Page 2 of 3
Re: Heat sink adhesive
Posted: Thu May 05, 2016 11:33 pm
by Heater
Rive,
OK, done that. What is supposed to happen? I get:
Code: Select all
frequency(45)=1200000000
temp=63.9'C
frequency(45)=1200000000
temp=63.9'C
frequency(45)=1200000000
temp=63.4'C
frequency(45)=1200000000
temp=62.8'C
frequency(45)=1200000000
temp=62.3'C
frequency(45)=1200002000
temp=63.4'C
--------------------------------------------------------------------------------
||Ax-b||_oo/(eps*(||A||_oo*||x||_oo+||b||_oo)*N)= 0.0025941 ...... PASSED
================================================================================
Finished 1 tests with the following results:
1 tests completed and passed residual checks,
0 tests completed and failed residual checks,
0 tests skipped because of illegal input values.
--------------------------------------------------------------------------------
End of Tests.
================================================================================
Re: Heat sink adhesive
Posted: Thu May 05, 2016 11:38 pm
by Heater
OK, I ran it again, much the same:
Code: Select all
frequency(45)=1200000000
temp=65.5'C
frequency(45)=1200002000
temp=66.1'C
frequency(45)=1200000000
temp=66.6'C
--------------------------------------------------------------------------------
||Ax-b||_oo/(eps*(||A||_oo*||x||_oo+||b||_oo)*N)= 0.0025941 ...... PASSED
================================================================================
Finished 1 tests with the following results:
1 tests completed and passed residual checks,
0 tests completed and failed residual checks,
0 tests skipped because of illegal input values.
--------------------------------------------------------------------------------
End of Tests.
================================================================================
Re: Heat sink adhesive
Posted: Fri May 06, 2016 12:09 am
by Heater
OK Rive,
What is that crazy program of yours doing?
You managed to crash my Pi 3. The last run went like so:
Code: Select all
- The matrix A is randomly generated for each test.
- The following scaled residual check will be computed:
||Ax-b||_oo / ( eps * ( || x ||_oo * || A ||_oo + || b ||_oo ) * N )
- The relative machine precision (eps) is taken to be 1.110223e-16
- Computational tests pass if scaled residuals are less than 16.0
temp=46.2'C
frequency(45)=1199998000
temp=48.9'C
frequency(45)=1200000000
temp=50.5'C
frequency(45)=1199998000
temp=50.5'C
frequency(45)=1200000000
temp=51.0'C
frequency(45)=1200000000
temp=51.5'C
frequency(45)=1200000000
temp=51.5'C
Took a power cycle to get control back after that.
Presumably nothing to do with overheating given the temps reported there.
But then, on a previous run the thing bombed at start up with an "illegal instruction" error.
Re: Heat sink adhesive
Posted: Fri May 06, 2016 12:13 am
by Heater
Rive,
Like so:
Code: Select all
frequency(45)=1200000000
temp=51.5'C
frequency(45)=1199998000
temp=52.1'C
Illegal instruction
pi@raspberrypi3:~ $ frequency(45)=600000000
WTF?
Re: Heat sink adhesive
Posted: Fri May 06, 2016 1:35 am
by Rive
On this, your pi will freeze/lock up and become unresponsive (crash) if you overheat.
Results like this (6.6 Gflops)
Code: Select all
@raspberrypi:~ $ ./xhpl
================================================================================
HPLinpack 2.1 -- High-Performance Linpack benchmark -- October 26, 2012
Written by A. Petitet and R. Clint Whaley, Innovative Computing Laboratory, UTK
Modified by Piotr Luszczek, Innovative Computing Laboratory, UTK
Modified by Julien Langou, University of Colorado Denver
================================================================================
An explanation of the input/output parameters follows:
T/V : Wall time / encoded variant.
N : The order of the coefficient matrix A.
NB : The partitioning blocking factor.
P : The number of process rows.
Q : The number of process columns.
