Pandaboard Support

@Rick
Ambient temperature is about 20C and there is no cooling. It is in a metal case, called Tigal (might still be on the CJE website where I bought it).
So, according to TI, Panda doesn’t ‘have a temperature’.

Just tried Raik’s PB_CPUdetect and it works.

Sorry is Kuemmel’s not mine!

I would say – completely off the top of my head and without an iota of anything resembling “proof” (this is the internet, after all) – that if your chip is a package-on-package (the RAM is on top of the CPU), that straying over 85C is “too much”. Not for the health of the processor, but more because things change shape slightly at different temperatures and a PoP design welds together two solid things at different temperatures. [this is why adding a heatsink is much less effective than forced air cooling (ie a fan) – a heatsink on top will be cooling the RAM, not the SoC]

Last year, after lots of tests, the Pi foundation decided that the ~700MHz Pi could be overclocked up towards 1GHz safely so long as two conditions were met:

1. It ramped up the speed when the device was busy (and not as a matter of course)
2. It backs down the moment the CPU temperature hits 85C.

More on that here: http://www.raspberrypi.org/archives/2008

Does your box have adequate ventilation and air space around the board? Are there any slots for air to pass through?
[update – looks like it ought to be okay – http://cjemicros.co.uk/micros/individual/newprodpages/prodinfo.php?prodcode=PAN-PAN-ES-TG ]

So they seem to think 85C is when it is getting too much. I’d prefer 75C just to be paranoid.
I would worry less about 55-65C, and below that – well – that’s barely even tickover, am I wrong? ;-)

BTW, that link above is worth reading just for the comment by the guy who misinterprets the overvolt value of ‘6’ (means ramping up the CPU power from 1.2V to 1.35V) as “need another 6 volts, where can I find an 11V power supply for my Pi?”. Eeeeeek! :-) :-)

The test is, of course, to fire up something computationally intensive and see how that affects the temperature. If it doesn’t rise much, stop worrying. ;-)

Now while you might expect a rise – playing a 480P (SD!) H.264 video on my eeePC causes a jump from ~55C to ~68C (well, it is an Intel chipset!); yesterday I watched an episode of Red Data Girl (720P (HD!) H.264+subs) on the Pi with RaspXBMC and according to the on-screen temperature report (overlaid on top of the video), the temperature rose from ~48C to 51C. You gotta love the Pi to do it better and not even break into a sweat… ;-)

How will your Panda fare? [though your post implies that something is keeping the CPU in fast mode as a matter of course?]

As at June 4th 2013:

There are also at least 2 changes that haven’t made it into CVS yet. The Smartreflex driver (CPU runs much cooler) and the CJE Power control module support. (This solves the problem of shutdown/restart not working).

I have tweaked !CPUTmpMon to show both processor speed and temperature (by adding the following lines to !RunImage) which will be useful once the SmartReflex driver makes it into CVS as I’ll be able to see the processor speed up and slow down as the SmartReflex driver operates:

!RunImage    WR/     BASIC     16:01:57 18-Feb-2012   35 kbytes

    1REM >!RunImage
    2REM For !CPUTmpMon
    3REM Version 2.01 (18.02.2012)
    4REM (c) WPB, 2010
    5REM OMAP4 version by Willi Theiß

