Why NBG6617 IPQ4018 bogoMIPS is so low?

root@OpenWrt:~# cat /proc/cpuinfo
processor	: 0
model name	: ARMv7 Processor rev 5 (v7l)
BogoMIPS	: 67.03
Features	: half thumb fastmult vfp edsp neon vfpv3 tls vfpv4 idiva idivt vfpd32 lpae evtstrm 
CPU implementer	: 0x41
CPU architecture: 7
CPU variant	: 0x0
CPU part	: 0xc07
CPU revision	: 5
[    0.000256] Calibrating delay loop (skipped), value calculated using timer frequency.. 96.00 BogoMIPS (lpj=480000)

Something is wrong.

No, BogoMIPS are just bogus and wrong. There have been several attempts to remove it from cpuinfo in the past (or to display only 0 or 1 as its value), but the need not to break (severely broken) userland software foiled those attempts. For new platforms (ARMv7/ ARMv8, etc.) there is no relation to the actual performance or clockspeed of the SOC, both for technical reasons (highres timers) just as well as the explicit intention to kill it with fire (there are no ABI guarantees for broken software requesting low-level CPU specific info from new chips).

Edit: https://lkml.org/lkml/2015/1/7/807 this thread (not just that single mail) explains the situation pretty well.


this value seem to depend on soc architecture and low-level pll setup, for example QCA953X that is very similar to QCA955X will display quite higher value at same clock rate than QCA955X, probably because QCA955X inverts CPU/DDR PLLs

This topic was automatically closed 10 days after the last reply. New replies are no longer allowed.