Performance iMX6ULL vs VF50?

Technical Support
,
1

Hello,
we are migrating from SOM module VF50 to i.MX6ULL and I’ve noticed that cat /proc/cpuinfo reports significantly worse BogoMIPS for module i.MX6ULL ( even driven at higher freq. ) than for VF50.
Our image is based on tdx-reference-minimal-image with some changes due to custom carrier board.

So I want to ask if performance of iMX6ULL is correct (i.e. it is really lower than on VF50 module), or shall we look for an issue with build kernel?

Values reported for VF50:
cat /proc/cpuinfo
processor : 0
model name : ARMv7 Processor rev 1 (v7l)
BogoMIPS : 262.96
Features : half thumb fastmult vfp edsp neon vfpv3 tls vfpv4 vfpd32
CPU implementer : 0x41
CPU architecture: 7
CPU variant : 0x0
CPU part : 0xc05
CPU revision : 1

Hardware : Freescale Vybrid VF5xx/VF6xx (Device Tree)
Revision : 0000
Serial : 02928057

Values reported for iMX6ULL:
cat /proc/cpuinfo
processor : 0
model name : ARMv7 Processor rev 5 (v7l)
BogoMIPS : 64.00
Features : half thumb fastmult vfp edsp neon vfpv3 tls vfpv4 idiva idivt vfpd32 lpae
CPU implementer : 0x41
CPU architecture: 7
CPU variant : 0x0
CPU part : 0xc07
CPU revision : 5

Hardware : Freescale i.MX6 Ultralite (Device Tree)
Revision : 0000
Serial : 07130558

Report from cpufreq-info for iMX6ULL:
cpufrequtils 008: cpufreq-info (C) Dominik Brodowski 2004-2009
Report errors and bugs to cpufreq@vger.kernel.org, please.
analyzing CPU 0:
driver: imx6q-cpufreq
CPUs which run at the same hardware frequency: 0
CPUs which need to have their frequency coordinated by software: 0
maximum transition latency: 86.0 us.
hardware limits: 198 MHz - 528 MHz
available frequency steps: 198 MHz, 396 MHz, 528 MHz
available cpufreq governors: conservative, userspace, powersave, ondemand, performance
current policy: frequency should be within 198 MHz and 528 MHz.
The governor “performance” may decide which speed to use
within this range.
current CPU frequency is 528 MHz (asserted by call to hardware).
cpufreq stats: 198 MHz:0.57%, 396 MHz:1.14%, 528 MHz:98.29% (458)

Thank you.

2

Hi @Janos !

Thanks for the question!

To be fair, I was not aware of this BogoMIPS measure.

Reading about it seems like there is no direct comparison between different processors.

From Wikipedia (BogoMips - Wikipedia):

It is not usable for performance comparisons among different CPUs.[4]

And then from Wikipedia’s reference (3. The frequently asked questions about BogoMips):

`MIPS is short for Millions of Instructions Per Second. It is a measure for the computation speed of a program. Like most such measures, it is more often abused than used properly (it is very difficult to justly compare MIPS for different kinds of computers).

I could not find documentation about BogoMIPS in Linux Kernel Documentation, though.

Do you have concerns about it?

Best regards,

3

bogomips = the number of million times per second a processor can do absolutely nothing

L2-cacheless VF50 is clearly much slower than iMX6ULL. If you asked about VF61, then it wouldn’t be so simple to answer. iMX6ULL has much smaller L2 cache than VF61, which may flip winner in some apps. As well VF61 in contrast to iMX6ULL includes precious Cortex-M4, which is not only a real-time enabler, but as well can be used to offload Linux from high interrupts rate, which is performance killer #1.

1 Like
4

Hi @Janos !

Was your question solved? If yes, please mark the related answer as the Solution :slight_smile:

Let us know.

Best regards,

5

Hi Henrique,

I don’t know if it is solved :), it was just a curiosity and answers are rather confusing for me.

I’ve installed image provided by Toradex to colibri imx6ull and reported BogoMIPS is the same as reported by image built by us so I suppose there is no issue with performance introduced by our customization.