Ubuntu 15.04 (Loboris) - Unable to install kodi

whitewind · Published in 2015-11-18 18:16:56

It seems that I solved the problem!
I limited CPU freq to 1200 and DDR freq to 600 and experienced no crashes so far

bronco · Published in 2015-11-18 21:57:35

whitewind replied at 2015-11-18 18:16
I limited CPU freq to 1200 and DDR freq to 600 and experienced no crashes so far

As expected. It's still the simple rule "Cheap, fast, reliable -- choose two".

Would be interesting to compare your still 'overclocked' settings with sane settings. Would you mind running the 3 short benchmarks listed here with your current settings and with sane settings (1200 or better 1008 MHz cpufreq and DRAM @ 480 MHz) to compare results?

whitewind · Published in 2015-11-19 10:32:31

bronco replied at 2015-11-18 21:57
As expected. It's still the simple rule "Cheap, fast, reliable -- choose two".

Would be intere ...

cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq
1200000
cat /sys/devices/platform/sunxi-ddrfreq/devfreq/sunxi-ddrfreq/cur_freq
600000

Copy code

sysbench --test=cpu --cpu-max-prime=20000 run --num-threads=4 && echo -e "\n\n" && 7za b && echo -e "\n\n" && mbw -t0 200 ; mbw -t1 200 ; mbw -t2 200
sysbench 0.4.12: multi-threaded system evaluation benchmark
Test execution summary:
total time: 154.4941s
total number of events: 10000
total time taken by event execution: 617.8966
per-request statistics:
min: 61.10ms
avg: 61.79ms
max: 138.14ms
approx. 95 percentile: 66.28ms
Threads fairness:
events (avg/stddev): 2500.0000/40.42
execution time (avg/stddev): 154.4741/0.02
7-Zip (A) 9.20 Copyright (c) 1999-2010 Igor Pavlov 2010-11-18
p7zip Version 9.20 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,4 CPUs)
RAM size: 1002 MB, # CPU hardware threads: 4
RAM usage: 850 MB, # Benchmark threads: 4
Dict Compressing | Decompressing
Speed Usage R/U Rating | Speed Usage R/U Rating
KB/s % MIPS MIPS | KB/s % MIPS MIPS
22: 1310 295 432 1275 | 27377 303 814 2470
23: 1236 280 449 1259 | 34295 385 815 3138
24: 1209 315 413 1300 | 33697 388 806 3126
25: 1159 340 389 1324 | 33095 387 804 3112
----------------------------------------------------------------
Avr: 307 421 1290 366 810 2961
Tot: 336 615 2125
Long uses 4 bytes. Allocating 2*52428800 elements = 419430400 bytes of memory.
Getting down to business... Doing 10 runs per test.
0 Method: MEMCPY Elapsed: 1.44191 MiB: 200.00000 Copy: 138.705 MiB/s
1 Method: MEMCPY Elapsed: 0.48774 MiB: 200.00000 Copy: 410.058 MiB/s
