Raspberry Pi Benchmarks

[dom]

The first example is included in the following benchmark report. This is for a benchmark running in a reliability testing mode, alongside my OpenGL benchmark and temperature measuring program. In this case, the RPi crashed due to overheating when overclocked.

James asked me to comment.

If I understand you were running a high overclock (CPU 1000 MHz, Core 500 MHz, SDRAM 600 MHz, 6 volts) when it crashed.
Obviously when busy with a high overclock the temperature rises.

When the temperature rises lots of physics change. Current may increase. Propagation delay may increase. Both these effects may affect overclock stability.

So, I don't think it's accurate to say the temperature reaching 70C results in a crash due to overheating.
You can say a borderline overclock setting may fail with increased temperature.

(when testing the temperature limiting overclock code, I deliberately tried to get the temperature as high as possible and got close to 80C.
for this specific Pi/power supply that was still stable, but with overclocking, your milage may vary).

[shuckle]

My pi crashes regularly when it reaches 80 °C. Need to find a way to cool it somehow.
I put it into the pink box from RS where it was bought. That probably hast too little cooling. First time i got that high temperature but it happens everytime.

[jamesh]

My pi crashes regularly when it reaches 80 °C. Need to find a way to cool it somehow.
I put it into the pink box from RS where it was bought. That probably hast too little cooling. First time i got that high temperature but it happens everytime.

Impressive you can get it to 80. I've never seen one anywhere near that. Is that under heavy load, high overclock?

[DavidS]

I would be very interested in the settings and code that is running to get 80 degrees, as well as the enviromental conditions.

I have not managed to get any of my RPi systems to reach greater than 62 degrees with an ambient temperature of up to 38 degrees C and humidity ranging from 20% to 85% saturation. This with force_turbo=1, and overclocked to maximum stable:

Code: Select all

force_turbo=1
over_voltage=6
arm_freq=1000
sdram_frep=500
core_freq=400

And even under heavy load for hours at a time. I have intentionaly tested at the highest ambient tempurature that one is likely to run there system in (38 degrees C) under heavy load (CPU at 100%, and heavily using GPU) in order to assure the stability of the system.

So I would be very interested to know your operating conditions, and what you were running in order to get your RPi to report anything close to 80 degrees.

[RoyLongbottom]

My pi crashes regularly when it reaches 80 °C. Need to find a way to cool it somehow.
I put it into the pink box from RS where it was bought. That probably hast too little cooling. First time i got that high temperature but it happens everytime.

The pink box that my RPi was supplied in would need large holes cutting in the top, besides some for cables and the SD card. Is your one a proper case?

[RoyLongbottom]

|
So I would be very interested to know your operating conditions, and what you were running in order to get your RPi to report anything close to 80 degrees.

Running my OpenGL benchmark, at 23°C room temperature, maximum CPU temperature was 67°C after 13 minutes and 69°C with full overclocking. Running my Livermore Loops program at the same time, also produced 69°C after 11 minutes, at normal frequency settings. On overclocking with the latter, 72.9°C was indicated (15 second sampling so could be higher) and one of the programs bombed out with an illegal instruction. See

http://www.roylongbottom.org.uk/Raspber ... m#anchor29

[DavidS]

Ok I guess there are some differences between RPi boards, and I have been very lucky.

[shuckle]

Yes, i cut holes for usb and network cables to my pink box. My pi is overclocked to 1000 Mhz arm and I run motion, which causes 70-80 % cpu load. The intention was to build an usb web camera box, but that failed due to the overheat.
I have other pi with higher load and higher overclock, so it has to be the pink box, which i now used first time. The other pis are in normal boxes and they stay under 60 °C.
The overload actually causes reboot.

[jamesh]

Yes, i cut holes for usb and network cables to my pink box. My pi is overclocked to 1000 Mhz arm and I run motion, which causes 70-80 % cpu load. The intention was to build an usb web camera box, but that failed due to the overheat.
I have other pi with higher load and higher overclock, so it has to be the pink box, which i now used first time. The other pis are in normal boxes and they stay under 60 °C.
The overload actually causes reboot.

Do you get the same problem with the other boards if you put them in the pink box?

[shuckle]

Just tested with another pi and it went only to 60 °C. I am surprised, but i can now continue my project. But the now problematic board has been in use and itbhas worked fine.

[RoyLongbottom]

sudo rpi-update

gets the latest released firmware. This is bleeding edge, so may have rough edges, but is usually fine. Lots of people use it. If you dont have rpi-update, get it here https://github.com/Hexxeh/rpi-update

The SoC itself is mostly GPU - that takes up the majority of the die area. The temp monitor is on the GPU and is pretty accurate IIRC. Not sure exactly how it works though. It is I believe what is used by the dynamic clocking to ensure the chip doesn't get too hot.

That did not work due to a missing file. I think that the SD card was corrupted, following an earlier recovery, as I managed to obliterate the Ext4 partition with more overclocking tests. After reinstalling Raspbian, rpi-update did its job but I still had a problem running my OpenGL tests on overclocking, where the system just froze. It only seems to occur when the OpenGL tests are run at the same time as other programs, not particularly temperature dependent, and not a problem at normal CPU MHz and voltage settings.