Time : Time in seconds to solve the linear system.
Gflops : Rate of execution for solving the linear system.
The following parameter values will be used:
N : 8000
NB : 256
PMAP : Row-major process mapping
P : 1
Q : 1
PFACT : Left
NBMIN : 2
NDIV : 2
RFACT : Right
BCAST : 2ring
DEPTH : 0
SWAP : Mix (threshold = 64)
L1 : transposed form
U : transposed form
EQUIL : yes
ALIGN : 8 double precision words
--------------------------------------------------------------------------------
- The matrix A is randomly generated for each test.
- The following scaled residual check will be computed:
||Ax-b||_oo / ( eps * ( || x ||_oo * || A ||_oo + || b ||_oo ) * N )
- The relative machine precision (eps) is taken to be 1.110223e-16
- Computational tests pass if scaled residuals are less than 16.0
================================================================================
T/V N NB P Q Time Gflops
--------------------------------------------------------------------------------
WR02R2L2 8000 256 1 1 51.47 6.634e+00
HPL_pdgesv() start time Sun Apr 24 09:14:53 2016
HPL_pdgesv() end time Sun Apr 24 09:15:44 2016
--------------------------------------------------------------------------------
||Ax-b||_oo/(eps*(||A||_oo*||x||_oo+||b||_oo)*N)= 0.0025941 ...... PASSED
================================================================================
Finished 1 tests with the following results:
1 tests completed and passed residual checks,
0 tests completed and failed residual checks,
0 tests skipped because of illegal input values.
--------------------------------------------------------------------------------
End of Tests.
================================================================================
pi@raspberrypi:~ $ ./xhpl
================================================================================
HPLinpack 2.1 -- High-Performance Linpack benchmark -- October 26, 2012
Written by A. Petitet and R. Clint Whaley, Innovative Computing Laboratory, UTK
Modified by Piotr Luszczek, Innovative Computing Laboratory, UTK
Modified by Julien Langou, University of Colorado Denver
================================================================================
An explanation of the input/output parameters follows:
T/V : Wall time / encoded variant.
N : The order of the coefficient matrix A.
NB : The partitioning blocking factor.
P : The number of process rows.
Q : The number of process columns.
Time : Time in seconds to solve the linear system.
Gflops : Rate of execution for solving the linear system.
The following parameter values will be used:
N : 8000
NB : 256
PMAP : Row-major process mapping
P : 1
Q : 1
PFACT : Left
NBMIN : 2
NDIV : 2
RFACT : Right
BCAST : 2ring
DEPTH : 0
SWAP : Mix (threshold = 64)
L1 : transposed form
U : transposed form
EQUIL : yes
ALIGN : 8 double precision words
--------------------------------------------------------------------------------
- The matrix A is randomly generated for each test.
- The following scaled residual check will be computed:
||Ax-b||_oo / ( eps * ( || x ||_oo * || A ||_oo + || b ||_oo ) * N )
- The relative machine precision (eps) is taken to be 1.110223e-16
- Computational tests pass if scaled residuals are less than 16.0
================================================================================
T/V N NB P Q Time Gflops
--------------------------------------------------------------------------------
WR02R2L2 8000 256 1 1 51.69 6.605e+00
HPL_pdgesv() start time Sun Apr 24 09:16:14 2016
HPL_pdgesv() end time Sun Apr 24 09:17:06 2016
--------------------------------------------------------------------------------
||Ax-b||_oo/(eps*(||A||_oo*||x||_oo+||b||_oo)*N)= 0.0025941 ...... PASSED
================================================================================
Finished 1 tests with the following results:
1 tests completed and passed residual checks,
0 tests completed and failed residual checks,
0 tests skipped because of illegal input values.
--------------------------------------------------------------------------------
End of Tests
Re: Heat sink adhesive
Posted: Fri May 06, 2016 1:43 am
by Heater
Rive,
If your pi locked up/crashed it probably overheated.