2 Method: MEMCPY Elapsed: 0.48344 MiB: 200.00000 Copy: 413.704 MiB/s
3 Method: MEMCPY Elapsed: 0.48413 MiB: 200.00000 Copy: 413.110 MiB/s
4 Method: MEMCPY Elapsed: 0.48390 MiB: 200.00000 Copy: 413.306 MiB/s
5 Method: MEMCPY Elapsed: 0.48377 MiB: 200.00000 Copy: 413.423 MiB/s
6 Method: MEMCPY Elapsed: 0.48406 MiB: 200.00000 Copy: 413.174 MiB/s
7 Method: MEMCPY Elapsed: 0.48439 MiB: 200.00000 Copy: 412.895 MiB/s
8 Method: MEMCPY Elapsed: 0.48377 MiB: 200.00000 Copy: 413.423 MiB/s
9 Method: MEMCPY Elapsed: 0.48416 MiB: 200.00000 Copy: 413.088 MiB/s
AVG Method: MEMCPY Elapsed: 0.58013 MiB: 200.00000 Copy: 344.753 MiB/s
Long uses 4 bytes. Allocating 2*52428800 elements = 419430400 bytes of memory.
Getting down to business... Doing 10 runs per test.
0 Method: DUMB Elapsed: 1.56488 MiB: 200.00000 Copy: 127.805 MiB/s
1 Method: DUMB Elapsed: 0.28296 MiB: 200.00000 Copy: 706.809 MiB/s
2 Method: DUMB Elapsed: 0.26870 MiB: 200.00000 Copy: 744.338 MiB/s
3 Method: DUMB Elapsed: 0.26941 MiB: 200.00000 Copy: 742.352 MiB/s
4 Method: DUMB Elapsed: 0.26937 MiB: 200.00000 Copy: 742.476 MiB/s
5 Method: DUMB Elapsed: 0.26952 MiB: 200.00000 Copy: 742.049 MiB/s
6 Method: DUMB Elapsed: 0.26910 MiB: 200.00000 Copy: 743.218 MiB/s
7 Method: DUMB Elapsed: 0.26919 MiB: 200.00000 Copy: 742.970 MiB/s
8 Method: DUMB Elapsed: 0.26969 MiB: 200.00000 Copy: 741.603 MiB/s
9 Method: DUMB Elapsed: 0.26973 MiB: 200.00000 Copy: 741.479 MiB/s
AVG Method: DUMB Elapsed: 0.40026 MiB: 200.00000 Copy: 499.681 MiB/s
Long uses 4 bytes. Allocating 2*52428800 elements = 419430400 bytes of memory.
Using 262144 bytes as blocks for memcpy block copy test.
Getting down to business... Doing 10 runs per test.
0 Method: MCBLOCK Elapsed: 0.57257 MiB: 200.00000 Copy: 349.299 MiB/s
1 Method: MCBLOCK Elapsed: 0.29024 MiB: 200.00000 Copy: 689.073 MiB/s
2 Method: MCBLOCK Elapsed: 0.29657 MiB: 200.00000 Copy: 674.388 MiB/s
3 Method: MCBLOCK Elapsed: 0.29116 MiB: 200.00000 Copy: 686.912 MiB/s
4 Method: MCBLOCK Elapsed: 0.28948 MiB: 200.00000 Copy: 690.882 MiB/s
5 Method: MCBLOCK Elapsed: 0.29001 MiB: 200.00000 Copy: 689.631 MiB/s
6 Method: MCBLOCK Elapsed: 0.28972 MiB: 200.00000 Copy: 690.329 MiB/s
7 Method: MCBLOCK Elapsed: 0.28939 MiB: 200.00000 Copy: 691.102 MiB/s
8 Method: MCBLOCK Elapsed: 0.28941 MiB: 200.00000 Copy: 691.056 MiB/s
9 Method: MCBLOCK Elapsed: 0.28953 MiB: 200.00000 Copy: 690.782 MiB/s
AVG Method: MCBLOCK Elapsed: 0.31881 MiB: 200.00000 Copy: 627.335 MiB/s

