pxgator
Posts: 105
Joined: Mon Feb 16, 2015 6:45 pm
Location: Southern Colorado, USA

Re: Cooling the Pi3 with a fan.

Sun Apr 10, 2016 3:13 am

Sorry.....I don't have a camera to take any pics so this is the best that
I can do:
http://www.ebay.com/itm/221950553658?_t ... EBIDX%3AIT
I can give you a tip though. Get rid of the thermal tape and use a very thin
coat of thermal glue. If possible you might also use a larger heatsink. Put
your fan close as possible so that the airflow is concentrated on the SOC.
Maybe a smaller diameter fan...??
So what's all this RPi stuff anyhow? Well folks, it's a feat of engineering from
the UK almost as remarkable as the De Havilland Mosquito and the Colossus.

User avatar
Rive
Posts: 586
Joined: Sat Mar 26, 2016 5:21 pm
Location: USA

Re: Cooling the Pi3 with a fan.

Sun Apr 10, 2016 3:25 am

pxgator wrote:Sorry.....I don't have a camera to take any pics so this is the best that
I can do:
http://www.ebay.com/itm/221950553658?_t ... EBIDX%3AIT
I can give you a tip though. Get rid of the thermal tape and use a very thin
coat of thermal glue. If possible you might also use a larger heatsink. Put
your fan close as possible so that the airflow is concentrated on the SOC.
Maybe a smaller diameter fan...??
I ripped of that crap adhesive tape soon as i got the heatsink.

I was running AS5 thermal paste on the SoC until today. it held it fairly well too. I took it off though, and replaced it with some better thermal adhesive i had, just to see if it made any difference... made no difference at all. Maybe 1C or so. Course, I didn't let the AS5 burn in...may be worth it to go back to the thermal paste (if I can get the heatsink off now).


Maybe I'll get a smaller fan at a higher speed/CFM, or even an aluminum heatsink...
DNPNWO

pxgator
Posts: 105
Joined: Mon Feb 16, 2015 6:45 pm
Location: Southern Colorado, USA

Re: Cooling the Pi3 with a fan.

Sun Apr 10, 2016 3:37 am

Rive wrote:
Maybe I'll get a smaller fan at a higher speed/CFM, or even an aluminum heatsink...
Good idea. I'm a retired (old geezer) electronic technician that spent many years dealing
with cooling issues on semiconductors. Such as IGBT modules in 3 phase motor drives.
Air flow direction, concentration, and escape paths can be crucial in obtaining proper
cooling. You're on the right track so don't give up until you get the results you are needing.

Edit: @ Rive...I sent you a pm.
So what's all this RPi stuff anyhow? Well folks, it's a feat of engineering from
the UK almost as remarkable as the De Havilland Mosquito and the Colossus.

User avatar
Rive
Posts: 586
Joined: Sat Mar 26, 2016 5:21 pm
Location: USA

Re: Cooling the Pi3 with a fan.

Wed Apr 13, 2016 6:05 pm

pxgator wrote:
Rive wrote:
Maybe I'll get a smaller fan at a higher speed/CFM, or even an aluminum heatsink...
Good idea. I'm a retired (old geezer) electronic technician that spent many years dealing
with cooling issues on semiconductors. Such as IGBT modules in 3 phase motor drives.
Air flow direction, concentration, and escape paths can be crucial in obtaining proper
cooling. You're on the right track so don't give up until you get the results you are needing.

Edit: @ Rive...I sent you a pm.

Thanks...Awesome!!

I managed to shave nearly 20C off Cpuburn-a53. This new heatsink (with same fan) cools at near 63C, and stress runs at 43C. I doubt I will see better results than that. idle is at 27C.


