My new case

In https://tjohann.wordpress.com/2015/06/16/i-have-a-thermal-problem-with-my-new-board/ I have described the thermal problem with my old case. Now I have a new one. It’s in military green … really really cool.

If the machine builds two kernel with „make -j 10“, so that it’s really 100% load for all cores, the temperature of the regulators are around 50-54 degree. The specification for the board say that temperatures above 70 degree are critical, so it seems to be okay with the 50-54 degree.

After some playing with the location and direction of the fans, I find a good airflow. It’s different from most other air flow systems. The upper fan in the side and the last one at the top pull into the case to have an airflow around the fins of the regulator. The lower fan in the side an the first at the top push the air out of the case. Normally all fan in the side or top push or pull together and are not split-ed between push and pull , but I need an airflow around the regulator fins, because they’re shadowed by the radiator of the watercooler.

I have a thermal problem with my new board

My new combination is really powerful, but it shows a problem -> the temperature of the board (not the processor) goes up to more than 70 degree after 1 hour of building different stuff (like kernel or so). So I ordered better fans -> my choice was Noctua-NF12 PWM (http://www.noctua.at/main.php?show=productview&products_id=42&lng=de). But i still had the same problem. So I hanged-up one of the fans that the air flow direct spotlight the upper left corner (where the heat sinks of the power regulators are):

But after all its better, but not what I need. I ended up with another order for a new case -> Corsair Vengeance Series … of course in military green (http://www.corsair.com/de-de/vengeance-c70-mid-tower-gaming-case-military-green)

My old board died -> RIP

For about 1,5 years I updated my old i7-920 to an ADM-FX8320 which was a real performance jump, but last weekend it died. I don’t know why, but it wont do a thing, nor beep nor post, nothing. So I need a hardware. After some searching i decided to buy one from my computer part dealer. He has a bundle of board, processor and cooling (AMD FX-Series FX-9590, ASUS Crosshair V Formula-Z and Corsair Cooling Hydro Series H80i Extreme Performance Water Cooling ). Everything arrived at Tuesday this week and of course I assembled all at that day. Now after 5 days and moving all components around in the case i find a more or less final configuration:

The machine is amazingly fast … really really fast … it crossbuilds an arm926 „full“ kernel within minutes!

Now after this interrupt its time to start again with my armv7hl slackware port.

Release of crosscompile toolchain for arm926 targets on x86_64 hosts

I found some problems with my first version of the toolchain, so i fixed them and uploaded (https://sourceforge.net/projects/arm926sdk/files/sdk-v1/) the final version for x86_64 hosts. A version for armv7l hosts will follow as soon as i have slackware running on A20.

Some toolchain updates

During the last week I decided to reorganize my arm926_sdk. So I start with a „new“ naming scheme for the toolchain -> it starts again with version v1 but it also supports the host and sysroot extensions I build to use libconfig or libconfuse.

General it needs some testing, especially the download scripts but it should work 🙂 … I will update all during this week and I hopefully add support for armv7hl to it.

The adventure of building slackware starts today with an repo for the scipts

Actual I have debian sid on the cluster nodes, but today I made a failure -> I want to install mate as my desktop environment (on my master node -> a cubietruck). The problem with the mate packages for debian is that they need systemd. So after many statements that systemd is quite simple and that it makes no problem I decided to move from sysv to systemd, but that was a failure. The removing the systemd entry (see http://without-systemd.org/wiki/index.php/How_to_remove_systemd_from_a_Debian_jessie/sid_installation) and an aptitude update stop the cubietruck from reboot the system again.

The kernel starts, systemd starts and after some „ok“ outputs i get a lot of „failures“. It’s amazing how simple it is to destroy a working system 😦 … I use Linux and slackware for nearly 18 years and I never had such a problem!

So after that bad experience I decided, first build slackware than do some funny stuff with my cluster. The good thing is that I can use distcc to distribute the build of needed packages and so I hope to get a complete slackware-armv7l at the end of this month.

Alien Bob already start to build slackware for his chromebook (http://alien.slackbook.org/blog/armport/). I will use the work as my starting point and will host all scripts on github (https://github.com/tjohann/my_slack_arm_build_env). The difference will be that i build all with -mcpu=cortex-a7 -mfpu=neon-vfpv4 -mfloat-abi=hard …

So i’m quite curious what i will learn 🙂

Some background info about my a20 based cluster

The cluster has actual 10 nodes

• 8  Banana-Pi
• 2 Cubietruck

On the picture (left side) you can see the power supply which i took from an old pc, it’s stable and powerful. In the middle there’s the Banana-Pi stack with the 8 devices. The cable which go from every device to the small pcb on the button are for serial console and as the jailhouse (*) printk output.  On the right side you can see the two cubietruck. There`re the master devices. The one on the top has a harddisk which provide the space needed. All devices are connected via Gigabit-Switch (you can see a little bit of it).

(*) jailhouse is a hypervisor. I will write about it in an later post

Start building a A20 based cluster

After a long time of inactivity i start again to blog more about my private projects … one is my A20 based cluster … You can find some pictures of it on https://plus.google.com/u/0/107871910566886283917/posts/p/pub and the stuff i will code on https://github.com/tjohann/my_a20_cluster_sdk …