Copy code

cat /sys/devices/system/cpu/cpu0/cpufreq/scaling_max_freq
1008000
cat /sys/devices/platform/sunxi-ddrfreq/devfreq/sunxi-ddrfreq/cur_freq
480000

Copy code

Test execution summary:
total time: 184.6397s
total number of events: 10000
total time taken by event execution: 738.4045
per-request statistics:
min: 72.76ms
avg: 73.84ms
max: 184.96ms
approx. 95 percentile: 78.61ms
Threads fairness:
events (avg/stddev): 2500.0000/31.95
execution time (avg/stddev): 184.6011/0.02
7-Zip (A) 9.20 Copyright (c) 1999-2010 Igor Pavlov 2010-11-18
p7zip Version 9.20 (locale=en_US.UTF-8,Utf16=on,HugeFiles=on,4 CPUs)
RAM size: 1002 MB, # CPU hardware threads: 4
RAM usage: 850 MB, # Benchmark threads: 4
Dict Compressing | Decompressing
Speed Usage R/U Rating | Speed Usage R/U Rating
KB/s % MIPS MIPS | KB/s % MIPS MIPS
22: 883 284 302 859 | 28271 383 666 2550
23: 1024 304 342 1043 | 28215 384 672 2582
24: 967 328 316 1040 | 28018 386 674 2599
25: 961 346 316 1097 | 27530 386 670 2589
----------------------------------------------------------------
Avr: 316 319 1010 385 670 2580
Tot: 350 495 1795
Long uses 4 bytes. Allocating 2*52428800 elements = 419430400 bytes of memory.
Getting down to business... Doing 10 runs per test.
0 Method: MEMCPY Elapsed: 1.95616 MiB: 200.00000 Copy: 102.241 MiB/s
1 Method: MEMCPY Elapsed: 0.66871 MiB: 200.00000 Copy: 299.082 MiB/s
2 Method: MEMCPY Elapsed: 0.61808 MiB: 200.00000 Copy: 323.584 MiB/s
3 Method: MEMCPY Elapsed: 0.61786 MiB: 200.00000 Copy: 323.698 MiB/s
4 Method: MEMCPY Elapsed: 0.61683 MiB: 200.00000 Copy: 324.237 MiB/s
5 Method: MEMCPY Elapsed: 0.61812 MiB: 200.00000 Copy: 323.563 MiB/s
6 Method: MEMCPY Elapsed: 0.61738 MiB: 200.00000 Copy: 323.951 MiB/s
7 Method: MEMCPY Elapsed: 0.61769 MiB: 200.00000 Copy: 323.785 MiB/s
8 Method: MEMCPY Elapsed: 0.61762 MiB: 200.00000 Copy: 323.825 MiB/s
9 Method: MEMCPY Elapsed: 0.61686 MiB: 200.00000 Copy: 324.222 MiB/s
AVG Method: MEMCPY Elapsed: 0.75653 MiB: 200.00000 Copy: 264.364 MiB/s
Long uses 4 bytes. Allocating 2*52428800 elements = 419430400 bytes of memory.
Getting down to business... Doing 10 runs per test.
0 Method: DUMB Elapsed: 1.79266 MiB: 200.00000 Copy: 111.566 MiB/s
1 Method: DUMB Elapsed: 0.33156 MiB: 200.00000 Copy: 603.215 MiB/s
2 Method: DUMB Elapsed: 0.34014 MiB: 200.00000 Copy: 588.002 MiB/s
3 Method: DUMB Elapsed: 0.34048 MiB: 200.00000 Copy: 587.413 MiB/s
4 Method: DUMB Elapsed: 0.33511 MiB: 200.00000 Copy: 596.810 MiB/s
5 Method: DUMB Elapsed: 0.34012 MiB: 200.00000 Copy: 588.031 MiB/s
6 Method: DUMB Elapsed: 0.34004 MiB: 200.00000 Copy: 588.159 MiB/s
7 Method: DUMB Elapsed: 0.33948 MiB: 200.00000 Copy: 589.136 MiB/s
8 Method: DUMB Elapsed: 0.33083 MiB: 200.00000 Copy: 604.535 MiB/s
9 Method: DUMB Elapsed: 0.33151 MiB: 200.00000 Copy: 603.296 MiB/s
AVG Method: DUMB Elapsed: 0.48219 MiB: 200.00000 Copy: 414.771 MiB/s
Long uses 4 bytes. Allocating 2*52428800 elements = 419430400 bytes of memory.
Using 262144 bytes as blocks for memcpy block copy test.
Getting down to business... Doing 10 runs per test.
0 Method: MCBLOCK Elapsed: 0.69250 MiB: 200.00000 Copy: 288.808 MiB/s
1 Method: MCBLOCK Elapsed: 0.42269 MiB: 200.00000 Copy: 473.160 MiB/s
2 Method: MCBLOCK Elapsed: 0.40659 MiB: 200.00000 Copy: 491.901 MiB/s
3 Method: MCBLOCK Elapsed: 0.40983 MiB: 200.00000 Copy: 488.008 MiB/s
4 Method: MCBLOCK Elapsed: 0.41297 MiB: 200.00000 Copy: 484.292 MiB/s
5 Method: MCBLOCK Elapsed: 0.40258 MiB: 200.00000 Copy: 496.799 MiB/s
6 Method: MCBLOCK Elapsed: 0.39701 MiB: 200.00000 Copy: 503.762 MiB/s
7 Method: MCBLOCK Elapsed: 0.41403 MiB: 200.00000 Copy: 483.054 MiB/s
8 Method: MCBLOCK Elapsed: 0.39970 MiB: 200.00000 Copy: 500.379 MiB/s
9 Method: MCBLOCK Elapsed: 0.39838 MiB: 200.00000 Copy: 502.029 MiB/s
AVG Method: MCBLOCK Elapsed: 0.43563 MiB: 200.00000 Copy: 459.107 MiB/s

