Welcome back!
A while ago, I posted three parts of Fast-Booting Qt Devices blog post series where we showcased 1,5 second boot-time, optimized the Qt application and finally showed you how we optimized the entire Linux stack. Today, we will show you that hardware selection and hardware architecture in general can have a big impact on the actual startup time even when using the same CPU. To demonstrate this, we have two boards with NXP i.MX6 Quadcore CPU. One is a board geared towards software development, and the other is a system-on-module board aimed to be used in the production as well.
So, let's a have a small Battle of the Boards! :)
On the left side, we have the NXP SABRE i.MX 6 Quad Development Board:
- NXP i.MX 6 Quadcore processor, running at 1GHz
- 1GB DDR3 RAM
- 8GB eMMC
On the right, we have Toradex Apalis i.MX 6 Computer on Module:
- NXP i.MX 6 Quad core processor, running at 1GHz
- 1GB DDR3 RAM
- 4GB eMMC
Both boards are running exactly the same Qt Cluster demo, kernel configurations and u-boot.
Toradex Computer on Module is a clear winner with 19% (294 ms) faster startup time. Our earlier fast-boot example with the NXP SABRE resulted in a very good 1560 ms from power up to display of the first full screen Qt Quick frame. Now, with the Toradex board, we got an even faster 1266 ms.
Where does the difference come from?
- Powering up of the board is faster with Toradex module
- Kernel is able to access eMMC earlier resulting into a faster kernel startup time
So, when designing your embedded devices, remember that hardware selection does matter too. If you need to reach blazing fast startup time, it certainly helps to have fast memory and memory bus, well optimized bootloader and kernel, as well as of course a powerful chip that can quickly crunch through the libraries you need to load. The rest is then up to your software--even with the optimized hardware you can ruin your boot-up time with a sloppy software design. For those tips, check out the earlier posts in this series.
If you are interested in hearing more, I will be talking about fast-boot of Qt based devices at the Qt World Summit in San Francisco, October 18-20. We are looking forwards to seeing you there, and hearing your feedback!
Commenting for this post has ended.
How was the performance of CPU x GPU VIVANTE and using Qt Quick 2. Using EGLFS some tests I did with toradex i.MX6DL had an above normal consumption.
It would be great to see something like that for a more "common" and cheaper SoC, like Raspberry Pi or Beagle Bone.