Anders Posted February 23, 2020 Share Posted February 23, 2020 I'm getting tired of Globalscale deleting my post from the Espressobin forums, instead of either fixing or acknowledging the problem. So I'll repost this here for future reference. This is the URL to the original thread: http://espressobin.net/forums/topic/mainline-linux-not-running-at-1ghz/ Hi All That the Espressobin aims to support mainline is awesome – a lot of the benefits are described here http://espressobin.net/mainline-linux/. But I’m unable to get my v7 board running at more than 800MHz using the mainline kernel (for some reason cpufreq-info insists on reporting 1GHz with actual benchmarking only showing 800MHz). For the sake of this post, I tried pulling the most recent linux-next kernel. I’m using the config file from armbian. export ARCH=arm64 export CROSS_COMPILE=aarch64-linux-gnu- git checkout next-20190614 curl https://raw.githubusercontent.com/armbian/build/master/config/kernel/linux-mvebu64-next.config > .config make oldconfig # Just picked the defaults for new options make -j7 Image dtbs modules cp arch/arm64/boot/Image /mnt/sdcard/boot/ cp arch/arm64/boot/dts/marvell/armada-3720-espressobin.dtb /mnt/sdcard/boot/dtb/marvell/ make modules_install INSTALL_MOD_PATH=/mnt/sdcard/ Here are the results. Notice the performance numbers and CPU Freq reported by 7-Zip: root@espressobin:~# uname -a Linux espressobin 5.2.0-rc4-next-20190614 #7 SMP PREEMPT Sun Jun 16 10:27:52 CEST 2019 aarch64 GNU/Linux root@espressobin:~# cpufreq-info cpufrequtils 008: cpufreq-info (C) Dominik Brodowski 2004-2009 Report errors and bugs to cpufreq@vger.kernel.org, please. analyzing CPU 0: driver: cpufreq-dt CPUs which run at the same hardware frequency: 0 1 CPUs which need to have their frequency coordinated by software: 0 1 maximum transition latency: 0.97 ms. hardware limits: 200 MHz - 1000 MHz available frequency steps: 200 MHz, 250 MHz, 500 MHz, 1000 MHz available cpufreq governors: conservative, userspace, powersave, ondemand, performance, schedutil current policy: frequency should be within 250 MHz and 1000 MHz. The governor "performance" may decide which speed to use within this range. current CPU frequency is 1000 MHz (asserted by call to hardware). cpufreq stats: 200 MHz:0.09%, 250 MHz:3.60%, 500 MHz:0.16%, 1000 MHz:96.15% (152) analyzing CPU 1: driver: cpufreq-dt CPUs which run at the same hardware frequency: 0 1 CPUs which need to have their frequency coordinated by software: 0 1 maximum transition latency: 0.97 ms. hardware limits: 200 MHz - 1000 MHz available frequency steps: 200 MHz, 250 MHz, 500 MHz, 1000 MHz available cpufreq governors: conservative, userspace, powersave, ondemand, performance, schedutil current policy: frequency should be within 250 MHz and 1000 MHz. The governor "performance" may decide which speed to use within this range. current CPU frequency is 1000 MHz (asserted by call to hardware). cpufreq stats: 200 MHz:0.09%, 250 MHz:3.60%, 500 MHz:0.16%, 1000 MHz:96.15% (152) root@espressobin:~# lscpu Architecture: aarch64 Byte Order: Little Endian CPU(s): 2 On-line CPU(s) list: 0,1 Thread(s) per core: 1 Core(s) per socket: 2 Socket(s): 1 NUMA node(s): 1 Model: 4 CPU max MHz: 1000.0000 CPU min MHz: 200.0000 BogoMIPS: 25.00 NUMA node0 CPU(s): 0,1 Flags: fp asimd evtstrm aes pmull sha1 sha2 crc32 cpuid root@espressobin:~# 7za b 7-Zip (a) [64] 16.02 : Copyright (c) 1999-2016 Igor Pavlov : 2016-05-21 p7zip Version 16.02 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,64 bits,2 CPUs LE) LE CPU Freq: 794 794 794 794 794 794 794 794 RAM size: 986 MB, # CPU hardware threads: 2 RAM usage: 441 MB, # Benchmark threads: 2 Compressing | Decompressing Dict Speed Usage R/U Rating | Speed Usage R/U Rating KiB/s % MIPS MIPS | KiB/s % MIPS MIPS 22: 709 148 467 690 | 17938 199 771 1532 23: 701 149 480 715 | 17737 199 770 1535 24: 696 149 503 749 | 