There is no such evidence in my results. See above.
I only carried out your instructions and got the results I presented.
Whatever failure is going on does not seem to be temperature related
.
Re: Heat sink adhesive
Posted: Fri May 06, 2016 1:47 am
by Rive
Heater wrote:Rive,
If your pi locked up/crashed it probably overheated.
There is no such evidence in my results. See above.
I only carried out your instructions and got the results I presented.
Whatever failure is going on does not seem to be temperature related
.
If you crashed it was either you overheated, or you have voltage issues. The benchmark heats you up so fast that it just crashes. Even the temp read out just locks up (as does everything else)
Show you Gflops (you did not show them in the above)
Re: Heat sink adhesive
Posted: Fri May 06, 2016 1:48 am
by Heater
Here we go again:
Code: Select all
frequency(45)=600000000
temp=60.1'C
frequency(45)=1200000000
temp=61.2'C
--------------------------------------------------------------------------------
||Ax-b||_oo/(eps*(||A||_oo*||x||_oo+||b||_oo)*N)= 0.0025941 ...... PASSED
================================================================================
Finished 1 tests with the following results:
1 tests completed and passed residual checks,
0 tests completed and failed residual checks,
0 tests skipped because of illegal input values.
--------------------------------------------------------------------------------
End of Tests.
================================================================================
Re: Heat sink adhesive
Posted: Fri May 06, 2016 1:52 am
by Heater
Rive,
Show you Gflops (you did not show them in the above)
I show all the output of the program as given. What else should I do?
Re: Heat sink adhesive
Posted: Fri May 06, 2016 1:57 am
by Heater
There it goes again. A lock up:
Code: Select all
- The matrix A is randomly generated for each test.
- The following scaled residual check will be computed:
||Ax-b||_oo / ( eps * ( || x ||_oo * || A ||_oo + || b ||_oo ) * N )
- The relative machine precision (eps) is taken to be 1.110223e-16
- Computational tests pass if scaled residuals are less than 16.0
frequency(45)=1200000000
temp=49.4'C
frequency(45)=1200000000
temp=49.9'C
frequency(45)=1200002000
temp=51.5'C
frequency(45)=1200000000
temp=51.5'C
frequency(45)=1200000000
temp=52.6'C
frequency(45)=1200000000
temp=52.6'C
Not such a high temperature apparently.
Re: Heat sink adhesive
Posted: Fri May 06, 2016 2:00 am
by Rive
It heats up too fast (unless you are running an insufficient power supply)
Look at mine, see what I posted? You are posting just a little, not all.
Run just this command, post all when test completes
Then run this for 15 min, and post all (if you don't crash before then, or overheat)
Code: Select all
wget https://raw.githubusercontent.com/ssvb/cpuburn-arm/master/cpuburn-a53.S
gcc -o cpuburn-a53 cpuburn-a53.S
while true; do vcgencmd measure_clock arm; vcgencmd measure_temp; sleep 2; done& ./cpuburn-a53
Re: Heat sink adhesive
Posted: Fri May 06, 2016 2:09 am
by Heater
Rive,
"Post all" is tricky as my Pi is headless and I talk to it via Dataplicity. But your first suggestion ends like this:
Code: Select all
================================================================================
T/V N NB P Q Time Gflops
--------------------------------------------------------------------------------
WR02R2L2 8000 256 1 1 98.30 3.473e+00
HPL_pdgesv() start time Fri May 6 05:03:46 2016
HPL_pdgesv() end time Fri May 6 05:05:24 2016
--------------------------------------------------------------------------------
||Ax-b||_oo/(eps*(||A||_oo*||x||_oo+||b||_oo)*N)= 0.0025941 ...... PASSED
================================================================================
Finished 1 tests with the following results:
1 tests completed and passed residual checks,
0 tests completed and failed residual checks,
0 tests skipped because of illegal input values.
