Not yet. The pinout should be compatible with the Bela cape. When we saw the datasheet for the chip about a year ago it was meant to have 4x PRUs on it, but they have now removed them, so the port of our current infrastructure would be pretty complicated. Overall this looks like a fairly expensive and very power-hungry device (notice the massive heatsink) which is more meant as a prototyping board for vision/AI class applications and not too well suited for being embedded into musical instruments ...

The bumph says: 2x 6-core Programmable Real-Time Unit and Industrial Communication SubSystem (PRU-ICSSG)

I was thinking of getting one to heat the office.

    AndyCap The bumph says: 2x 6-core Programmable Real-Time Unit and Industrial Communication SubSystem (PRU-ICSSG)

    Well ... they may be wrong: the chip's datasheet doesn't agree:

    alt text

    alt text

    alt text

    🙂

    HOWEVER, I just contacted the beagleboard people and they say that the PRUs are there, but not supported by TI. Beagleboard people have got them to work anyhow and the relevant patch is almost mainlined: https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/arm64/boot/dts/ti/k3-j721e-main.dtsi?h=v5.19-rc2#n1812

    Interesting that TI have them disabled, I wonder why.

    Did you get Bela running on the older AI in the end or was it not worth it with the out of control temperatures without active cooling?

      AndyCap Did you get Bela running on the older AI in the end or was it not worth it with the out of control temperatures without active cooling?

      We got it to run yes, there was a GSoC project last year (https://dhruvag2000.github.io/Blog-GSoC21/logs/). There is a kernel issue that makes it crash at boot half of the times, but it essentially worked. The work done there should make it relatively straightforward to port to the new BBAI-64, as we were using the rproc driver.

      Nice (apart from the boot issue!).

      If I manage to get one I will let you know how it goes, if the passive cooling is up to it then I could have a play with Bela on it...

      So with a very basic test looking at temps:

      Temperature in room is around 18 degrees at the moment I guess. I have put some stands on so the heatsink is sitting about 2 cm above desk.

      Without any fan:

      At idle I am getting 58 degrees.

      After loading the two arm cores at 100% (using yes) I am getting 68 degrees and it seems to be sitting there with slight variations.

      I can't find my box of fans currently so I just pointed a desk fan at it:

      idle: 34 degrees
      100% load : 38 degrees

      I have no idea what speed the cores are running as the cpufreq stuff seems to be missing.

      And a quick basic benchmark (prime numbers Sieve)

      Taking the BB Black as the base

      AI32, no fan and suffering from heatstroke is slower at 82%

      AI64, single core, no fan is faster at 366% (pretty sure cpu is not at max freq here)

      So I have now tried to get it to throttle down by sticking it in a box with no ventilation.

      After an hour I have managed to get it up to 90 degrees with no throttling. It seems to be sticking to 90 degrees.

      I'm quite impressed with the temperatures compared to the AI32, they have done a better job.

      I'm going to try to get the PRUs, C7s and R5s spinning as well to see what effect that has, this may take some time...

        Thanks for taking the time of investigating and sharing

        AndyCap up to 90 degrees with no throttling.

        is that good or bad? ... makes me wonder whether throttling is implemented at all?

        Weird you don't get the CPU frequency ... what does /proc/cpuinfo say?

        Just says:

        processor	: 0
        BogoMIPS	: 400.00
        Features	: fp asimd evtstrm aes pmull sha1 sha2 crc32 cpuid
        CPU implementer	: 0x41
        CPU architecture: 8
        CPU variant	: 0x1
        CPU part	: 0xd08
        CPU revision	: 0
        
        processor	: 1
        BogoMIPS	: 400.00
        Features	: fp asimd evtstrm aes pmull sha1 sha2 crc32 cpuid
        CPU implementer	: 0x41
        CPU architecture: 8
        CPU variant	: 0x1
        CPU part	: 0xd08
        CPU revision	: 0

        I asked on the forum and Robert got back saying the cpufreq stuff is not yet there:

        debian@BeagleBone:~$ ls -ltr /sys/devices/system/cpu/cpufreq
        total 0

        I'm guessing it is running schedutil policy:

        debian@BeagleBone:~$ cat /sys/module/cpufreq/parameters/default_governor 
        schedutil

        So I am thinking it is already throttled slightly.

        Mooching around on the net TI say the processor can handle 125 degrees (junction temperature), how that relates to the sensors I am not sure but have found someone complaining that the processor shutdown at 115 degrees from the sensors.

        With ventilation and both arm cores flat out it sticks below 70, I'm happily messing around with it with no fan.

        Compared to the A!-32 which can't even idle without shutting itself down after a while it is a breath of fresh air!

        I have asked on the forum if anyone has looked at Xenomai on it, no bites yet...

          AndyCap BogoMIPS : 400.00

          does that not mean 400MHz?

          AndyCap I asked on the forum and Robert got back saying the cpufreq stuff is not yet there:

          but he'll know the current frequency for sure!

          AndyCap I have asked on the forum if anyone has looked at Xenomai on it, no bites yet...

          seems unlikely ... what kernel is it running? The latest BB xenomai kernel is 5.4 and it doesn't actually build right now.

            debian@BeagleBone:~$ uname -r
            5.10.109-ti-arm64-r49

              I see your "ti-linux-xenomai-4.19.y doesn't build" issue has been open for over a year without response, not looking hopeful.

              couldn't expect much as it was an issue that only happened with my config file.... I eventually got it to build somehow though, as that's what was used for the BBAI(-32) (https://github.com/BelaPlatform/Bela/pull/669).

              AndyCap 5.10.109-ti-arm64-r49

              Right, I had the feeling that all the work they had done for this board was done on 5.10 , so there's not even a branch to build a Xenomai kernel for this board. One could always try to run Bela without Xenomai ... (ah-ehm, needs a few changes to the source code, but a minimal example shouldn't be too hard to put together). The other challenge is the PRU you need to get to work. The good thing is that (PRU-wise) the code done last year for the BBAI should work also here with only minor changes. Trying to rebase it right now.

              Thanks, I will look at all that once I actually get some PRU stuff actually running, everything seems different with this kernel and added to that I don't really know what I am doing 🙂