   52freq%=10000
  141DIM iconblk% 48+16
  152iconblk%!12 = 196
  229REM We need to read the temperature, so poll normally until that's complete.
     B%=_std_fuse_prod_id_1%
     C%=0
     silicon_type%= (USR(ReadReg) >> 16) AND %11
     B%=_mpu_cm%
     C%=_cm_clksel_dpll_mpu%
     dpll_mult%=(USR(ReadReg) >> 8) AND &7FF
     B%=_mpu_cm%
     C%=_cm_clksel_dpll_mpu%
     dpll_div%=USR(ReadReg) AND &3F
     B%=_mpu_cm%
     C%=_cm_div_m2_dpll_mpu%
     dpll_clkout_m2%=USR(ReadReg) AND &1F
     freq%=38.4*2*dpll_mult%/(dpll_div%+1)/(2*dpll_clkout_m2%)
  230PROCtemp_start_conv
  270DEF FNtemp_to_string(temp%)
  271="f "+STR$(freq%)+"MHz "+STR$(temp% DIV 100) + "." + LEFT$(STR$(ABS(temp% MOD 100)), 1) + CHR$(176) + "C"
  323_control_idcode_reg% = &4A002204
     _std_fuse_prod_id_1% = &4A002218
     _mpu_cm%             = &4A004000
     _cm_clksel_dpll_mpu% = &16C
     _cm_div_m2_dpll_mpu% = &170
  325B% = _control_idcode_reg%
     C%=0
  326REM Hawkeye Number is 16-bit number in bits 12:27...
  354.ReadReg
      MOV R5,R2
  355SWI "OS_EnterOS"            ; SVC mode
  358SWI "OS_Memory"             ; R2 -> logical address, R3 -> reference key
  359ADD R2,R2,R5

I have also renamed my ‘MLO’ and used the ‘MLO1500’ instead. My Pandaboard ES is now running at 1500MHz and about 56°C. Otherwise I am running the RComp 32Gbyte dual-partition PandaLand SD card with RISC OS 5.19 (30-Sep-2012) which is the latest supported image (AIUI).

I have tried RISC OS 5.21 (03-Sep-2012) on my Pandaboard. The SmartReflex driver seems to switch between 350MHz and 1200MHz as you do more or less intensive tasks. I am using the MLO1500 (rather than the MLO1200) but the SmartReflex driver overrides this and seems to have 1200MHz as the fastest speed. However the processor runs much cooler.

The new !Boot structure has confused !NewIconsX (the sprite name for the drive icon can live at FCxxxxxx and a $wimp%!32="Sname"+CHR$0 therefore fails with an abort, so I commented out that line of BASIC in the RunImage).

Moving from 5.19 (30-Sep-2012) [the version from PandaLand] to 5.21 (03-Sep-2013) causes the sound to be much worse – rather than just getting the timing wrong, the pitch is now worse. It would be nice to get sound sorted out. I am using !SignalBox to test the sound (which uses sound scheduling to add future sounds).

I’ve found the SmartReflex driver causes my Pandaboard to crash, not immediately but it’s not stable enough to use :-( . Mine seems to crash also if I try and use non smart reflex with the MLO1500.

I used MLO1500 extensively with no crash. Also I have not had a crash with SmartReflex.

My Pandaboard has an RTC card fitted with 3V button battery; but, if the time is set manually in ‘Time and Date’ configure, the time is not correct when Panda is next powered up. For one SD card running RISC OS 5.19 of 9/2/13, the time shown after boot up is the time at the last shutdown, while for another SD card running RISC OS 5.21 of 22/9/13, the time shown after boot up is for 1/1/00. :-((
I have tested the battery with my digital multimeter and it shows 2.97V
Has anyone got the RTC working on a Pandaboard or knows how to get it working?

For a RTC card to work something needs to be changed https://www.riscosopen.org/wiki/documentation/show/HALDeviceSysPeri_RTC

Thanks for this. Following the link at the bottom of that page,
https://www.riscosopen.org/wiki/documentation/show/HAL%20TODO#nonsensical ,
it seems that there is some hope that this might be corrected soon.

Finally got RISC OS 5.21 running on my Panda.
Broke a 8 GB SDCard on the way which had a backup.. So now I need to install from scratch. (not funny)
USB Flash drives are more reliable when used with RISC OS it seems.

First impressions:
- getting a working setup with MLO, uboot etc. seems quite hairy.
- disk speed is still not up to par, there surely could be some speed bumps in the stack somewhere waiting to be fixed.
- networkcard is not recognized by Network Configuration.
- SDcards formatted on Linux (FAT32) are not recognized. (formatting in a camera helps)
- partitions on SDCards are not recognized.
- fomatting filecore (!HForm) on SDcard can kill SDCards.