Then I thought that I would see would happen at 80°C with no overclocking (no real concerns at RPi cost). I heated up the RPi using close proximity of a 60W bulb, whilst running an integer test program, running at 580 to 640 MIPS. On reaching 80°C, the Temperature Monitor Panel Applet changes to red. I kept running until nearly 84°C and there was no change if processing speed. Maybe I will have another go with the OpenGL program until it crashes - for the fun of affordable destruction?

[mcgyver83]

Hi,

I'm looking for a test suite that can provide some feelings regarding performances inprovment reached by overclock.
Can you suggest a suite or a single utility that can run a set of test?

My benchmarks are described in the following, with normal and overclocked results shown.

http://www.roylongbottom.org.uk/Raspber ... hmarks.htm

Benchmarks and source codes are in the following. Downloaded benchmarks need execution permission setting. You can change source codes if you want - all FREE.

http://www.roylongbottom.org.uk/Raspber ... hmarks.zip

After setting execution permission there is a way to run the tests like a suite or I have to run them like

Code: Select all

test1 && test2 && test3...

I would be very interested in the settings and code that is running to get 80 degrees, as well as the enviromental conditions.

I have not managed to get any of my RPi systems to reach greater than 62 degrees with an ambient temperature of up to 38 degrees C and humidity ranging from 20% to 85% saturation. This with force_turbo=1, and overclocked to maximum stable:

Code: Select all

force_turbo=1
over_voltage=6
arm_freq=1000
sdram_frep=500
core_freq=400

And even under heavy load for hours at a time. I have intentionaly tested at the highest ambient tempurature that one is likely to run there system in (38 degrees C) under heavy load (CPU at 100%, and heavily using GPU) in order to assure the stability of the system.

So I would be very interested to know your operating conditions, and what you were running in order to get your RPi to report anything close to 80 degrees.

David why you force the rpi to stay at the highest frequency with "force_turbo"?

[RoyLongbottom]

Hi,
I'm After setting execution permission there is a way to run the tests like a suite or I have to run them like

Code: Select all

test1 && test2 && test3...

Using the & between programs, runs them all at the same time and mixes up output in one window. The following script file (classic.sh) will run my Classic Benchmarks sequentially, The NoPause parameter avoid stopping for user input comments at the end. The script file needs execution permission and is run from a command line (./classic.sh).

Code: Select all

./dhrystonePiA6 NoPause
./linpackPiA6 NoPause
./linpackPiSP NoPause
./liverloopsPiA6 NoPause
./whetstonePiA6 NoPause

To run them all at the same time from different LXTerminal windows, use:

Code: Select all

lxterminal -e ./dhrystonePiA6 NoPause
lxterminal -e ./linpackPiA6 NoPause
lxterminal -e ./linpackPiSP NoPause
lxterminal -e ./liverloopsPiA6 NoPause
lxterminal -e ./whetstonePiA6 NoPause

[DavidS]

I would be very interested in the settings and code that is running to get 80 degrees, as well as the enviromental conditions.

I have not managed to get any of my RPi systems to reach greater than 62 degrees with an ambient temperature of up to 38 degrees C and humidity ranging from 20% to 85% saturation. This with force_turbo=1, and overclocked to maximum stable:

Code: Select all

force_turbo=1
over_voltage=6
arm_freq=1000
sdram_frep=500
core_freq=400

And even under heavy load for hours at a time. I have intentionaly tested at the highest ambient tempurature that one is likely to run there system in (38 degrees C) under heavy load (CPU at 100%, and heavily using GPU) in order to assure the stability of the system.

So I would be very interested to know your operating conditions, and what you were running in order to get your RPi to report anything close to 80 degrees.

David why you force the rpi to stay at the highest frequency with "force_turbo"?

As far as I know no one has yet created a module for dynamic overclocking on RISC OS.

And I have not looked into if this could be handled by the GPU MailBox protocal as of yet.

[DavidS]

I would be very interested in the settings and code that is running to get 80 degrees, as well as the enviromental conditions.

I have not managed to get any of my RPi systems to reach greater than 62 degrees with an ambient temperature of up to 38 degrees C and humidity ranging from 20% to 85% saturation. This with force_turbo=1, and overclocked to maximum stable:

Code: Select all

force_turbo=1
over_voltage=6
arm_freq=1000
sdram_frep=500
core_freq=400

And even under heavy load for hours at a time. I have intentionaly tested at the highest ambient tempurature that one is likely to run there system in (38 degrees C) under heavy load (CPU at 100%, and heavily using GPU) in order to assure the stability of the system.

So I would be very interested to know your operating conditions, and what you were running in order to get your RPi to report anything close to 80 degrees.

David why you force the rpi to stay at the highest frequency with "force_turbo"?

As far as I know no one has yet created a module for dynamic overclocking on RISC OS.

And I have not looked into if this could be handled by the GPU MailBox protocal as of yet.