--------------------------------------------------------------------------------
End of Tests.
================================================================================
Much the same as before.
Re: Heat sink adhesive
Posted: Fri May 06, 2016 2:13 am
by Rive
Heater wrote:Rive,
"Post all" is tricky as my Pi is headless and I talk to it via Dataplicity. But your first suggestion ends like this:
Code: Select all
================================================================================
T/V N NB P Q Time Gflops
--------------------------------------------------------------------------------
WR02R2L2 8000 256 1 1 98.30 3.473e+00
HPL_pdgesv() start time Fri May 6 05:03:46 2016
HPL_pdgesv() end time Fri May 6 05:05:24 2016
--------------------------------------------------------------------------------
||Ax-b||_oo/(eps*(||A||_oo*||x||_oo+||b||_oo)*N)= 0.0025941 ...... PASSED
================================================================================
Finished 1 tests with the following results:
1 tests completed and passed residual checks,
0 tests completed and failed residual checks,
0 tests skipped because of illegal input values.
--------------------------------------------------------------------------------
End of Tests.
================================================================================
Much the same as before.
Is this a pi3, because 3.4 Gflops is really bad (6 Gflops is average for a stock pi3). May indicate throttling due to overheating, or you are underpowered.
Re: Heat sink adhesive
Posted: Fri May 06, 2016 2:17 am
by Heater
Part 2:
Code: Select all
$ while true; do vcgencmd measure_clock arm; vcgencmd measure_temp; sleep 2; done& ./cpuburn-a53
[1] 844
frequency(45)=600062000
temp=47.8'C
frequency(45)=600000000
temp=52.6'C
frequency(45)=600000000
temp=55.8'C
frequency(45)=600000000
temp=58.5'C
frequency(45)=600000000
temp=60.1'C
frequency(45)=600000000
temp=61.2'C
frequency(45)=600000000
temp=62.3'C
frequency(45)=600000000
temp=63.9'C
Yep. You crashed my Pi again!
Re: Heat sink adhesive
Posted: Fri May 06, 2016 2:19 am
by Rive
Heater wrote:Part 2:
Code: Select all
$ while true; do vcgencmd measure_clock arm; vcgencmd measure_temp; sleep 2; done& ./cpuburn-a53
[1] 844
frequency(45)=600062000
temp=47.8'C
frequency(45)=600000000
temp=52.6'C
frequency(45)=600000000
temp=55.8'C
frequency(45)=600000000
temp=58.5'C
frequency(45)=600000000
temp=60.1'C
frequency(45)=600000000
temp=61.2'C
frequency(45)=600000000
temp=62.3'C
frequency(45)=600000000
temp=63.9'C
Yep. You crashed my Pi again!
What is your power supply? Because your Pi3 is slow as molasses. You didnt even hit 64C (@ 600 MHz ) before crashing. Should have scored way higher for Gflops too. You have slightly better performance than a stock Pi2 (which scores around 1 Gflop).
I run that at 50C btw...
Re: Heat sink adhesive
Posted: Fri May 06, 2016 2:26 am
by Heater
Rive,
Is this a pi3,
The board says: Raspberry Pi 3 Model B V1.2
Power supply I can look into. Just now I have to sleep. Be back later...
Re: Heat sink adhesive
Posted: Fri May 06, 2016 2:27 am
by Rive
Heater wrote:Rive,
Is this a pi3,
The board says: Raspberry Pi 3 Model B V1.2
Power supply I can look into. Just now I have to sleep. Be back later...
I think you have voltage issues (are you stock or overclocked?), That would also explain why your not overheating... Pi3 is under powered, and will literally crash before it can throttle/overheat...Or it is just defective. You score half the Gflops that a normal Pi3 scores on the bench (major red flag).