Copy code

bronco · Published in 2015-11-19 22:48:48

whitewind replied at 2015-11-19 10:32

Would you mind giving cpufreq-ljt-stress-test a try as outlined here: http://linux-sunxi.org/Hardware_Reliability_Tests#CPU

whitewind · Published in 2015-11-20 17:53:06

bronco replied at 2015-11-19 22:48
Would you mind giving cpufreq-ljt-stress-test a try as outlined here: http://linux-sunxi.org/Hardw ...

Testing CPU 0
1536 MHz SKIPPED
1440 MHz SKIPPED
1344 MHz SKIPPED
1200 MHz ............................................................ OK
1104 MHz ............................................................ OK
1008 MHz ............................................................ OK
912 MHz ............................................................ OK
816 MHz ............................................................ OK
720 MHz ............................................................ OK
648 MHz ............................................................ OK
600 MHz ............................................................ OK
504 MHz ............................................................ OK
480 MHz ............................................................ OK
408 MHz SKIPPED
312 MHz SKIPPED
240 MHz SKIPPED
120 MHz SKIPPED
60 MHz SKIPPED

Copy code

All cores produce the same result. No failures as expected: I use heatsink with fan

bronco · Published in 2015-11-20 18:35:16

whitewind replied at 2015-11-20 17:53
All cores produce the same result. No failures as expected: I use heatsink with fan

Sorry, I don't understand... what do you mean with "all cores produce the same result"?

A realistic approach to check both stability and data integrity on a multi core system with n cores is to run n-1 cpuburn-a7 tasks while executing 1 cpufreq-ljt-stress-test and see what's happening then.

Same applies to lima-memtester.

The so called 'default' clockspeeds used with the H3 based Orange Pis are so insanely high that I would simply call them overclocked. But where's the point when you run your system even when idle at voltage/clockspeed combinations that would only be useful if you plan to do high performance computing. And then not trying out how the hardware behaves in such HPC situations when utilising all cores with 100%.

If HPC is NOT the goal then it would make much more sense to use conservative settings (adjusting clockspeeds in sysfs is not enough, it would also require more operating point's inside the fex file's dvfs table) since the need for a fan or even a heatsink might be gone. And if HPC is the goal then a reliability test with HPC conditions has to proof that the settings are sane for such situations (but I still doubt that an ultra-slow Cortex-A7 design is the right foundation for HPC

but since the problem also applies to other situations like gaming it's still necessary to test what's happening when running with full utilisation)

Anyway: In a few weeks my Orange Pi PC will arrive and I will then give it a try with something like this http://pastebin.com/Wa879G0Y and report back the other thread.

whitewind · Published in 2015-11-20 20:44:44

bronco replied at 2015-11-20 18:35
Sorry, I don't understand... what do you mean with "all cores produce the same result"?

A realist ...

cpuburn-a7 starts 4 threads and doesn't have an option.
Anyway, I when I run cpuburn-a7 and cpufreq-ljt-stress-test in parallel, I have no errors and max temp is 56°

bronco · Published in 2015-11-20 20:55:31

Edited by bronco at 2015-11-20 21:00

whitewind replied at 2015-11-20 20:44
when I run cpuburn-a7 and cpufreq-ljt-stress-test in parallel, I have no errors and max temp is 56°

Ok, when lima-memtester also exits without an error while cpuburn-a7 is running simultaneously your cooling approach seems to be safe regarding data integrity and stability. Problems with decreased lifespan may remain due to high voltage settings but maybe that's not of interest to Orange Pi H3 users...
Anyway, thx for the tests. I will wait until my Orange Pi PC arrives and then do some own testings and report back here. And good news: Siarhei in the meantime received his first H3 device and that might improve hardware specific issues also a lot (he's the one who wrote the tools you used)

loboris · Published in 2015-11-20 21:40:20

Edited by loboris at 2015-11-20 15:07

cpufreq-ljt-stress-test (script modified to show cpu temperature and skip tests for 504 to 912 MHz)
OparngePi-2, heatsink&fan(3x3cm,5V), Ubuntu 14.04, kernel 3.4.39-01-lobo, ambient temperature 24 C

@bronco  I dont think there will be any "decreased lifespan" when running at 1200/1536 MHZ, H3 SoC is capable of running at 80 C continuously.

