and fix all patches) and should be possible to boot kernel 4.3.3 and 4.4 - next.

 

Network performance (over TP link switch). 

Windows desktop -> Clearfog

[  3]  0.0-10.0 sec  1.09 GBytes   938 Mbits/sec
[  3]  0.0-10.0 sec  1.07 GBytes   922 Mbits/sec
[  3]  0.0-10.0 sec  1.09 GBytes   938 Mbits/sec

Clearfog -> Windows desktop

[  4]  0.0-10.0 sec  1.04 GBytes   889 Mbits/sec
[  5]  0.0-10.0 sec  1.02 GBytes   878 Mbits/sec
[  4]  0.0-10.0 sec  1.03 GBytes   882 Mbits/sec

To be continued ...

 

Edit #1: Board boots with 4.3.3 / Eth: no, USB: yes, mPci: yes

Edit #2: Jessie image download

Edit #3: Legacy kernel upgraded to 3.10.94

Edit #4: Network performance

Edit #5: If eth0 connected to gigabit, + cca. 2°C more heat on CPU in idle

Edit #6: mSATA is working with patched U-boot

Edit #7: mSATA and mPCI can be enabled in any combination. Tested one wireless card ...

Edit #8: boot time when installed to low performance 32Gb mSATA, power -> Login prompt <= 10s, (-3s waiting at u-boot) = cca. 7sec

Edit #9: I2C working out of the box. Tested with display, communication is slow. Perhaps only the settings

Edit #10: USB somehow buggy on current legacy kernel. I was trying to use Temper to collect temperature ... it worked once, next reading hang the device and I need to unplug / plug. Haven't debug. USB flash memory working normal.

Edit #11: Added cpu freq scalling 800/1600Mhz -> less heat in idle

Edit #12: Kernel 4.3.3 & 4.4 / Eth: yes, USB: yes, mPci: yes, mSata: yes ... and doing preliminary testing with Atheros AR9380 N wireless card // Kernel not ready yet

Edit:#13: Added mSATA to SATA card, patch a driver to unlock cheap - (consumer grade!, 13 USD shipped) AR9380 AR5BXB112 wireless card to operate at 5Ghz. Running STABLE

Edit:#14: Booting from m2 120G SATA drive

Edit:#15: Measuring temperature on the edge of heatsink = 39-42 while ambient is around 20.

Edit:#16: Added external 2.5 inch mechanical SATA drive powered from Clearfog +5V from header

5
Link to comment
Share on other sites
Tido

Tido

Posted
Posted

The picture is a bit dark, but a nice idea with the R1 in the background to compare the size of the PCB. It is quite large.

 

What about the LEDs on the Ethernetports, I haven't attached any device on my R1 lately, but they didn't work in the past.

Do these LEDs work on ClearFog?

1
Link to comment
Share on other sites
tkaiser

tkaiser

Posted
Posted

To be continued ...

 

Curious whether you had the time to do some benchmarking already? SATA and Ethernet performance? :)

 

Also the output of the following for both kernel variants would be great (to prepare RPi-Monitor support for Armada 38x):

find /sys -iname "*temp*"
Link to comment
Share on other sites
Igor

Igor

Posted
  • Author
Posted

I do have mSATA drive around but according to Solid-run staff:

 

"If you want to replace the mini pcie with msata then you can do it and it will work with some tweaks in u-boot.
The pcie port can be configured to becomes sata without adding any extra hardware."

 

I don't have that info yet, neither M2 drive, ...

I'll make better photos tomorrow on day light.

Link to comment
Share on other sites
tkaiser

tkaiser

Posted
Posted

And all the information (inluding schematics and hardware details like the heatsink design) fully available: http://wiki.solid-run.com/doku.php?id=products:a38x:documents

 

By looking at the pictures and in the schematics I just learned that there's an onboard UART-to-USB adapter utilising the Micro USB jack next to the SD card slot. That's convenient (and maybe the best use for micro USB anyway ;) )

 

BTW: Do you have any of these 'Clickboards' lying around? I fear most if not all will not fit inside Solid-Run's enclosure?

Link to comment
Share on other sites
Igor

Igor

Posted
  • Author
Posted

BTW: Do you have any of these 'Clickboards' lying around? I fear most if not all will not fit inside Solid-Run's enclosure?

 

Nope. I'll skip this part ... Not fan of such solutions. 

 

Yes, UART is "plug and play" ;)

Link to comment
Share on other sites
tkaiser

tkaiser

Posted
Posted

Nope. I'll skip this part ... Not fan of such solutions. 

