I have no idea if some of the projects developers read this but I hope somebody does.I was quite shocked to find not a single health/biology based project supports Linux + ARM (Raspberri Pi, etc).

This is pretty surprising.

Now there aren't that many Android + ARM biology project either (WCG has a couple and only one active near it's end ...OpenZika).

Is it really that hard to support Linux + ARM? Why no support outside of the Physics, Astronomy and Math projects?

There are quite a few ARM + Linux devices out there and I'm sure people would run BOINC Biology/health related projects on them IF there were any....

So....what gives?

There are quite a few ARM + Linux devices out there and I'm sure people would run BOINC Biology/health related projects on them IF there were any....

It isn't just the health/biology projects where support is lacking for ARM+Linux. There are several other projects that don't support ARM +Linux/Android. CPDN my favoured project doesn't either and unless someone donates a serious amount of money to enable development and maintenance to so it isn't likely to happen when the project people are working flat out all the time as it is.

I don't know if this is the issue with the health/biology projects but suspect it may be.

Projects will use whatever OS the project people are familiar with.
And if that's Windows, and it does the job, then there's no big conspiracy about not using ARM processors.

There's also the question of what computational power the projects need for health/biology research.

One of my NVidia GTX 970 GPUs - still a reasonably powerful device - is closing in on completion of a GPUGrid health/biology task - after 20 hours of solid computing.

I hate to think how long a Raspberri Pi would have taken for that task - certainly longer than the project researchers are willing to wait.

I hate to think how long a Raspberri Pi would have taken for that task - certainly longer than the project researchers are willing to wait.

The CPDN tasks running on my laptop are going to be between 10 and about 45 days total computing time. While not the fastest machine on the block by quite a ways, it is still a lot faster than a Raspberry Pi. And I think if I didn't know the estimates were way out running under WINE I would have given up on seeing them on it never mind with a Pi. (One that I estimate based on time and work done will take 16 to 17 days has an estimate of 188 days! I shudder to think what that would be on a Pi?

I think when we talk about ARM we definitely have to talk about WHICH ARM processor. Some are far faster than others. I have a bunch of Android tablets that have processors far slower than the one in the Pi 3+ that I have that crunch OpenZika on WCG, usually more than one a day.

Here is the Linpack performance of the latest Pi 3+ in comparison to older versions

Here is the Linpack performance of the latest Pi 3+ in comparison to older versions

About a fifth of the performance of my machines which are still slow by today's standards. ARM need to bring out something a lot faster before there would be any point in the projects which have longer tasks developing for it. Those with short tasks are another matter as there are so many android and other arm processors out there they could make a significant contribution.

It's not that these ARM processors are slow per se, but they're hampered by the design and what the device manufacturer puts on as operating system.

I have two devices here that have a fast 64bit CPU in them but since the Android on it is 32bit, and it lacks the 32bit libraries to run stuff on a 64bit CPU, they're practically useless. Or they run SDK < 26, which is the bare minimum these days for applications distributed via the Google Play Store.
Aside from that, even if I could run these full bore, they have no way to get rid of all that heat in a sufficient way. Although my other two devices now have a 140mm fan standing next to them that blows cool air over them.

Here is the Linpack performance of the latest Pi 3+ in comparison to older versions

Unfortunately, I can't find any results for Linpack tests on my NVidia GPUs. BOINC has the theoretical figure as '4087 GFLOPS peak', which we know is ludicrous in benchmark terms, but I'd guess we might be in the low hundreds of GigaFLOPS.

Very roughly, that's a thousand times faster than even the best Pi. My 20 hour job would take 20,000 hours (if the Pi could handle the memory needed): that's over 2 years. We'd all be dead before they found a cure.

Raspberry Pi3+ is still with old 40nm cpu. Standard cpu are moving to 7nm.
And, during 2019, no new version.
If you want to crunch seriously with little board it's better to use Odroid or Pine64 or others

TN-Grid does genetic research and supports ArmHF and Arm64.

I am running 12 RPi3's with an Arm64 Kernel and Debian Buster (Headless accessed via SSH and a couple with GUI's via RDP), they are happily crunching this project along with Universe, Yoyo and DHEP.

With these Arm boards you have to also factor in purchase and power costs, at £30 and max of 10W (More like 5W with nothing plugged in, wired LAN and headless) they are still reasonably efficient. They have so many other uses as well that does not stress them that you can always run boinc in the background. I'm running NAS, DHCP, DNS, NTP and Web servers with mine along with a bunch of test stuff and they just stay on 24/7. Netbooted with ZRAM swap means no SD card required either.

A more modern board will be more powerful as the 3B was released back in 2012 (the 3B+ was a disappointment) but it's still pretty good as the Pine64 or ODroids can cost double and suck more power as well.

By the way comparing CPU and GPU compute is comparing Apples to Oranges.

Edit: Power Consumption

A more modern board will be more powerful as the 3B was released back in 2012 (the 3B+ was a disappointment) but it's still pretty good as the Pine64 or ODroids can cost double and suck more power as well.

Oh, well, i don't think that 70£ are a bleeding (and consume, at full load, 5W like Raspberry).
And, for example, Odroid N2 is 9/10 times faster in a lot of benchmarks (and 3/4 times in all the others).

A more modern board will be more powerful as the 3B was released back in 2012 (the 3B+ was a disappointment) but it's still pretty good as the Pine64 or ODroids can cost double and suck more power as well.

Oh, well, i don't think that 70£ are a bleeding (and consume, at full load, 5W like Raspberry).
And, for example, Odroid N2 is 9/10 times faster in a lot of benchmarks (and 3/4 times in all the others).

Check the benchmarks carefully. If you don't run Raspberryian 32 bit, but run a 64 bit Linux on the Pi the benchmarks will be very different.

Check the benchmarks carefully. If you don't run Raspberryian 32 bit, but run a 64 bit Linux on the Pi the benchmarks will be very different.

Oh, well, 20% more
Bench

Raspberry Pi3 was great 5y ago, not now.

I have a top of the line (or used to be, anyway) OnePlus 6t here.
With all 8 cores active on the Snapdragon 845, I barely manage to hit what a core i5-3317U does from 6 years ago.

There's a reason these bigger tasks for projects don't use arm. I turned my core count down to 3 so I wasn't going to end up with a swollen battery over time.

I turned my core count down to 3 so I wasn't going to end up with a swollen battery over time.

Even running just one core on my phone with WCG I see the message, "computing suspended, waiting for battery to cool down." This has gotten worse since I put the phone in a case but having wrecked a few phones over the years I decided I wanted to keep my fairphone for more than 18 months!