Image
Image
Image


Image

Overclocks:

Code: Select all

dtparam=sd_overclock=100
arm_freq=1300
core_freq=500
over_voltage=4
sdram_freq=575
sdram_schmoo=0x02000020
over_voltage_sdram_p=6
over_voltage_sdram_i=4
over_voltage_sdram_c=4
v3d_freq=500
h264_freq=333
gpu_mem=256
cpuburn-a63 (63C):

Code: Select all

[email protected]:~ $ while true; do vcgencmd measure_clock arm; vcgencmd measure_temp; sleep 10; done& ./cpuburn-a53
[1] 1514
frequency(45)=600000000
temp=30.6'C
frequency(45)=1300000000
temp=50.5'C
frequency(45)=1300000000
temp=53.7'C
frequency(45)=1299998000
temp=56.4'C
frequency(45)=1300002000
temp=58.0'C
frequency(45)=1300000000
temp=58.5'C
frequency(45)=1300000000
temp=59.1'C
frequency(45)=1300000000
temp=60.7'C
frequency(45)=1300000000
temp=61.2'C
frequency(45)=1299998000
temp=61.2'C
frequency(45)=1300000000
temp=61.2'C
frequency(45)=1300000000
temp=62.3'C
frequency(45)=1300002000
temp=62.3'C
frequency(45)=1300000000
temp=62.3'C
frequency(45)=1300000000
temp=62.3'C
frequency(45)=1299998000
temp=62.3'C
frequency(45)=1300000000
temp=62.3'C
frequency(45)=1299998000
temp=62.3'C
frequency(45)=1300000000
temp=62.3'C
frequency(45)=1300000000
temp=62.3'C
frequency(45)=1300000000
temp=62.3'C
frequency(45)=1300000000
temp=62.8'C
frequency(45)=1299998000
temp=63.4'C
frequency(45)=1300000000
temp=62.8'C
frequency(45)=1300000000
temp=62.8'C
frequency(45)=1300000000
temp=62.3'C
frequency(45)=1300000000
temp=62.3'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=62.8'C
frequency(45)=1299998000
temp=62.8'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=62.8'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=62.8'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=62.8'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.9'C
frequency(45)=1300000000
temp=62.8'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=62.3'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300002000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1299998000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=62.8'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=62.8'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=62.8'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1299998000
temp=63.4'C
frequency(45)=1299998000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=62.3'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300002000
temp=63.4'C
frequency(45)=1299998000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=62.8'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=62.3'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C
frequency(45)=1300000000
temp=63.4'C

Code: Select all

top - 13:52:09 up 32 min,  1 user,  load average: 4.00, 4.00, 3.40
Tasks: 137 total,   5 running, 132 sleeping,   0 stopped,   0 zombie
%Cpu0  : 99.7 us,  0.3 sy,  0.0 ni,  0.0 id,  0.0 wa,  0.0 hi,  0.0 si,  0.0 st
%Cpu1  :100.0 us,  0.0 sy,  0.0 ni,  0.0 id,  0.0 wa,  0.0 hi,  0.0 si,  0.0 st
%Cpu2  :100.0 us,  0.0 sy,  0.0 ni,  0.0 id,  0.0 wa,  0.0 hi,  0.0 si,  0.0 st
%Cpu3  :100.0 us,  0.0 sy,  0.0 ni,  0.0 id,  0.0 wa,  0.0 hi,  0.0 si,  0.0 st
KiB Mem:    753132 total,   512096 used,   241036 free,    34036 buffers
KiB Swap:   102396 total,        0 used,   102396 free.   353964 cached Mem

  PID USER      PR  NI    VIRT    RES    SHR S  %CPU %MEM     TIME+ COMMAND     
 1517 pi        20   0    1680     84     16 R 100.0  0.0  16:07.04 cpuburn-a53 
 1515 pi        20   0    1680    328    272 R 100.0  0.0  16:06.10 cpuburn-a53 
 1519 pi        20   0    1680     88     16 R  99.6  0.0  15:57.17 cpuburn-a53 
 1518 pi        20   0    1680     84     16 R  95.7  0.0  16:04.50 cpuburn-a53 
 1032 pi        20   0   46852  19112  16416 S   2.3  2.5   0:03.97 lxterminal  
  639 root      20   0  184828  50420  26516 S   1.7  6.7   0:14.60 Xorg        
  869 pi        20   0   91680  24620  20360 S   0.7  3.3   0:14.16 lxpanel     
  864 pi        20   0   20644  11936   9568 S   0.3  1.6   0:02.05 openbox     
 1167 pi        20   0    5112   2532   2164 R   0.3  0.3   0:05.66 top         
    1 root      20   0   23904   3984   2736 S   0.0  0.5   0:06.75 systemd     
    2 root      20   0       0      0      0 S   0.0  0.0   0:00.00 kthreadd    
    3 root      20   0       0      0      0 S   0.0  0.0   0:00.06 ksoftirqd/0 
    5 root       0 -20       0      0      0 S   0.0  0.0   0:00.00 kworker/0:+ 
    7 root      20   0       0      0      0 S   0.0  0.0   0:00.27 rcu_sched   
Last edited by Rive on Wed Apr 13, 2016 6:36 pm, edited 3 times in total.
DNPNWO