As for GPU related stuff I should add, that even if I boot into Linux I strongly prefer to keep the setings the same, as it provides consistancy.

[mcgyver83]

Thanks for the script.

Raspberry and other system (correct me if is not a feature of rpi but something related to the cpu/gpu) is able to scale the frequency when not user.
In config.txt you can set core_freq=x and core_freq_min=y

[RoyLongbottom]

Thanks for the script.

Raspberry and other system (correct me if is not a feature of rpi but something related to the cpu/gpu) is able to scale the frequency when not user.
In config.txt you can set core_freq=x and core_freq_min=y

The MHz settings can be changed using the following command via overclock

sudo raspi-config

I am just looking at the following command that displays current MHz settings, with a view to including it in some of my benchmarks (command new to me too).

vcgencmd get_config int

[dom]

As far as I know no one has yet created a module for dynamic overclocking on RISC OS.

And I have not looked into if this could be handled by the GPU MailBox protocal as of yet.

That will work. The linux cpufreq driver uses the mailbox interface.

[scruss]

Using the & between programs, runs them all at the same time …

A single ampersand would. Double ampersand is logical AND (to sh-like shells, at least) so it would run them sequentially until one failed.

[mcgyver83]

Using the & between programs, runs them all at the same time …

A single ampersand would. Double ampersand is logical AND (to sh-like shells, at least) so it would run them sequentially until one failed.

Thanks,I'm used to run command like ap-get update && apt-get upgrade and it works as I expect: when the first command ends the second is executed if no errors happened.

[RoyLongbottom]

I have revised my DriveSpeed benchmark, used for measuring performance of SD cards and USB drives. The original used the same C code as a version to run via Android but produced some strange results on RasPi, with random access and small files tests. The revised version produces similar results to the original Linux benchmark and has, itself, been compiled and run via Linux. Further details, results, source code and execution file are available in:

http://www.roylongbottom.org.uk/Raspber ... m#anchor20
http://www.roylongbottom.org.uk/Raspber ... hmarks.zip

SD system cards - many can have similar large file reading speeds and booting times, but there are wide variations on writing speeds.

USB Sticks Format - Ext4 is faster than with FAT formatting, mainly on writing speeds, but particularly with small files, where the difference can be 4 to 5 times.

RPi vs Linux - not much difference on slower USB drives, but Rpi becomes more bus speed limited on faster devices.

As a reminder, the benchmark measures writing and reading speeds of large files, small files and random access. I will be pleased to receive results for newer faster devices.

[dom]

USB Sticks Format - Ext4 is faster than with FAT formatting, mainly on writing speeds, but particularly with small files, where the difference can be 4 to 5 times.

There are some benchmarks in this thread:
http://forum.xbmc.org/showthread.php?ti ... pid1577721

including a ridiculously good f2fs score for copying many small files. I'd be interested in if you can test f2fs with your benchmark.

[RoyLongbottom]

There are some benchmarks in this thread:
http://forum.xbmc.org/showthread.php?ti ... pid1577721

including a ridiculously good f2fs score for copying many small files. I'd be interested in if you can test f2fs with your benchmark.[/quote]

I will as soon as I can find how to install f2fs to format USB drives (if it is appropriate) and SD cards, including a suitable OS on the RPi. I spent an hour yesterday and it seemed that numerous people are experimenting but found no simple downloads or instructions. I can install a suitable version of Linux to use a PC to format drives [more googling later] .

See:

http://www.raspberrypi.org/phpBB3/viewt ... 31&t=65330

[CalcProgrammer1]

Thanks for these benchmarks! I tested them on other ARM boards and they work as well (haven't tried recompiling to take advantage of armv7 compiling yet). I just compared my Model B Pi to my Galaxy Note 3 with a Debian chroot (Qualcomm Snapdragon S800) and wow, huge difference in performance. Have a CubieBoard that I need to fix the rootfs on and I'll test on that as well.

dhrystonePiA6 on RasPi - 855.06 MIPS rating
dhrystonePiA6 on Note3 - 4856.76 MIPS rating

liverloopsPiA6 on RasPi - 149.75 max / 64.92 avg
liverloopsPiA6 on Note3 - 1410.10 max / 546.87 avg

linpackPiA6 on RasPi - 43.43 Mflops
linpackPiA6 on Note3 - 647.42 Mflops

linpackPiSP on RasPi - 57.87 Mflops
linpackPiSP on Note3 - 834.65 Mflops

Do these take advantage of multicore processors?

[RoyLongbottom]

The benchmarks are also available for Android. See results in the following to see how good
the Snapdragon S800 really is:

http://www.roylongbottom.org.uk/android ... nchorStart

There are MP benchmarks for RPi and the same for Android see:

http://www.roylongbottom.org.uk/Raspber ... hmarks.htm

http://www.roylongbottom.org.uk/android ... hmarks.htm

http://www.roylongbottom.org.uk/android ... hmarks.htm

Note Linpack NEON MP is not much use as it is based on the original N100 code. Users are allowed to use their own linear equation solver for real MP use.

Some of Android benchmarks are for ARM V5, ARM V7 and NEON instructions.