Recommended heatsink for panda?

[Karel Gardas]

Hello,
I'm using Pandaboard as a softwrae development platform. For me it
enables make -j2 on hours of compilation of Haskell code.
Unfortunately build sometimes fails with strange error, which does not
show again if restarted. I suspect CPU/RAM over-heating and so would
like to ask people here how they solve this issue themselves? I know
just http://therandomlab.blogspot.com/2011/03/pandaboard-is-hot.html
but this heatsink is not sufficient. I quite like something like this:
http://shop.tntrade.cz/primecooler-pc-nb1-northbridge-heatsink_d57600.html
-- but it looks the base of this is too big to get between SD card
slot and LS Research chip.
Any idea what to use for better cooling CPU and RAM?
Thanks!
Karel

[DJ Delorie]

I've got one of these on my pandaboard (it just sticks on):

http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=ATS1315-ND

Unfortunately, it didn't solve the particular problem I was seeing
(random hangs during heavy compiles) despite the cpu being cooler (it
seems, based on bottom-of-board temp).

I also got one of these but fried it before I could put it in place:

http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=563-1108-ND

(stupid smart fans ;)

[Karel Gardas]

This heatsink looks nice. Fan quite dangerously -- 8k RPM! :-)
Well, do you use this temp monitoring program? http://hardwarebug.org/files/omap4_temp.c
-- do you think it provides temperatures in Celsius or Fahrenheit? If
in C, then my board is in real danger I guess while compiling for a
few minutes where it goes up to 91.

How exactly helped your heatsink? i.e. what was the temperature delta
after installing it?

Thanks!
Karel

On Jun 24, 9:01 pm, DJ Delorie <d...@redhat.com> wrote:
> I've got one of these on my pandaboard (it just sticks on):
>

> http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=ATS1...

>
> Unfortunately, it didn't solve the particular problem I was seeing
> (random hangs during heavy compiles) despite the cpu being cooler (it
> seems, based on bottom-of-board temp).
>
> I also got one of these but fried it before I could put it in place:
>

> http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=563-...
>
> (stupid smart fans ;)

[DJ Delorie]

I didn't use a temperature monitoring program, I just stuck my finger
on it.

[Måns Rullgård]

Karel Gardas <karel....@gmail.com> writes:

> This heatsink looks nice. Fan quite dangerously -- 8k RPM! :-) Well,
> do you use this temp monitoring program?
> http://hardwarebug.org/files/omap4_temp.c -- do you think it provides
> temperatures in Celsius or Fahrenheit? If in C, then my board is in
> real danger I guess while compiling for a few minutes where it goes up
> to 91.

Those temperatures are Celsius. 91 is a bit high, but still within a
reasonable range. TJmax is typically over 100 C for devices like this,
although I can't find a figure for the OMAP4430.

--
Måns Rullgård
ma...@mansr.com

[Rob Clark]

btw, which kernel, and what bootargs?

This sounds a bit similar to highmem bugs that were seen on 2.6.35 kernel

BR,
-R

[Karel Gardas]

$ cat /proc/cmdline
ro elevator=noop vram=32M mem=456M@0x80000000 mem=512M@0xA0000000
root=UUID=41f00c7f-ab45-4243-8559-5d4e032878ea fixrtc quiet splash
console=ttyO2,115200n8

$ uname -a
Linux panda 2.6.38-1208-omap4 #11-Ubuntu SMP PREEMPT Fri Apr 15
16:34:35 UTC 2011 armv7l armv7l armv7l GNU/Linux

This is standard distro kernel from Ubuntu 11.04 headless
installation.

Thanks!
Karel

On Jun 24, 10:27 pm, Rob Clark <robdcl...@gmail.com> wrote:

> On Fri, Jun 24, 2011 at 11:25 AM, Karel Gardas <karel.gar...@gmail.com> wrote:
> > Hello,
> > I'm using Pandaboard as a softwrae development platform. For me it
> > enables make -j2 on hours of compilation of Haskell code.
> > Unfortunately build sometimes fails with strange error, which does not
> > show again if restarted. I suspect CPU/RAM over-heating and so would
> > like to ask people here how they solve this issue themselves? I know

> > justhttp://therandomlab.blogspot.com/2011/03/pandaboard-is-hot.html

> > but this heatsink is not sufficient. I quite like something like this:

> >http://shop.tntrade.cz/primecooler-pc-nb1-northbridge-heatsink_d57600...

[Måns Rullgård]

Apparently my temp reading tool for OMAP4 sometimes hangs. Any ideas
why this might be happening?