 

Well, at least it's good to know that some sort of a standard is used to expose a few useful interfaces and 3.3V/5V on this header:

 

mikrobus_pinout_b.jpg

 

BTW: To get an idea about integer performance of Armada 388, could you provide the results of the following please if you find some spare time?

sysbench --test=cpu --cpu-max-prime=20000 run --num-threads=2
Link to comment
Share on other sites
Igor

Igor

Posted
  • Author
Posted 
root@armada:~# sysbench --test=cpu --cpu-max-prime=20000 run --num-threads=2
sysbench 0.4.12:  multi-threaded system evaluation benchmark

Running the test with following options:
Number of threads: 2

Doing CPU performance benchmark

Threads started!
Done.

Maximum prime number checked in CPU test: 20000


Test execution summary:
    total time:                          187.9687s
    total number of events:              10000
    total time taken by event execution: 375.9061
    per-request statistics:
         min:                                 37.49ms
         avg:                                 37.59ms
         max:                                 37.90ms
         approx.  95 percentile:              37.64ms

Threads fairness:
    events (avg/stddev):           5000.0000/2.00
    execution time (avg/stddev):   187.9531/0.01

I don't have troubles with header but modules ... I rather DIY ;). Header is O.K., most important stuff is here.

Link to comment
Share on other sites
tkaiser

tkaiser

Posted
Posted 
root@armada:~# sysbench --test=cpu --cpu-max-prime=20000 run --num-threads=2
...
    execution time (avg/stddev):   187.9531/0.01

 

Thx, that's comparable with RPi 2 (4 cores @ 900 MHz) and means that single-threaded performance is very good (to be more precise: way better than any of the quad-core SoCs we tested here the last few months). Hopefully we get CESA accelerated encryption and due to better hardware not that much CPU load when utilising I/O and network.

Link to comment
Share on other sites
Guest Rabeeh Khoury

Guest Rabeeh Khoury

Posted
Posted

Hi,

 

I had previously done some I/O benchmark; running dd with direct I/O (i.e. data doesn't go through processor caches) i was able to reach ~1.5GB/sec on 3 SSDs (2 on mini pcie and one m.2) which is maxing out the 3 SATA links (each SATA link is ~540MB/sec).

 

Igor - i can provide patch for modifying the processor I/O from pcie to msata (without changing anything on the hardware side).

 

Link to comment
Share on other sites
Guest Rabeeh Khoury

Guest Rabeeh Khoury

Posted
Posted

Just looked it up; you need to modify the #if 1 to #if 0 in the following two places -

 

https://github.com/SolidRun/u-boot-armada38x/blob/u-boot-2013.01-15t1-clearfog/tools/marvell/bin_hdr/src_phy/a38x/mvHighSpeedTopologySpec-38x.c#L92

 

and

 

https://github.com/SolidRun/u-boot-armada38x/blob/u-boot-2013.01-15t1-clearfog/tools/marvell/bin_hdr/src_phy/a38x/mvHighSpeedTopologySpec-38x.c#L98

 

This will modify the SERDES of the PEX (pcie) to become a SATA port.... A magical feature from Marvell :)

 

Link to comment
Share on other sites
Skygod

Skygod

Posted
Posted (edited)

"There is no CPU governor (yet?), so I assume it's running full speed (1.6Ghz) all the time."


Interesting to see on their website.....

 

There will be overclocking options too in the future to 1.8ghz/2.0ghz with newer DDR devices - but this is not public for now.

 

 

 

From a quick read of their website, it would appear that this is what the Banana R1 should have been.

If I could find a mini pcie ADSL2+ modem, this could be an interesting all-in-one gateway device project

 

 

 

EDIT : Actually, thinking about it, exposing the motherboard to the telecomms network might not be such a good idea, as I've already had a modem/router fried during a heavy storm.

Edited by Skygod
Link to comment
Share on other sites
tkaiser

tkaiser

Posted
Posted

I had previously done some I/O benchmark; running dd with direct I/O (i.e. data doesn't go through processor caches) i was able to reach ~1.5GB/sec on 3 SSDs (2 on mini pcie and one m.2) which is maxing out the 3 SATA links (each SATA link is ~540MB/sec).

 

This sounds great. Really looking forward to these Armada devices. Did you already look into making this SERDES stuff configurable through DTS?

 

@Skygod: The Clearfog Pro is a totally different beast compared to the Lamobo R1. If you want something that limited as the R1 you could take Solid-Run's MicroSom A380 and combine it with a cheap switch IC and crappy slow Wi-Fi: http://wiki.solid-run.com/doku.php?id=products:a38x:microsom

 

The single-core A380 is as fast as the dual-core A20 (but since single-threaded speed is nearly twice as fast it will outperform the A20 in most areas) but since the Marvell Armada SoC features 2 GbE MAC implementations you could really build a high speed router with this setup (unlike the R1 where you have to use a separate USB-Ethernet dongle to achieve this)

Link to comment
Share on other sites
tkaiser

tkaiser

Posted
Posted

fiber? In the boonies of N.E.Thailand?

