Battery Comparison Between Kernels in a Rom From An Engineer’s Perspective
Posted January 6, 2011 at 06:00 pm by egzthunder1
While this is a bit uncommon, we will be talking about a post in a particular thread rather than the whole thread itself. This post reminded us about the proper way of providing feedback on things as important as battery life. A sole post saying “battery life on rom xyz sucks” isn’t really going to help the developer (or anyone) determine the cause and provide a fix or a solution for the problem. XDA member frankpreyes decided to post a review on Myn’s Warm TwoPointTwo ROM thread for the EVO 4G. The comparison is carried out by keeping as many variables constant (no flashing other nighties, keeping radios, etc) and only focused on the kernels themselves as there were conflicting reports of many users claiming that battery life was better/worse with one or another. The report was done from an engineering perspective, so we are positive that many of you will agree with the approach.
Please leave feedback in this article since feedback in the thread will cause nothing more than make that monster thread even longer. And that is why the whole post will be featured in this article
I decided to see, so far, the difference between the HTC 15 and the older, HTC 17 kernel, as provided by Myn. Testing involved heavy usage, and light usage. Although the dataset is small, (From 1/3/2011 to 1/5/2011), there is definitely a quantitative difference between the two kernels.
For this comparison, I ran the same nightly, provided by Myn. I have a 0002 hardware. When I was running the HTC 15 kernel, I was using all the latest radios, wimax, NV (What the heck is an NV anyway). When I switched back to HTC 17, from Myns first post, I downgraded my radio. I decided not to downgrade my PRI (so far), to the previous revision, to see if I can quantitatively assess the impact, once I had baseline (HTC 15), and had modified as few variables as possible. Also, I started out with a fully charged seidio big battery, charged with the phone completely off, and cleared battery stats before using my evo. I also waited 1 hour after turning my evo on, before I used the % battery data, so as to allow the battery to stabilize.
Morning usage is light, typically reading emails, keeping up with this forum, looking stuff up, while afternoon usage is what I consider heavy, as I am using my Evo as a Wifi station (Sprint Hotspot, 3G), when I use my CR-48 netbook. U-Tube, This forum, Google reader, etc.
Using the battery monitor widget, I plotted %battery remaining per hour of use.
Here are the test results so far:
______________HTC 15_________HTC 17______% Difference
What I am reporting is the slope of a regression line drawn through the data for each day. The slope is the rate of change. In each case, the regression line matched very well with the data (R-squared = 0.96 to 0.99). The closer this number gets to 1, the more accurate the fit of the regression line to the data.
This provides some evidence that the HTC 15 kernel (Flashed on the first nightly) consumes more battery per hour than the HTC 17 kernel.
Todays data appears to be even better in favor of the HTC 17 kernel. This is proof, as some were saying that the #17 kernel gets better with use.
I will probably revert back to the previous PRI, to see what changes this makes to battery drain per hour. Once that is done, if there is any interest, I can post the graph I created that shows battery drain.
From a subjective standpoint, HTC 17 appears faster, and does not result in the morning sluggishness I was experiencing with HTC 15. . On HTC 15, I would get a force close in Launcher Pro Paid edition (Myns version) after first boot in the morning. After that, no other problems. With HTC 17, this no longer occurs. No programs were added or updated during this test.
You can find the full comparison in this post.
Want something published in the Portal? Contact any News Writer.
Want something on the XDA Portal? Send us a tip!