On the bright side:
- 1 GHz ARM running RISC OS 5.21 :-)
- 1 GB RAM (max 128MB RAMDisk?? 512MB would be lovely when compiling RISC OS!)
- my Beagleboard still runs strong so got a lot of stuff on that drive :-)
- 1920×1080 16M on DVI works :-)
- 35MB of 8GB used on harddisk .. so time to get that filled up.

Hmm.. the dreaded ‘ambigious name’ has popped up when using a USBstick. apparently i need to reboot before i can use it again.
I did not expect this to be still a problem, disconnecting a disk should surely remove the reference. I know that for old-time sake floppies could be swapped etc.
Rebooting to use a memorystick? Ah well..

Hmm.. the dreaded ‘ambigious name’ has popped up when using a USBstick. apparently i need to reboot before i can use it again.
I did not expect this to be still a problem, disconnecting a disk should surely remove the reference. I know that for old-time sake floppies could be swapped etc.
Rebooting to use a memorystick? Ah well..Is there a guide to properly format a SDcard for use with SDFS. combining a Filecore partition and a fat32 for booting?

Is there a guide to properly format a SDcard for use with SDFS. combining a Filecore partition and a fat32 for booting?

Two options – download the Raspberry Pi RC11 SD card image and tweak the FAT partition contents to include MLO, riscos (OMAP4 version), etc. This gives 2Gbytes available to RISC OS. Otherwise buy a 32Gbyte Pandaland SD card from R-Comp (ready to go, 32Gbyte available to RISC OS, fast SD card).

SDcard formatting:
- format the card with !HForm using 20 cilinders less than defined. ( in this case: 255 heads 63 sectors/track 993-20=973 cilinders)
- add a 128MB partition at the end of the disk with gparted (Linux)
- format the 128MB partition with FAT32.
- copy MLO, uboot, riscos, cmos and uenv.txt to fat32 part

The card i use is Sandisk 8GB Extreme III and does 20MB/s read 5.5MB/s write :-)

just installed !NetSurf so i am typing this from the Panda.

To get the network adapter recognized I needed to modify Autosense/EtherUSB into a Obey file with the string:

Set InetSetup$Driver$USB EtherUSB:ej0:EtherUSB:0.26:EtherUSB

And then configure as usual. :-)

!MoonFish and !SunFish are not doing as well as !smbserver for file sharing.
the latter will do up to 2.6 MB/s over 100 Mbit Full duplex.
it should be possible to squeeze a bit more performance out of that. So will look into recompiling !smbserver.
Hopefully the source code will give some clues on how to improve !MoonFish ans !SunFish.

Oh. Aemulor on the Panda! :-) Sweeeet!!
Apart from it killing the mode selection icon and generating an error it seems to work ok. (not everything runs but !Artworks 1.7f and my all time favorite !ProjMan runs.)
Can’t register on the Spellings Forum, seems it is broken. The forum has a feature request since april this year but I doubt anyone can login or add a comment..

The Pandaboard ROMs have disappeared from https://www.riscosopen.org/content/downloads/other-zipfiles. Where can I find the most recent one?

I got smartreflex working fine now, it seems I’d borked my !Boot with duplicate files.
All stable and smooth now for awhile :-D

The Pandaboard ROMs have disappeared

Yes, We had a query relating to licensing and decided the best option was to temporarily take down the OMAP4 build. Sorry. :( It’s pretty trivial so we hope normal service will be restored in the next few days (although attending the London Show means we can’t sort it out before next week).

Is https://www.riscosopen.org/viewer/revisions/logs?ident=1383093681-831331.html the licensing problem fix? Can normal service be restored?

I suspect that the changes were too late for last nights autobuild but will appear after tonight’s build.

Hopefully the OMAP4 builds will be back again soon.

@Rob: do you need a more recent ROM? i have built one from sources, 26 oct.

@Jan: Yes please! I don’t have a working RISC OS machine setup, so I can’t build my own at the moment. Any OMAP4 ROM would do, so I can get my Pandaboard ES setup again. My Email address is: rob at robheaton.co.uk

Cheers!

@Rob. done.