pxgator
Posts: 105
Joined: Mon Feb 16, 2015 6:45 pm
Location: Southern Colorado, USA

Re: Cooling the Pi3 with a fan.

Wed Apr 13, 2016 6:16 pm

Rive wrote:
Thanks...Awesome!!

I managed to shave nearly 20C off Cpuburn-a53. This new heatsink (with same fan) cools at near 63C, and stress runs at 43C. I doubtt I will see better results than that. idle is at 27C.
Your quite welcome......hope you have fun with your 'extreme pi' :D .

Edit: Some pics of your new setup??
So what's all this RPi stuff anyhow? Well folks, it's a feat of engineering from
the UK almost as remarkable as the De Havilland Mosquito and the Colossus.

User avatar
Rive
Posts: 586
Joined: Sat Mar 26, 2016 5:21 pm
Location: USA

Re: Cooling the Pi3 with a fan.

Wed Apr 13, 2016 6:32 pm

pxgator wrote:
Rive wrote:
Thanks...Awesome!!

I managed to shave nearly 20C off Cpuburn-a53. This new heatsink (with same fan) cools at near 63C, and stress runs at 43C. I doubtt I will see better results than that. idle is at 27C.
Your quite welcome......hope you have fun with your 'extreme pi' :D .

Edit: Some pics of your new setup??
Image

Image

Image
DNPNWO

User avatar
Gerd
Posts: 66
Joined: Wed Mar 16, 2016 10:48 am
Location: Europe

Re: Cooling the Pi3 with a fan.

Sat Apr 23, 2016 9:33 am

Rive wrote:I give you 5 min tops, before you hit the ceiling, slow to a crawl, and revert to 600MHz (Hell, you probably couldn't even manage to do it with 'Stress' (which is a lightweight contender compared to cpuburn)).

C'mon, show me how wrong I am. I am waiting.
No vid, sreenshot must suffic.
cpuburn passive.png
cpuburn passive.png (32.4 KiB) Viewed 3296 times
Now here is my challenge for you:
Get linpack installed and get over to the linpack thread and post some numbers. My guess: with your current oc settings you are posting around you will fail. Hard.

User avatar
Rive
Posts: 586
Joined: Sat Mar 26, 2016 5:21 pm
Location: USA

Re: Cooling the Pi3 with a fan.

Sat Apr 23, 2016 12:52 pm

Gerd wrote: Now here is my challenge for you:
Get linpack installed and get over to the linpack thread and post some numbers. My guess: with your current oc settings you are posting around you will fail. Hard.
I am not gonna hunt it down, put it here what you want me to install/run for linpack and I will give it ago.

for example, If I wanted someone to run cpuburn-a53:

Code: Select all

wget https://raw.githubusercontent.com/ssvb/cpuburn-arm/master/cpuburn-a53.S
gcc -o cpuburn-a53 cpuburn-a53.S
DNPNWO

User avatar
Rive
Posts: 586
Joined: Sat Mar 26, 2016 5:21 pm
Location: USA

Re: Cooling the Pi3 with a fan.

Sat Apr 23, 2016 1:11 pm

Gerd wrote:
Rive wrote:I give you 5 min tops, before you hit the ceiling, slow to a crawl, and revert to 600MHz (Hell, you probably couldn't even manage to do it with 'Stress' (which is a lightweight contender compared to cpuburn)).