17395 199 768 1527 25: 688 149 528 786 | 17149 198 769 1526 ---------------------------------- | ------------------------------ Avr: 149 494 735 | 199 770 1530 Tot: 174 632 1133 The following are the performance numbers I’m looking for, achieved using 4.4.8-armada-17.02-espressobin kernel downloaded from http://espressobin.net/tech-spec/. That one, and 4.4.52-armada-17.10.4-g719fc86-dirty are the only kernels I have been able to actually get 1GHz performance with: root@espressobin:~# uname -a Linux espressobin 4.4.52-armada-17.10.4-g719fc86-dirty #7 SMP PREEMPT Tue Jul 3 10:59:53 UTC 2018 aarch64 GNU/Linux root@espressobin:~# 7za b 7-Zip (a) [64] 16.02 : Copyright (c) 1999-2016 Igor Pavlov : 2016-05-21 p7zip Version 16.02 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,64 bits,2 CPUs LE) LE CPU Freq: 734 997 998 998 998 992 996 996 RAM size: 927 MB, # CPU hardware threads: 2 RAM usage: 441 MB, # Benchmark threads: 2 Compressing | Decompressing Dict Speed Usage R/U Rating | Speed Usage R/U Rating KiB/s % MIPS MIPS | KiB/s % MIPS MIPS 22: 870 150 563 847 | 22431 198 969 1915 23: 860 151 582 876 | 22154 198 968 1918 24: 854 151 607 918 | 21735 198 963 1908 25: 848 152 639 969 | 21341 198 960 1900 ---------------------------------- | ------------------------------ Avr: 151 598 903 | 198 965 1910 Tot: 174 781 1406 Following http://wiki.espressobin.net/tiki-index.php?page=Build+From+Source+-+Kernel to build 4.4.52-armada-17.10.4-g719fc86-dirty also works fine at 1GHz. I’ve tried a bunch of different U-boots, all with the same result, so I really think the bug is in Linux. (FYI this is not the only problem I have been having with the v7, but lets focus on one bug at a time. Not to mention that Globalscale claims 1.2 GHz for v7 which seems entirely out of reach as any 1200 u-boot bricks my board). 1 Link to comment Share on other sites More sharing options...
Werner Posted February 24, 2020 Share Posted February 24, 2020 I know this is more a self-defence agains censorship but I would like to know how Armbians kernels behave. Feel free to give all three branches (legacy, current and dev) a shot. Link to comment Share on other sites More sharing options...
lanefu Posted February 24, 2020 Share Posted February 24, 2020 @Anders you'll see similar frustrations on this thread. The v7s seem particularly flawed. Stability seems to vary board by board as well. Link to comment Share on other sites More sharing options...
sfx2000 Posted March 20, 2020 Share Posted March 20, 2020 On 2/23/2020 at 3:36 AM, Anders said: Following http://wiki.espressobin.net/tiki-index.php?page=Build+From+Source+-+Kernel to build 4.4.52-armada-17.10.4-g719fc86-dirty also works fine at 1GHz. I’ve tried a bunch of different U-boots, all with the same result, so I really think the bug is in Linux. (FYI this is not the only problem I have been having with the v7, but lets focus on one bug at a time. Not to mention that Globalscale claims 1.2 GHz for v7 which seems entirely out of reach as any 1200 u-boot bricks my board). MV3720 can definitely be stable at 1.0GHz with DDR3 running at 400Mhz with a fair amount of heat sinkage - but that's not espressobin Bare board - you're going to run into thermal issues straight up at 800MHz, based on experience (espressobin was our ref device for my board) Link to comment Share on other sites More sharing options...
Pali Posted October 10, 2020 Share Posted October 10, 2020 Hello! There is a bug in kernel's cpufreq driver which cause that Armada 3720 SOC is running only at 800 MHz, even it reports 1 GHz. Marek sent patches which fix this issue. See link https://lore.kernel.org/linux-arm-kernel/20201009125711.0176752a@kernel.org/ @Anders: Would you be able to test these patches and check if they are stable on your Espressobin V7 boards? Link to comment Share on other sites More sharing options...
Recommended Posts