How on earth did you guys get it to get a higher cpu sysbench with newer software? I can understand optimizing things, but to benchmark higher with the same hardware is quite impressive.
I have been very happy with the software development on the librem 5. This is honestly what I feel I’m paying for when I get a librem 5. Even with the delays on the hardware, seeing things moving forward with the software is really nice!
My biggest compliments for the work you do.
I am increasingly looking forward to the day when PureOS 10 Byzantium is released.
This post inspired me to get the latest x86 image and play around in QEMU again, haven’t done that in a while.
Was the Evergreen’s brightness up higher than the Chesnut in this video? Or is it some other change that made it look so much brighter and crisper?
Oh man, called out.
The screens are about the same brightness; however, I made a mistake by leaving my circular polarized lens filter on. In some shots, the filter can help by cutting down on all the light reflecting back at the camera. But, it is just about the worst thing for recording two subjects that both put out polarized light. So, I quickly pulled the filter once I noticed I still had it on. 
I’m really pleased to see Byzantium Evergreen with the screen on and WiFi and modem off using about 0.7W. With a 17.1 Wh (4.5Ah * 3.8V) battery, that means about 24 hours of battery life. I would love to see if anyone is managing to get such long battery life in their testing. It’s nice to see such progress, and I would love to hear what specifically allowed such a big jump in energy efficiency in Byzantium.
Of course, most people will want to have either the WiFi or cellular modem on, so the question is how energy efficient can Purism make those components. Even if it takes Purism a while to implement suspend-to-RAM, and the BM818 being able to wakeup the system when getting a call, I will be happy without suspend-to-RAM, if the L5 can get to a 24 hr battery life with the BM818 on and the screen off most of the time, except when getting phone calls.
By the way, did anyone notice that Purism has now raised $7.16M in convertible notes? I’m hoping this means that Purism will be able hire a developer to focus on power management issues.
The room for improvement is still pretty large from a software optimization stand point. It is not that they are getting higher scores, per se, rather than they are getting closer to what the hardware should have gotten to begin with.
That said optimizing software is one of the hardest aspect of software development so this is a great accomplishment!
Unrelated question: is this beautiful voice yours @david.hamner? I always love listening to that voice in Purism videos
Take a look at the updated graph. I made an error using the wrong device tree on Evergreen hardware.
The value reported by power gauge is divided by r_sns; on Evergreen that should be 1667. But with the on other batches the value used was 5000, which is what I was using in my testing.
The fix for this was really simple: power*3 when using the old r_sns value. So the value: 0.6876W should be: 2.0628
I’m really sorry for letting this error slip by.
Yep, that is my voice in the video. But the videos are a team effort for sure! 
With a 17.1Wh battery and 2.06W of energy consumption, I calculate that we should get 8.3 hours of on screen time in airplane mode.
Not bad, considering that is about what my Android phone gets (Xiaomi Redmi Note 7, 4000mAh battery), but not as exciting as the old numbers (24.8 hours at 0.6876W).
At any rate, a 26.4% reduction (2.8W -> 2.06W) in energy consumption is still pretty impressive. Definitely worth publicizing. It would be great to have another article where you show some battery life graphs with Byzantium.