C'mon, show me how wrong I am. I am waiting.
No vid, sreenshot must suffic.
cpuburn passive.png
No pic of the pi3? What, no web cam, camera, or cell phone? Hmmm. Also, you didn't do it for the 15 min or state the ambient (you did it according to top for 8 min), or use this to show all temps (records temp and freq every 10 sec for 15 min):

Code: Select all

while true; do vcgencmd measure_clock arm; vcgencmd measure_temp; sleep 10; done& ./cpuburn-a53
Don't cheat by say putting in a chilled room, where ambient is not room temp, or under a ceiling fan/other fan, etc (that is why a vid is better (even if you just make it with your phone, and then simply upload it to google drive (if you have gmail, you have google drive) and share link https://drive.google.com/drive/my-drive ), makes it harder to fake/skew the results)

Here is the complete quote BTW:
http://www.raspberrypi.org/forums/viewt ... 62#p948462
I'd like you see one of you with the pi3 stock @ 1200 MHz (or even overclocked at a higher arm freq ), without a case, no fan, and with just a heat sink show the temps and freq in a video while running the synthetic CPUBurn-a53 for a non-stop/continuous duration of 15 min without hitting 85°C.

I give you 5 min tops, before you hit the ceiling, slow to a crawl, and revert to 600MHz (Hell, you probably couldn't even manage to do it with 'Stress' (which is a lightweight contender compared to cpuburn)).

C'mon, show me how wrong I am. I am waiting.
DNPNWO

User avatar
Gerd
Posts: 66
Joined: Wed Mar 16, 2016 10:48 am
Location: Europe

Re: Cooling the Pi3 with a fan.

Sat Apr 23, 2016 2:27 pm

Rive wrote: I am not gonna hunt it down, put it here what you want me to install/run for linpack and I will give it ago.
I just followed viewtopic.php?p=926798#p926798 .

User avatar
Gerd
Posts: 66
Joined: Wed Mar 16, 2016 10:48 am
Location: Europe

Re: Cooling the Pi3 with a fan.

Sat Apr 23, 2016 2:39 pm

Rive wrote:No pic of the pi3?
I thought you would follow that thread.
viewtopic.php?p=957234#p957234
No cheating. Ambient 20°C.
And the cooler isn't mounted yet, just lying planar an the CPU. Glued and vertical it will show its full potential.

User avatar
Rive
Posts: 586
Joined: Sat Mar 26, 2016 5:21 pm
Location: USA

Re: Cooling the Pi3 with a fan.

Sat Apr 23, 2016 2:48 pm

Gerd wrote:
Rive wrote: I am not gonna hunt it down, put it here what you want me to install/run for linpack and I will give it ago.
I just followed viewtopic.php?p=926798#p926798 .

I see links and instructions for downloading, and compiling code....I don't have time for all that. I am not gonna jump through hoops for linpack (make it simple, do step by step instructions...like this example for setting up sdbench)

for example, sdbench:
open, and 'select all':
http://www.nmacleod.com/public/sdbench.sh

Copy the text, then open a terminal and type...

Code: Select all

sudo nano sdbench.sh

...and paste the text in, then save with Ctrl-X, Y, Enter.

Make it executable with...

Code: Select all

sudo chmod +x sdbench.sh
Then run:

Code: Select all

sudo ./sdbench.sh
DNPNWO

User avatar
Gerd
Posts: 66
Joined: Wed Mar 16, 2016 10:48 am
Location: Europe

Re: Cooling the Pi3 with a fan.

Sat Apr 23, 2016 3:30 pm

Rive wrote:
Gerd wrote:
Rive wrote: I am not gonna hunt it down, put it here what you want me to install/run for linpack and I will give it ago.
I just followed viewtopic.php?p=926798#p926798 .
I see links and instructions for compiling code....I don't have time for all that. I am not gonna jump through hoops for linpack.
No compiling, what is so hard to follow: "Alternately, you can install gfortran and libmpich-dev libs and then the following binary shoud work:
http://web.eece.maine.edu/~vweaver/junk/pi3_hpl.tar.gz " in the post i linked?"

OK, i prepare some fast food:

Code: Select all

sudo apt-get install libmpich-dev
gfortran should install also.