Karel Gardas <karel....@gmail.com> writes:

> 2011/6/27 Måns Rullgård <ma...@mansr.com>:
>> Karel Gardas <karel....@gmail.com> writes:
>>
>>> Hello,
>>>
>>> am I right you are the author of this excellent temp monitoring tool
>>> for Pandaboard? If so, then I've observed one issue with it. If you
>>> keep it running in a loop like:
>>>
>>> while(true)
>>> do
>>>   sudo ./omap4_temp
>>>   sleep 5
>>> done
>>>
>>> for some time, it'll stop working and in fact start consume 100% CPU.
>>> When this happen there is no way to recover temp reading again.
>>> Perhaps reboot will be able to help, but I've not rebooted my board
>>> yet.
>>>
>>> Have you seen this behavior already?
>>
>> No, I haven't seen that happen.  Could you try to find out where it is
>> hanging?  There are two loops that spin until a particular bit in a
>> register is set or cleared, and it seems like this isn't happening for
>> some reason.  Attaching a debugger should tell where it is stuck.
>>
> It's the first cycle, line 90:
>
> 88
> 89 *temp = 0x200;
> 90 while (!(*temp & 0x100)) __asm__ ("" ::: "memory");
> 91 *temp = 0;
> 92 while (*temp & 0x100) __asm__ ("" ::: "memory");
> 93 tempval = *temp;
> 94

--
Måns Rullgård
ma...@mansr.com

[Vladimir Pantelic]

M�ns Rullg�rd wrote:
> Apparently my temp reading tool for OMAP4 sometimes hangs. Any ideas
> why this might be happening?

overheated?

>
> Karel Gardas<karel....@gmail.com> writes:
>
>> 2011/6/27 M�ns Rullg�rd<ma...@mansr.com>:

>>> Karel Gardas<karel....@gmail.com> writes:
>>>
>>>> Hello,
>>>>
>>>> am I right you are the author of this excellent temp monitoring tool
>>>> for Pandaboard? If so, then I've observed one issue with it. If you
>>>> keep it running in a loop like:
>>>>
>>>> while(true)
>>>> do
>>>> sudo ./omap4_temp
>>>> sleep 5
>>>> done
>>>>
>>>> for some time, it'll stop working and in fact start consume 100% CPU.
>>>> When this happen there is no way to recover temp reading again.
>>>> Perhaps reboot will be able to help, but I've not rebooted my board
>>>> yet.
>>>>
>>>> Have you seen this behavior already?
>>>
>>> No, I haven't seen that happen. Could you try to find out where it is
>>> hanging? There are two loops that spin until a particular bit in a
>>> register is set or cleared, and it seems like this isn't happening for
>>> some reason. Attaching a debugger should tell where it is stuck.
>>>
>> It's the first cycle, line 90:
>>
>> 88
>> 89 *temp = 0x200;

>> 90 while (!(*temp& 0x100)) __asm__ ("" ::: "memory");
>> 91 *temp = 0;
>> 92 while (*temp& 0x100) __asm__ ("" ::: "memory");

[Karel Gardas]

Not at all! In fact when this happened temp was quite low (by panda
standard). And from this time it does not depend on the temperature,
it simply does not work but consume 100% CPU.

Karel

On Jun 27, 1:34 pm, Vladimir Pantelic <vlado...@gmail.com> wrote:
> M ns Rullg rd wrote:
> > Apparently my temp reading tool for OMAP4 sometimes hangs.  Any ideas
> > why this might be happening?
>
> overheated?
>
>
>

> > Karel Gardas<karel.gar...@gmail.com>  writes:
>
> >>  2011/6/27 M ns Rullg rd<m...@mansr.com>:
> >>>  Karel Gardas<karel.gar...@gmail.com>  writes:

[Karel Gardas]

I've just rebooted my panda and temp reading tool is working again.

I have also purchased primecooler pc-nb1 northbirdge cooler for
cooling CPU. For now I've just put it down on the CPU w/o any thermal
grease nor thermal stick and it cools nicely. It fits really tightly
inbetween SD card slot and WiFi chip. It also lies not only on CPU,
but also on small soldered piece with metal (gold) surface. I don't
what's this, but it's near to the CPU. For now I've just lay it down
for test if Panda works well with it. Later I'll probably use thermal
stick to fasten it to the CPU.

