Since we have our DC measurements ready and just recently received our copy of the GUDE 1202-1 AC Power meter we thought it would be very interesting to reproduce the measurements of the first Software that received the Blauer Engel certificate so far: The free PDF viewer Okular from KDE.
In this post we will be mostly looking at the measurement part of the Blauer Engel, which however is really a minor part of what is relevant to get the label.
The label itself is really more of a sustainability label in terms of the lifecycle of the software. This means the software should be uninstallable, it should not lead to planned obsolescence, run on 5+ years old machines, export data to standard formats to prevent lock-in etc.
To get the Blauer Engel label NO minimum software performance metrics are enforced. This is planned for future iterations.
All the data from the measurements is available are in the Blue Angel Application repository from the KDE Eco Team.
The system used is a :
The most relevant piece for the usage data is probably the Annex 1 of the application form.
In here we find that:
These values are in a way very confusing as the unit %s is not something one has to deal with daily.
The idea in the original specification documents is to move all the merely descriptive values like % to the same dimension as energy (like power * time).
This however makes perfect sense if you have longer running and shorter running scenarios as it makes them better comparable.
What is very confusing though is the very small value of the idle average electrical power consumption. This value is found in other documents of the appliation (de-uz-215-eng-annex-3-okular-idle.pdf) as 5.07 W.
When looking a bit deeper in the application documents (de-uz-215-eng-annex-3-okular-scenario.pdf) we find that the 0.07 W originate from two measurements called Baseline and Idle. They represent the state in idle without the installed software and one in idle with the software installed. Both measuerements and then subtracted to get the theoretical overhead through the software itself. This explains the small value.
Als in this file we find the load average power 5.73 W which is what we will use as it represents our displayed measurement data better.
To make the calculation similar to our reported values we thus get:
[5.73 W (load) - 4.77 W (idle)] x 217 s / 3.6 = 57.86 mWh
Info: In the official application documents from the Blauer Engel specification this Baseline is called Grundlast.
In the Annex 1 the application mentions that Ubunutu 18.04 is required, but older versions should work too.
We opted for porting it to Ubuntu 22.04 as we believe this may encourage more people to have a go in trying out a certification for the Blauer Engel and get some code running as 18.04 does not get any Hardware and maintenance updates anymore and thus might not be something that people are eager to isntall anymore.
In order to have it running we made some adjustments to the Actiona script and even created a port to xdotool* which made it easier to work with for us.
Ubuntu 22.04 proved to be very relaxed here in terms of needed changes as only the pop-under behaviour of the window manager had to be accounted for which always opened Okular in the background.
For our screen also the clicking positions where greatly off and we ported it to our used resolution: 2560x1440
Since the Blauer Engel is quite restrictive on the test systems we had a bit problems getting our hands on one of these machines and opted for the only one still available to get on eBay, which was a:
Sadly we do not know what PSU was in the machine from the Blauer Engel official application for Okular, but all the PSUs from the Esprimo machines are either 80+ or 90+ in efficiency.
We ran the measurement with our setup by encapsulating xdotool into a docker container.
Okular sadly could not be boxed into a container as the UI was broken afterwards. So we resorted for having it run non containerized.
This means we have a net power draw of 2.15 W.
Since the Blauer Engel reports mostly on the energy let’s look at the Watt-hours:
The measurement took 213.877365 seconds. So we have to deduct a 213.877365 * 16.65 W / 3.6 = ~989.18 mwH from the measurement of the Standard Usage Scenario.
Since the Standard Usage Scenario used an energy of 1116.98 mWh this leaves us with a net total of ~ 127.80 mWh aka 0.127 Wh when removing the idle energy.
The difference in total energy of a factor of 2.2 seems a bit much for such simlar systems.
We are currently seeing ~20 % difference in similar machines for similar tasks so far, so this would be very interesting to have this machine tested with our lab equipment to drill deeper.
Nevertheless we are not sure if this derived value of 57.86 mWh holds, as the average power draw of a desktop computer of 5 W seems very unplausible. On this point in particular we have contacted the Umweltcampus Birkenfeld and will update the article with more details on that.
Apart from that we had some problems reproducing the measurements as some setup conditions posed not to be optimal for reproduing:
Althoug we have many critical points on the measurement part in this article we strongly believe in the approach of the Blue Angel as it is layed out so far.
The focus on the meta criteria and a Standard Usage Scenario is what really can help end users navigate through the complexity that modern software applications have if they want to make a sustainable digital decision.
Measuring all code paths and eventualities is neither scalable nor practical.
But by focusing on the major use cases and molding them into a Standard Usage Scenario is an approach to not only measure an application task, but also make it comparable.
We hope to see the Blue Angel move forward and streamline the application process as well as put a tighter grip on the measurement criteria. Also by introducing a central measurement service (available as SaaS) so that people interested in certifiying can use instead of investing into measurement hardware.
The restriction on the Esprimo machines makes perfect sense for us in order to allow people to reproduce the results, but not for scaling the certification process in general.