Code: Select all

cd ~

Code: Select all

wget http://web.eece.maine.edu/~vweaver/junk/pi3_hpl.tar.gz

Code: Select all

tar -xvzf pi3_hpl.tar.gz 

Code: Select all

chmod +x xhpl 

Code: Select all

./xhpl 
Later with

Code: Select all

nano HPL.dat
you can change the N from 8000.

Enjoy!

User avatar
Rive
Posts: 586
Joined: Sat Mar 26, 2016 5:21 pm
Location: USA

Re: Cooling the Pi3 with a fan.

Sat Apr 23, 2016 3:36 pm

Finished 1 tests with the following results:
0 tests completed and passed residual checks,
0 tests completed and failed residual checks,
1 tests skipped because of illegal input values.
--------------------------------------------------------------------------------

End of Tests..
HPL.dat

Code: Select all

HPLinpack benchmark input file
Innovative Computing Laboratory, University of Tennessee
HPL.out      output file name (if any)
6            device out (6=stdout,7=stderr,file)
1            # of problems sizes (N)
8000            Ns
1            # of NBs
256             NBs
0            PMAP process mapping (0=Row-,1=Column-major)
1            # of process grids (P x Q)
1         Ps
1         Qs
16.0         threshold
1            # of panel fact
0          PFACTs (0=left, 1=Crout, 2=Right)
1            # of recursive stopping criterium
2           NBMINs (>= 1)
1            # of panels in recursion
2            NDIVs
1            # of recursive panel fact.
2         RFACTs (0=left, 1=Crout, 2=Right)
1            # of broadcast
2            BCASTs (0=1rg,1=1rM,2=2rg,3=2rM,4=Lng,5=LnM)
1            # of lookahead depth
0            DEPTHs (>=0)
2            SWAP (0=bin-exch,1=long,2=mix)
64           swapping threshold
0            L1 in (0=transposed,1=no-transposed) form
0            U  in (0=transposed,1=no-transposed) form
1            Equilibration (0=no,1=yes)
8            memory alignment in double (> 0)

Code: Select all

[email protected]:~ $ ./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

HPL ERROR from process # 0, on line 173 of function HPL_pdtest:
>>> [0,0] Memory allocation failed for A, x and b. Skip. <<<

================================================================================

Finished      1 tests with the following results:
              0 tests completed and passed residual checks,
              0 tests completed and failed residual checks,
              1 tests skipped because of illegal input values.
--------------------------------------------------------------------------------

End of Tests.
DNPNWO

User avatar
Gerd
Posts: 66
Joined: Wed Mar 16, 2016 10:48 am
Location: Europe

Re: Cooling the Pi3 with a fan.

Sat Apr 23, 2016 3:57 pm

Rive wrote:something isnt right (do i need to manually set memory allocation?
/
Never had this, maybe because i run my Pi3 headless with gpu_mem=48. I guess you run gpu mem with 256. I would try lower it for this test to 64 in config.txt
Edit: or lower to N=6000 for the start.

User avatar
Rive
Posts: 586
Joined: Sat Mar 26, 2016 5:21 pm
Location: USA

Re: Cooling the Pi3 with a fan.

Sat Apr 23, 2016 4:21 pm

I set gpu mem to 64, and used 6000

I then disabled all overclocks, and ran with pi3 stock speeds and values ONLY for sdcard, core, sdram, voltages, and the arm. The test failed with the below parameter values.


Code: Select all

[email protected]:~ $ ./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      :    6000 
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        6000   256     1     1              24.34              5.919e+00
HPL_pdgesv() start time Sat Apr 23 12:18:55 2016

HPL_pdgesv() end time   Sat Apr 23 12:19:19 2016

--------------------------------------------------------------------------------
||Ax-b||_oo/(eps*(||A||_oo*||x||_oo+||b||_oo)*N)= 4802590957.1838751 ...... FAILED
||Ax-b||_oo  . . . . . . . . . . . . . . . . . =          29.178269
||A||_oo . . . . . . . . . . . . . . . . . . . =        1540.184305
||A||_1  . . . . . . . . . . . . . . . . . . . =        1543.558216
||x||_oo . . . . . . . . . . . . . . . . . . . =           5.921421
||x||_1  . . . . . . . . . . . . . . . . . . . =        5989.406088
||b||_oo . . . . . . . . . . . . . . . . . . . =           0.499944
================================================================================