***

CPU stress test, which is doing JPEG decoding by libjpeg-turbo
at different cpufreq operating points.

Testing CPU 0
1536 MHz ............................................................ T=66 OK
1440 MHz ............................................................ T=65 OK
1344 MHz ............................................................ T=63 OK
1200 MHz ............................................................ T=49 OK
1104 MHz ............................................................ T=47 OK
1008 MHz ............................................................ T=46 OK
  912 MHz SKIPPED
  816 MHz SKIPPED
  720 MHz SKIPPED
  648 MHz SKIPPED
  600 MHz SKIPPED
  504 MHz SKIPPED
  480 MHz ............................................................ T=37 OK
  408 MHz SKIPPED
  312 MHz SKIPPED
  240 MHz SKIPPED
  120 MHz SKIPPED
60 MHz SKIPPED

Testing CPU 1
1536 MHz ............................................................ T=65 OK
1440 MHz ............................................................ T=64 OK
1344 MHz ............................................................ T=63 OK
1200 MHz ............................................................ T=48 OK
1104 MHz ............................................................ T=47 OK
1008 MHz ............................................................ T=46 OK
  912 MHz SKIPPED
  816 MHz SKIPPED
  720 MHz SKIPPED
  648 MHz SKIPPED
  600 MHz SKIPPED
  504 MHz SKIPPED
  480 MHz ............................................................ T=37 OK
  408 MHz SKIPPED
  312 MHz SKIPPED
  240 MHz SKIPPED
  120 MHz SKIPPED
60 MHz SKIPPED

Testing CPU 2
1536 MHz ............................................................ T=65 OK
1440 MHz ............................................................ T=64 OK
1344 MHz ............................................................ T=63 OK
1200 MHz ............................................................ T=48 OK
1104 MHz ............................................................ T=47 OK
1008 MHz ............................................................ T=46 OK
  912 MHz SKIPPED
  816 MHz SKIPPED
  720 MHz SKIPPED
  648 MHz SKIPPED
  600 MHz SKIPPED
  504 MHz SKIPPED
  480 MHz ............................................................ T=37 OK
  408 MHz SKIPPED
  312 MHz SKIPPED
  240 MHz SKIPPED
  120 MHz SKIPPED
60 MHz SKIPPED

Testing CPU 3
1536 MHz ............................................................ T=65 OK
1440 MHz ............................................................ T=64 OK
1344 MHz ............................................................ T=63 OK
1200 MHz ............................................................ T=49 OK
1104 MHz ............................................................ T=47 OK
1008 MHz ............................................................ T=46 OK
  912 MHz SKIPPED
  816 MHz SKIPPED
  720 MHz SKIPPED
  648 MHz SKIPPED
  600 MHz SKIPPED
  504 MHz SKIPPED
  480 MHz ............................................................ T=37 OK
  408 MHz SKIPPED
  312 MHz SKIPPED
  240 MHz SKIPPED
  120 MHz SKIPPED
60 MHz SKIPPED

Overall result : PASSED

Without fan
(when 1536 test runs successfully, it means that it actually run at 1200 MHz,
because of termal throttling the frequeny is imediately after start lowered to 1200 MHz
[ 9674.230309] CPU Budget:update CPU 0 cpufreq max to 1200000 min to 480000
[ 9675.770205] CPU Budget: Temperature: 77 Limit state:0 item[1536000,4,-1,0 0]
)

CPU stress test, which is doing JPEG decoding by libjpeg-turbo
at different cpufreq operating points.