Power supply should be 5.1V 2.5A or 5.25V 2.4A
Re: Heat sink adhesive
Posted: Fri May 06, 2016 6:50 am
by Pithagoros
MarkHaysHarris777 wrote:Pithagoros wrote:
[heatsink] Definitely not required on my Pi 3Bs, I don't use heatisinks and the highest temperature I've seen is 68C.
You haven't loaded it yet (this is not a matter of debate either... well proven, and well understood).
Your assertion that you made unconditionally was that heatsinks are required. Regardless of whether I've loaded my Pi3Bs, heatsinks are clearly not required. But in fact they do come under a high load (although you claim to know that they don't and it's "not a matter for debate and well proven and understood that I haven't loaded my Pis
)
I don't use a case, a case has absolutely no useful purpose in my use of the Pi so I don't waste my time and money on them.
Other circumstances do exist.
Re: Heat sink adhesive
Posted: Fri May 06, 2016 11:33 am
by Heater
Rive,
Yesterday I was a happy bunny. I had a Pi 3. It worked. I calculated 100 million digits of PI for me.
Today, thanks to you, I'm miserable. I have a Pi 3 that immediately crashes when I run ./xhpl
I have tried every combination of power supply and USB cable I have around here.
Just now it's running of a Sandsrøm USB Power Bank, 12000mAh. capable of delivering 3.4 Amps. (Same one I used last night but now full charged). It puts 5.04 volts on the 5v header pin.
There is no over clocking going on here. All stock, out of the box.
Now to scour the junk pile for a suitable heat sink. Be back later....
Re: Heat sink adhesive
Posted: Fri May 06, 2016 12:12 pm
by jojopi
Heater wrote:I have a Pi 3 that immediately crashes when I run ./xhpl
There was a thread about that application, which you even posted in.
viewtopic.php?f=63&t=139712
Have you tried the official suggestion of a modest over_voltage in config.txt? Personally I prefer a slight underclock. It is not absolutely necessary to go to freon.
Re: Heat sink adhesive
Posted: Fri May 06, 2016 1:23 pm
by Heater
So I did, jojopi. I didn't have a Pi 3 at that time I wasn't paying so much attention.
Re: Heat sink adhesive
Posted: Fri May 06, 2016 1:53 pm
by Rive
It puts 5.04 volts on the 5v header pin
Yep. So your issue is insufficient power for the Pi3. That is why it crashes before the Pi3 can even fully utilize the SoC /ARM (and why you claim to never suffer from overheating. It can't even get close to the 80C throttling threshold
).
A Pi3 needs at least 5.1v (I use 5.25V), and also, you need to make sure you have enough Amps (2.4A/2.5A)
Or, just get used to the idea of an under powered pi3 that performs marginally better than a Pi2 (at 3 Gflops...where a properly powered stock Pi3 is at 6 Gflops, and an overclocked pi2 is around 1 Gflops). On the bright side, your under performing/under powered Pi3 will not ever require, or need a heatsink!
Re: Heat sink adhesive
Posted: Fri May 06, 2016 2:31 pm
by joan
Rive wrote:
...
A Pi3 needs at least 5.1v (I use 5.25V),
...
I don't believe that. Have you any evidence?
Re: Heat sink adhesive
Posted: Fri May 06, 2016 2:55 pm
by Rive
joan wrote:Rive wrote:
...
A Pi3 needs at least 5.1v (I use 5.25V),
...
I don't believe that. Have you any evidence?
Pi3 power supply draw (at idle with only 19" monitor, keyboard, mouse):
https://www.element14.com/community/com ... hmark-ever
5.19 V 1.141 W
0.22 A 00006 mAh
Re: Heat sink adhesive
Posted: Fri May 06, 2016 3:45 pm
by Heater
Rive,
A Pi3 needs at least 5.1v
That can't be right. The SoC does not run off of 5v it runs at 3.3v. There is an on board 3.3v regulator which should cater for any little variation in the 5v.
For reference I measure 3.32v on the header pin.
Edit: Unless there is some other regulator for the SoC that I am not aware of.