Finished      1 tests with the following results:
              0 tests completed and passed residual checks,
              1 tests completed and failed residual checks,
              0 tests skipped because of illegal input values.
--------------------------------------------------------------------------------

End of Tests.
DNPNWO

User avatar
Rive
Posts: 586
Joined: Sat Mar 26, 2016 5:21 pm
Location: USA

Re: Cooling the Pi3 with a fan.

Sat Apr 23, 2016 6:15 pm

passed @ 6000 with gpu_mem=64 for following overclock settings:

Code: Select all

dtparam=sd_overclock=100
arm_freq=1275
core_freq=500
over_voltage=4
sdram_freq=575
sdram_schmoo=0x02000020
over_voltage_sdram_p=6
over_voltage_sdram_i=4
over_voltage_sdram_c=4
v3d_freq=500
h264_freq=333

Code: Select all

[email protected]:~ $ ./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      :    6000 
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        6000   256     1     1              21.99              6.552e+00
HPL_pdgesv() start time Sat Apr 23 14:11:12 2016

HPL_pdgesv() end time   Sat Apr 23 14:11:34 2016

--------------------------------------------------------------------------------
||Ax-b||_oo/(eps*(||A||_oo*||x||_oo+||b||_oo)*N)=        0.0028294 ...... 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.
=================
passed @ 8000 with gpu_mem=64 for following overclock settings:

Code: Select all

dtparam=sd_overclock=100
arm_freq=1250
core_freq=500
over_voltage=4
sdram_freq=575
sdram_schmoo=0x02000020
over_voltage_sdram_p=6
over_voltage_sdram_i=4
over_voltage_sdram_c=4
v3d_freq=500
h264_freq=333

Code: Select all

[email protected]:~ $ ./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              51.15              6.675e+00
HPL_pdgesv() start time Sat Apr 23 14:23:43 2016

HPL_pdgesv() end time   Sat Apr 23 14:24:34 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.
passed @ 8000 with gpu_mem=64 for following overclock settings:

Code: Select all

dtparam=sd_overclock=100
arm_freq=1260
core_freq=500
over_voltage=4
sdram_freq=575
sdram_schmoo=0x02000020
over_voltage_sdram_p=6
over_voltage_sdram_i=4
over_voltage_sdram_c=4
v3d_freq=500
h264_freq=333

Code: Select all

[email protected]:~ $ ./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              50.63              6.743e+00
HPL_pdgesv() start time Sat Apr 23 14:38:56 2016

HPL_pdgesv() end time   Sat Apr 23 14:39:47 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.
Last edited by Rive on Sat Apr 23, 2016 7:03 pm, edited 2 times in total.
DNPNWO

User avatar
Gerd
Posts: 66
Joined: Wed Mar 16, 2016 10:48 am
Location: Europe

Re: Cooling the Pi3 with a fan.

Sat Apr 23, 2016 6:40 pm

As I said before, you had to lower clock from your posted settings. Why not put it the other way round. NO oc at all, NO cpu , NO ram, just NO; only over_voltage. And then lets see when you get the first correct pass.
Edit: And remember: I did N=10000 with passive cooling viewtopic.php?p=959556#p959556
Last edited by Gerd on Sat Apr 23, 2016 7:10 pm, edited 1 time in total.

User avatar
Rive
Posts: 586
Joined: Sat Mar 26, 2016 5:21 pm
Location: USA

Re: Cooling the Pi3 with a fan.

Sat Apr 23, 2016 6:43 pm

Gerd wrote:As I said before, you had to lower clock from your posted settings. Why not put it the other way round. NO oc at all, NO cpu , NO ram, just NO; only over_voltage. And then lets see when you get the first correct pass.
Ok. So, you mean run stock, BUT with just the over voltage?