Testing CPU 0
1536 MHz ............................................................ T=74 OK
1440 MHz SKIPPED
1344 MHz SKIPPED
1200 MHz ............................................................ T=73 OK
1104 MHz ............................................................ T=73 OK
1008 MHz ............................................................ T=73 OK
  912 MHz SKIPPED
  816 MHz SKIPPED
  720 MHz SKIPPED
  648 MHz SKIPPED
  600 MHz SKIPPED
  504 MHz SKIPPED
  480 MHz ............................................................ T=62 OK
  408 MHz SKIPPED
  312 MHz SKIPPED
  240 MHz SKIPPED
  120 MHz SKIPPED
60 MHz SKIPPED

Testing CPU 1
1536 MHz ............................................................ T=76 OK
1440 MHz SKIPPED
1344 MHz SKIPPED
1200 MHz ............................................................ T=78 OK
1104 MHz ............................................................ T=77 OK
1008 MHz ............................................................ T=77 OK
  912 MHz SKIPPED
  816 MHz SKIPPED
  720 MHz SKIPPED
  648 MHz SKIPPED
  600 MHz SKIPPED
  504 MHz SKIPPED
  480 MHz ............................................................ T=65 OK
  408 MHz SKIPPED
  312 MHz SKIPPED
  240 MHz SKIPPED
  120 MHz SKIPPED
60 MHz SKIPPED

Testing CPU 2
1536 MHz . T=77 FAILED
1440 MHz SKIPPED
1344 MHz SKIPPED
1200 MHz ............................................................ T=78 OK
1104 MHz ............................................................ T=78 OK
1008 MHz ............................................................ T=77 OK
  912 MHz SKIPPED
  816 MHz SKIPPED
  720 MHz SKIPPED
  648 MHz SKIPPED
  600 MHz SKIPPED
  504 MHz SKIPPED
  480 MHz ............................................................ T=65 OK
  408 MHz SKIPPED
  312 MHz SKIPPED
  240 MHz SKIPPED
  120 MHz SKIPPED
60 MHz SKIPPED

Testing CPU 3
1536 MHz ............................................................ T=79 OK
1440 MHz SKIPPED
1344 MHz SKIPPED
1200 MHz ............................................................ T=80 OK
1104 MHz ............................................................ T=79 OK
1008 MHz ............................................................ T=78 OK
  912 MHz SKIPPED
  816 MHz SKIPPED
  720 MHz SKIPPED
  648 MHz SKIPPED
  600 MHz SKIPPED
  504 MHz SKIPPED
  480 MHz ............................................................ T=65 OK
  408 MHz SKIPPED
  312 MHz SKIPPED
  240 MHz SKIPPED
  120 MHz SKIPPED
60 MHz SKIPPED

Overall result : FAILED

Click on image to show full size

bronco · Published in 2015-11-20 22:23:42

loboris replied at 2015-11-20 21:40
cpufreq-ljt-stress-test (script modified to show cpu temperature and skip tests for 504 to 912 MHz)
...

Wow! Great comprehensive reply. Thank you a lot.

Very good idea to show the temperature as well. Might be an idea to add /sys/devices/system/cpu/cpu0/cpufreq/cpuinfo_cur_freq also to the output and publish the modified script somewhere?

Regarding temperatures: based on my (very limited electrical) knowledge it makes a huge difference if a SoC stays at sane temperatures on its own or whether you need advanced heat dissipation techniques (AKA heatsink + fan) to achieve these temperatures. You increase the core voltage a bit and temperatures rise almost exponentially. Based on discussions on the linux-sunxi list it seems to be common knowledge that slight increases in core voltage negatively affect the chip's achievable lifespan.

The fex files from you I looked into just define 2 dvfs table entries with rather high voltages. I would suspect that many of the heat problems disappear automagically if one uses more entries with lower voltages for lower clockspeeds (I will start with these entries when my board arrives in a few weeks, but can not recommend to test these by anyone else who's not experienced with modifying fex files). I'll report back with some insights then.
At the moment I'm happy that the heat problems seem to be solveable by adjusting dvfs stuff and clockspeeds. And that's all I need to evaluate the H3 for home automation and IoT stuff.

In my eyes the H3 only has two interesting features: H.265 capable (Android only) and cheap as hell (no PMU, integrated 100 MBit Ethernet PHY). Since I'm not interested in Android only the latter applies...

Ubuntu 15.04 (Loboris) - Unable to install kodi

This thread contains more resources

Viewed section