Karel

[Vladimir Pantelic]

Karel Gardas wrote:
> I've just rebooted my panda and temp reading tool is working again.
>
> I have also purchased primecooler pc-nb1 northbirdge cooler for
> cooling CPU. For now I've just put it down on the CPU w/o any thermal
> grease nor thermal stick and it cools nicely. It fits really tightly
> inbetween SD card slot and WiFi chip. It also lies not only on CPU,
> but also on small soldered piece with metal (gold) surface. I don't
> what's this, but it's near to the CPU.

it's a crystal oscillator.

[dasankir]

Hi,

I've just posted this:

http://therandomlab.blogspot.com/2011/06/heatsink-performance-analysis-on.html

On Jun 24, 12:25 pm, Karel Gardas <karel.gar...@gmail.com> wrote:
> Hello,
> I'm using Pandaboard as a softwrae development platform. For me it
> enables make -j2 on hours of compilation of Haskell code.
> Unfortunately build sometimes fails with strange error, which does not
> show again if restarted. I suspect CPU/RAM over-heating and so would
> like to ask people here how they solve this issue themselves? I know

> justhttp://therandomlab.blogspot.com/2011/03/pandaboard-is-hot.html
> but this heatsink is not sufficient. I quite like something like this:http://shop.tntrade.cz/primecooler-pc-nb1-northbridge-heatsink_d57600...

[Karel Gardas]

Nice! In fact I've been inspired by your postings and thermal analysis
of panda quite a lot. Anyway, my data are:

100% CPU load (this means make -j2 here): ~91C w/o heatsink while ~51
with heatsink -- I'm using primecooler pc-nb1, which is 37x37x46mm and
it just lies on CPU, no thermal grease nor stick yet.

I guess your 100% load means just 1 core loaded, right?

Karel

On Jun 29, 5:28 pm, dasankir <i...@dasaki.com> wrote:
> Hi,
>
> I've just posted this:
>

> http://therandomlab.blogspot.com/2011/06/heatsink-performance-analysi...

[Karel Gardas]

Hi,

I've also put some of the pictures of my panda with PC-NB1 on top of
it into this blogpost: http://ghcarm.wordpress.com/2011/07/03/new-hardware-pandaboard/

Karel

On Jun 29, 5:28 pm, dasankir <i...@dasaki.com> wrote:

> Hi,
>
> I've just posted this:
>

> http://therandomlab.blogspot.com/2011/06/heatsink-performance-analysi...

[Benjamin Rose]

Would you recommend this heatsink? Have you found any issues with wifi attenuation?

[Karel Gardas]

Hi Benjamin,

I would certainly recommend this heatsink although it might look as a
overloaded for OMAP4. I'm sorry, but I'm not using WiFi here so I
cannot check.

Karel

On Jul 4, 2:19 am, Benjamin Rose <br...@webergyltd.com> wrote:
> Would you recommend this heatsink? Have you found any issues with wifi attenuation?
>

[Bryan Smith]

Hey Karel,

I checked out your site can you clarify something for me? You said that the Pandaboard's SD card/ MicroSD card Stability is "very weak and very sensitive to card in use" 

Are you basing this on the fact that the Ubuntu installer fails to resize certain SD's or the test you preformed here:

http://groups.google.com/group/pandaboard/browse_thread/thread/88894079fae87701#

Have you tried this with Angstrom or another distro besides Ubuntu to rule out it being a software issue? Any new or old SD/MicroSD I get my hands on boots and runs my Panda.

Bryan

--
A healthy diet  includes Linux, Linux, and more Linux.

[Karel Gardas]

Hi Bryan,

indeed, I made this claim based on my own test when both Solaris'
based thinkpad and Ubuntu 10.04.2 based i.MX53 are capable to read and
check SD card and just Panda does have problems with it. Anyway, in
the description below the table I admit that "but some people also
think that it might be software issue." -- see Frantisek's link to the
patch. I would be very glad if you can smash this my hardware argument
and proof that it's actually a kernel bug. Just use some of "non-
compatible" card which re reported on this board and use it together
with your Angstorm and write your experience here.

Thanks!
Karel
PS: on the other hand, for software developer like me if something is
hardware or kernel issue is quite irrelevant at the end since system
simply doesn't work like it should... see, i.MX53 rock stable with
linux freescale recommends to install. Panda is not stable with linux
its recommended to install. I hope you get this my point...

On Jul 4, 8:26 pm, Bryan Smith <bryansmit...@gmail.com> wrote:
> Hey Karel,
>
> I checked out your site can you clarify something for me? You said that the
> Pandaboard's SD card/ MicroSD card Stability is "very weak and very
> sensitive to card in use"
>
> Are you basing this on the fact that the Ubuntu installer fails to resize

> certain SD's or the test you preformed here:http://groups.google.com/group/pandaboard/browse_thread/thread/888940...