Also, third 8000 passed @ 1260 MHz
DNPNWO

User avatar
Rive
Posts: 586
Joined: Sat Mar 26, 2016 5:21 pm
Location: USA

Re: Cooling the Pi3 with a fan.

Sat Apr 23, 2016 6:52 pm

Why not put it the other way round. NO oc at all, NO cpu , NO ram, just NO; only over_voltage. And then lets see when you get the first correct pass.
passed at 8000 stock overvolt/overvolt sdram only as shown (Abient 20C, linpack @55C)

Code: Select all

over_voltage=4
over_voltage_sdram_p=6
over_voltage_sdram_i=4
over_voltage_sdram_c=4

Code: Select all

[email protected]:~ $ ./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              54.70              6.241e+00
HPL_pdgesv() start time Sat Apr 23 14:49:46 2016

HPL_pdgesv() end time   Sat Apr 23 14:50:41 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.

passed at 8000 stock with over_voltage only

Code: Select all

over_voltage=4

Code: Select all

[email protected]:~ $ ./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              54.59              6.255e+00
HPL_pdgesv() start time Sat Apr 23 14:55:29 2016

HPL_pdgesv() end time   Sat Apr 23 14:56:23 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.
Last edited by Rive on Sat Apr 23, 2016 7:12 pm, edited 2 times in total.
DNPNWO

User avatar
Gerd
Posts: 66
Joined: Wed Mar 16, 2016 10:48 am
Location: Europe

Re: Cooling the Pi3 with a fan.

Sat Apr 23, 2016 7:04 pm

Yes, but i meant NO oc, not even ram. What I expect you will find, is that you need over_voltage = 2 or 1 to run it at stock (=NO oc) settings. This means, that you have to overvolt your pi to get the correct answer for a linear equation. This should not hapen.

User avatar
Rive
Posts: 586
Joined: Sat Mar 26, 2016 5:21 pm
Location: USA

Re: Cooling the Pi3 with a fan.

Sat Apr 23, 2016 7:05 pm

Gerd wrote:Yes, but i meant NO oc, not even ram. What I expect you will find, is that you need over_voltage = 2 or 1 to run it at stock (=NO oc) settings. This means, that you have to overvolt your pi to get the correct answer for a linear equation. This should not hapen.
passed at 8000 stock with over_voltage=4 only

Do you want me to retest with over-voltage at 2? I have already verified it fails at stock with no over_voltage at all.

Code: Select all

over_voltage=4

Code: Select all

[email protected]:~ $ ./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              54.59              6.255e+00
HPL_pdgesv() start time Sat Apr 23 14:55:29 2016

HPL_pdgesv() end time   Sat Apr 23 14:56:23 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.
DNPNWO

User avatar
Rive
Posts: 586
Joined: Sat Mar 26, 2016 5:21 pm
Location: USA

Re: Cooling the Pi3 with a fan.

Sat Apr 23, 2016 7:16 pm

passes at 8000 stock with over voltage =1

Code: Select all

over_voltage=1

Code: Select all

[email protected]:~ $ ./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              55.37              6.166e+00
HPL_pdgesv() start time Sat Apr 23 15:14:17 2016

HPL_pdgesv() end time   Sat Apr 23 15:15:12 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.
DNPNWO

User avatar
Rive
Posts: 586
Joined: Sat Mar 26, 2016 5:21 pm
Location: USA

Re: Cooling the Pi3 with a fan.

Sat Apr 23, 2016 8:09 pm

Instead of hijacking this thread for linpack bench, I have started a new overclocking linpack/cpuburn thread here:
viewtopic.php?p=959884#p959884
DNPNWO

User avatar
Gerd
Posts: 66
Joined: Wed Mar 16, 2016 10:48 am
Location: Europe

Re: Cooling the Pi3 with a fan.

Sat Apr 23, 2016 8:32 pm

Rive wrote:Instead of hijacking this thread for linpack bench, I have started a new overclocking linpack/cpuburn thread here:
viewtopic.php?p=959884#p959884
Instead of opening annother thread you could as well have posted your results in the referenced linpack thread viewtopic.php?f=63&t=139712

Return to “General discussion”