 

A media converter like TP-Link MC220L could do the job. Since the Clearfog Pro can be fed with 9V- 32V one 9V PMU providing enough current is sufficient to power both board and MC220L.

Link to comment
Share on other sites
Skygod

Skygod

Posted
Posted

A media converter like TP-Link MC220L could do the job. Since the Clearfog Pro can be fed with 9V- 32V one 9V PMU providing enough current is sufficient to power both board and MC220L.

 

You've lost me. How can I connect this to my incoming ADSL2+ presentation?

Link to comment
Share on other sites
tkaiser

tkaiser

Posted
Posted

You've lost me. How can I connect this to my incoming ADSL2+ presentation?

 

By using a rather 'dumb' DSL modem that 'bridges' the protocol your ISP uses (IPoA, PPPoE, PPPoA, whatever) to 100BaseTX. The media converter and a fiber connection provide at least safe galvanic isolation and the next storm fries only modem and media converter. But I would believe a simple surge protector would be cheaper ;)

Link to comment
Share on other sites
Igor

Igor

Posted
  • Author
Posted

Some more measurement:

 

M2 SSD TRANSCEND MTS400 128GB
 
Write:
1073741824 bytes (1.1 GB) copied, 6.65401 s, 161 MB/s
Read:
1073741824 bytes (1.1 GB) copied, 2.07552 s, 517 MB/s
 
Mechanical 2.5" drive Western Digital 500Gb / 5400rpm
 
Write:
1073741824 bytes (1.1 GB) copied, 9.56753 s, 112 MB/s
Read:
1073741824 bytes (1.1 GB) copied, 10.4673 s, 103 MB/s
 
SD card Sandisk Extreme PRO 8Gb
 
Write:
1073741824 bytes (1.1 GB) copied, 68.9829 s, 15.6 MB/s
Read
1073741824 bytes (1.1 GB) copied, 68.4105 s, 15.7 MB/s
1
Link to comment
Share on other sites
tkaiser

tkaiser

Posted
Posted

M2 SSD TRANSCEND MTS400 128GB

 
Write:
1073741824 bytes (1.1 GB) copied, 6.65401 s, 161 MB/s

 

The write speed must be limited by the SSD (small capacity --> less parallelism --> slow write speeds). I hope these little beasts arrive soon:

 

Bildschirmfoto%202016-01-28%20um%2011.49

Will then try to do some tests with some SSDs in parallel. BTW: 15 minutes ago a packet from Solid-Run arrived:

 

IMG_5901.JPG

 

Wow, up and running within 10 minutes including download of your Armbian image: http://pastebin.com/Y3vJmX3H

 

Great first experience. Everything already provided to start through (onboard UART-to-USB adapter, PSU adapter, TF card).

1
Link to comment
Share on other sites
Igor

Igor

Posted
  • Author
Posted

Great!  :D

 

BTW:

Yes, this M2 disk can't write faster in general. All my other high performance SSD drives are in use ;)

 

Remember to recompile u-boot to make use of mSATA instead of mPci when your little beasts arrive. Patch is prepared but disabled.

 

Edit: If you run into MAC address / Eth remapping problem, set MAC in u-boot - simply use this boot.cmd,

1
Link to comment
Share on other sites
tkaiser

tkaiser

Posted
Posted

Yes, this M2 disk

 

LOL! I've overseen the M.2 socket on the back completely!

 

Ok, also ordered another mechanical M.2/NGFF-SATA adapter. So there is one M.2 socket and 2 x mSATA/mini-PCIe on the other side of the PCB. Plenty of disk connections and I would believe the Armada SoC can cope well with port multipliers (to be tested).

 

Did you had a look whether the issues with 4.4 are resolved in the meantime?

Link to comment
Share on other sites
Igor

Igor

Posted
  • Author
Posted

No, I haven't got nowhere with 4.4 ... Solidrun official devs are also giving up and moving to 4.1.x, Don't know the exact reason - need to do more research and talk do people who are working on this.

 

There should be 4x SATA but I guess the 4th is nc ;)

 

PMP

http://www.dx.com/p/1-to-5-sata-3-0-hard-drive-adapter-expansion-card-blue-black-370801#.VqsR-egrLAQ

 

This one with some big fat heat sink should work?

Link to comment
Share on other sites
tkaiser

tkaiser

Posted
Posted

 

I will give it a try even if it almost hurts to combine such a quality board with cheap JMB321 crap ;)

 

But it will take some weeks for the mechanical adapters to arrive to be able to use 'plain SATA' together with the Clearfog.

 

Thx for the update regarding kernel. Then let's see how far we get with 4.1.x (when I start testing I want to rely on btrfs to detect possible corruptions and there the kernel version matters)

1
Link to comment
Share on other sites
Guest
This topic is now closed to further replies.
 Share
Go to topic listing
×
×
  • Create New...

Important Information

Terms of Use - Privacy Policy - Guidelines

  I accept