4 ms·
I noticed something fishy. The /proc/cpuinfo BogoMIPS value reads off as 569.5 in the video. A PandaBoard 1Ghz Cortex-A9 reads around 2000 per processor[1]. I r
by angusgr 16y ago
I noticed something fishy. The /proc/cpuinfo BogoMIPS value reads off as 569.5 in the video. A PandaBoard 1Ghz Cortex-A9 reads around 2000 per processor[1]. I realise the Bogo means _bogus_ but AFAIK within a CPU model it should still scale linearly against clock speed.
Probably a sensible explanation for this, but at the same time this is the class of device where there really is no guarantee between spec claims and reality. For instance the advertised "600 & 800Mhz" WM8505s turned out to be 350Mhz CPUs in reality.
[1] http://www.gentoo.org/proj/en/base/embedded/handbook/?part=4&chap=9 http://www.gentoo.org/proj/en/base/embedded/handbook/?part=4...
- angusgr 16y agoProbably a sensible explanation It could be as simple as "runs at 250Mhz during early boot" but it also might not be. :)
- whiskers 16y agoIf you look at the CPU info it also shows it supports a range of frequencies from 200Mhz to 800Mhz and that the current running speed is ~280Mhz. So presumably the CPU scales between 200 and 800Mhz based on demand to improve battery life and the bogomips score was produced at a lower clock speed.
- angusgr 16y agoExcept cpufreq (kernel CPU frequency scaling) isn't running during the early boot stages when the BogoMIPS value is calculated. Like I said, it's possible the bootloader starts up at @250Mhz or so and leaves the kernel to later switch up to 800Mhz. I don't know why you'd design a board like that though (normally you want to boot nice and quick), but you never know. Also, even though cpufreq reports in Mhz those values will be extrapolated in the kernel from a base frequency & clock multiplier registers somewhere else, so it can easily be out by a constant factor. Once some benchmarks come out it should become more obvious if there's anything to this.