>
> Have you tried this with Angstrom or another distro besides Ubuntu to rule
> out it being a software issue? Any new or old SD/MicroSD I get my hands on
> boots and runs my Panda.
>
> Bryan
>

[alxx]

I'm using the same ram/video type heatsink I used on omap3 (on gumstix
overo omap3530)

They are about $3 each from jaycar (they have a US web site)
http://www.jaycar.com.au/productView.asp?ID=HH8580&keywords=heatsink&form=KEYWORD

I use a small 12V fan powered from a usb hub to keep a constant
airflow over the heat sink.


Alex

On Jun 25, 2:25 am, Karel Gardas <karel.gar...@gmail.com> wrote:
> Hello,
> I'm using Pandaboard as a softwrae development platform. For me it
> enables make -j2 on hours of compilation of Haskell code.
> Unfortunately build sometimes fails with strange error, which does not
> show again if restarted. I suspect CPU/RAM over-heating and so would
> like to ask people here how they solve this issue themselves? I know

> justhttp://therandomlab.blogspot.com/2011/03/pandaboard-is-hot.html
> but this heatsink is not sufficient. I quite like something like this:http://shop.tntrade.cz/primecooler-pc-nb1-northbridge-heatsink_d57600...

[dasankir]

these are my compiler options:

-fexceptions -march=armv7-a -mtune=cortex-a9 -mfpu=neon -mfloat-
abi=softfp

I read the load from the all'-CPU line of /proc/stat. I try to match
the desired load by enabling / disabling threads and by changing the
length of the loop inside of the threads: for (int i = 0; i < loops; i+
+) { sqrt (rand ());}

[dasankir]

In my opinion the PC-NB1 is a bit overshoot.

This will keep temp below 60 C (at room temp 25 C):

http://www.lightobject.com/3cmx3cm-heatsink-P163.aspx

And this below 55 C (at room temp 25 C):

http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=ATS1307-ND

[Karel Gardas]

Sure, PC-NB1 is overshoot, but it does have few advantages (1) was
available next day in the local shop behind the corner and (2) was
cheap in comparison with getting heatsink from DigiKey.

So I went this route and I'm happy with the choice...

Karel

On Jul 5, 3:59 pm, dasankir <i...@dasaki.com> wrote:
> In my opinion the PC-NB1 is a bit overshoot.
>
> This will keep temp below 60 C (at room temp 25 C):
>
> http://www.lightobject.com/3cmx3cm-heatsink-P163.aspx
>
> And this below 55 C (at room temp 25 C):
>

> http://search.digikey.com/scripts/DkSearch/dksus.dll?Detail&name=ATS1...

[dasankir]

@Karel

It wasn't my intention to argue. Another advantage of the PC-NB1 is
that it will ensure the best possible amount of cooling (passive). I
just personally don't like the size.

[cody.fr...@gmail.com]

Sorry for the late reply. The max temp I've reached (using omap4_temp.c and the command: stress --cpu 2 --io 1 --vm 1 --vm-bytes 128M --timeout 600s) is 54C in a room which is 27C. Idle (0.25 Load Average) temperature is 46C. The Sink i'm using now looks like this: http://i.imgur.com/WDOr5.jpg and came off an old graphics card. I had to trim the sides to make it fit, but it works great. The board is passively cooled and the heatsink is stuck to the processor using this tape here: http://www.sparkfun.com/products/9771. I imagine If I got a fan it would be much cooler, but since 54C is more than acceptable I haven't worried about it. Somewhat concerned about the temp of the ram, but I've seen no problems and haven't worried about it since I'm not sure if there even is a sensor to check the ram temp or if I'd have to do it manually. Hope this helps!

[Karel Gardas]

Hello Code,

have you cut the tape in order to cover just the CPU? Or have you just
stuck to heat-sink whole (5x5cm) and then just stuck to CPU?

Thanks,
Karel

On Jul 16, 11:46 am, cody.frolan...@gmail.com wrote:
> Sorry for the late reply. The max temp I've reached (using omap4_temp.c and
> the command: stress --cpu 2 --io 1 --vm 1 --vm-bytes 128M --timeout 600s) is
> 54C in a room which is 27C. Idle (0.25 Load Average) temperature is 46C. The

> Sink i'm using now looks like this:http://i.imgur.com/WDOr5.jpgand came

[cody.fr...@gmail.com]

I cut the tape and stuck it to the CPU, though you could do it either way. I just wanted to save the extra thermal tape for other projects!