I discovered that if I crank up the raspberry pi 4 (4GB version) to 1750 Mhz (from default 1500) the average credit is at this point just over 1000 points, it runs 3 simultaneous WUs for R@H, also I embellished the device with a CPU cooler (1/4 of a harvested piece from another mainboard) and a makeshift cooling fan.

This is the device: https://boinc.bakerlab.org/rosetta/show_host_detail.php?hostid=4116810

I used instructions from this site: https://magpi.raspberrypi.org/articles/how-to-overclock-raspberry-pi-4

/boot/config.txt

over_voltage=2
arm_freq=1750

temperature is now around 54 degrees Celsius, active cooling is required since the PI4 already has a cooling problem out of the box.

I've been curious, how are the pi's doing now that the database is only stored once for the project rather than once in each active WU slot? Does the memory footprint of the active WUs get reduced and allow you to activate another core?

---

Rosetta Moderator: Mod.Sense

I've been curious, how are the pi's doing now that the database is only stored once for the project rather than once in each active WU slot? Does the memory footprint of the active WUs get reduced and allow you to activate another core?

You can easily get 4 Rosetta processes running on a Pi 4 4GB with a 2GB zram config (2GB seems to be overkill, but it also doesn't seem to harm the system either).

For a Pi 3B+, 3 Rosetta processes really pushes the host with using swap. I think it would actually slow down the processing to go with 4 processes, and likely 2 processes would be a better fit.

I have 1 stock Pi 4 4GB for a baseline comparison, one Pi 4 4GB overclocked, and 1 Pi 3B using a large ZRAM configuration (250% of system ram). All are using Flirc cases and stay well under 60C.

If you want to track my Pis, here are the direct links to the systems:

Stock Pi 4 4GB:
https://boinc.bakerlab.org/rosetta/show_host_detail.php?hostid=4215281

Pi 4 4GB @ over_volt=4 and arm_freq=2015 mhz:
https://boinc.bakerlab.org/rosetta/show_host_detail.php?hostid=4269102

Stock Pi 3B+ with Zram and 3 Rosetta Processes
https://boinc.bakerlab.org/rosetta/show_host_detail.php?hostid=4244063

I'm using exactly this setup, but i've throttled my Pi4 cpu to run 3 threads at any one time
https://boinc.bakerlab.org/show_host_detail.php?hostid=4235245

I cut back to 3 threads rather than 4 as i found 4 threads running on all cores run hotter, there are 'hot spots' at times where cpu loads runs high (probably when a thread finish and a new one starts while the previous is uploading) and actually returns less credits per task.

<options>
  <ncpus>3</ncpus>
</options>

i reasoned that it could be either be due to memory or cache contention. There is less cpu cache shared between 4 busy threads, that makes it hit memory more often and with dram, it is more common to need wait states before data become available. so overall things run slower. 3 threads works 'just right' it runs well and stable overclocked to 1.75 ghz

i've actually not checked on the memory foot print prior to 4.20, but i noticed that some tasks, e.g. the Junior_HalfRoid* seem to consume more memory, perhaps it is larger etc. and for the larger models, cutting down on running too many threads seem to help with caches etc. if i were to run 4 such tasks, the credit per tasks tend to be lesser than if i only run 3.

the main benefit of 4.20 is it significantly cut back on the disk space used and network bandwidth. i think collectively it probably saved terabytes for everyone and bakerlab. and i think more than likely the database stay in the memory cache and hence access is faster

Wait, one can overclock a pi and run BOINC?
I've been planning to obtain a couple of raspberries pis (for BOINC, camera, learning, etc).
Hmm, I wonder how a Turing PLUG & PLAY PI Cluster would fare. Basically, it's an ITX sized motherboard with 7 slots for Raspberry Compute Modules, which are like compact RAM sticks but with a raspberry pi's compute related guts. One could have 28 ARM cores at 1200MHz in an ITX PC case. The only issue, currently, is that the most powerful Raspberry CM, the CM3+, is equivalent to a PI 3 B+, which only has 1GB of ram. IF they made a CM4+, that would potentially be a pint-sized powerhouse at not too high a price.

I've been curious, how are the pi's doing now that the database is only stored once for the project rather than once in each active WU slot? Does the memory footprint of the active WUs get reduced and allow you to activate another core?

I don't think the memory has reduced, but disk space certainly has. Here is top from one of my Pi4 4GB nodes. I've limited it to 3 tasks at a time. One of the tasks was swapped out waiting for memory. I adjusted the BOINC settings to allow 92% memory when not in use to get it going.

top - 18:09:18 up 3 days, 12 min,  1 user,  load average: 3.21, 3.07, 3.02
Tasks: 102 total,   4 running,  98 sleeping,   0 stopped,   0 zombie
%Cpu(s):  0.0 us,  0.2 sy, 75.0 ni, 24.7 id,  0.0 wa,  0.0 hi,  0.0 si,  0.0 st
MiB Mem :   3857.7 total,    430.2 free,   2707.8 used,    719.6 buff/cache
MiB Swap:    100.0 total,     77.0 free,     23.0 used.   1076.2 avail Mem

  PID USER      PR  NI    VIRT    RES    SHR S  %CPU  %MEM     TIME+ COMMAND
 4254 boinc     39  19  594.0m 467.5m 101.0m R  99.7  12.1 333:46.39 rosetta_4+
 4440 boinc     39  19 1383.5m   1.2g 105.6m R  99.7  32.8  66:17.44 rosetta_4+
 4448 boinc     39  19 1297.2m   1.2g 105.5m R  99.3  30.6  64:39.34 rosetta_4+
  319 boinc     30  10   49.2m   8.0m   4.1m S   1.0   0.2  43:50.60 boinc

---

MarksRpiCluster