Flaming Raspberries

[a1call]

Hi,
I was introduced to Raspberry Pis by another thread on this board. Ordered one and find it quite interesting. Looks like a very low cost computer farming unit.
Off the shelf installed Pari-GP from synaptic packages from the default OS.

Haven't had time to explore any further.

Does the limited memory makes prime hunting of arbitrary Precision useless?
Can the memory be increased?
Are there really 4 cores available?
Any comments or heads ups?

Thanks in advance.

[ET_]

Originally Posted by a1call

Hi,
I was introduced to Raspberry Pis by another thread on this board. Ordered one and find it quite interesting. Looks like a very low cost computer farming unit.
Off the shelf installed Pari-GP from synaptic packages from the default OS.

Haven't had time to explore any further.

Does the limited memory makes prime hunting of arbitrary Precision useless?
Can the memory be increased?
Are there really 4 cores available?
Any comments or heads ups?

Thanks in advance.

While prime hunting of arbitrary precision is feasible, it is still a bit outy of reach; but factoring is something the PIs would love. Yes they have 4 real cores available, and scale fairly well. Memory can'tbe increased, but you could buy a Odroid-C2, that is a PI-clone board computer clocked higher (1.5 GHz) and with 2 GB of memory. I bought 5 of them to run a s a mini-cluster.

[M344587487]

Originally Posted by ET_

While prime hunting of arbitrary precision is feasible, it is still a bit outy of reach; but factoring is something the PIs would love. Yes they have 4 real cores available, and scale fairly well. Memory can'tbe increased, but you could buy a Odroid-C2, that is a PI-clone board computer clocked higher (1.5 GHz) and with 2 GB of memory. I bought 5 of them to run a s a mini-cluster.

A Renegade could also be a good choice if memory is a limitation, comes in 1 2 or 4 gig varieties of DDR4: https://www.indiegogo.com/projects/r...s/x/17824624#/

I can try and bench the memory performance of a Renegade vs a pi3b and a pi3b+ if someone can suggest a good way to do that (my renegade is the 1 gig version).

[ET_]

Addendum: if you plan to run 24/7 CPU-intensive programs, consider adding a passive cooler and/or some air fans.

https://www.picocluster.com/index.php

[EdH]

Just one additional comment:

Unless something has changed and I missed it, all the sponsored OSs are 32-bit. SUSE has a 64-bit OS for the Pi3b

[wombatman]

Originally Posted by ET_

Addendum: if you plan to run 24/7 CPU-intensive programs, consider adding a passive cooler and/or some air fans.

https://www.picocluster.com/index.php

https://www.amazon.com/gp/product/B0...?ie=UTF8&psc=1

I can highly recommend this case. It has a small (quiet) fan and heatsinks that keep the Pi literally about 15-25 C cooler than the passively cooled ones I have. The price is a bit steep, but the temperature difference can be the difference between the Pi shutting down to stop from overheating or not when under full load.

Do note that it's not for the newest Pi (the 3B+), though I'm sure they'll come out with another one.

[ewmayer]

Originally Posted by ET_

While prime hunting of arbitrary precision is feasible, it is still a bit outy of reach; but factoring is something the PIs would love.

Factoring as in TF? I was under the impression that GPUs are so overwhelmingly faster (by any sensible metric, including per-watt throughput) at TF that using any CPU-based platform for that is a waste of electricity. Whereas the ARM-core-based micros do parallel-scale quite well at LL-testing, so can be used effectively for DC assignments, even if first-time testing is unrealistic (due to the 6-month assignment expiry) for all but the highest-end ARM-based micros, meaning at least 8 cores and capable of 128-bit vector SIMD arithmetic (a.k.a. asimd).

As far as 32-bit OSes go, that's a problem if the CPU supports asimd (specifically the aarch64 variety) since Mlucas requires 64-bit-ness for the asimd assembly code. Generic-C non-SIMD builds on 32-bit should be OK, though I don't have access to such a system to actually check that.

[a1call]

Thank you for all the replies.
Being 32 bit means there is not possible to install primo on it.
Also from what I saw, PARI operations get very slow when you go into 100k dds. Probably due to 32 bit architecture being very slow.

Regarding increasing the memory, is it possible to add ramdrive to it from say a USB stick and effectively increase the memory?

Thank you very much for all the insights.

[M344587487]

There's 64 bit for all of them, it's just that for raspberry pi's you need to go with an unofficial distro ( https://github.com/sakaki-/gentoo-on-rpi3-64bit ) as they are stubbornly against supporting 64 bit and as first movers can get away with it. No you can't increase the working memory with a USB stick, you would have a very slow swap space.

[EdH]

Originally Posted by M344587487

There's 64 bit for all of them, it's just that for raspberry pi's you need to go with an unofficial distro ( https://github.com/sakaki-/gentoo-on-rpi3-64bit ) as they are stubbornly against supporting 64 bit and as first movers can get away with it. No you can't increase the working memory with a USB stick, you would have a very slow swap space.

Thank you for the link. I haven't tried this 64-bit distro yet. But I will soon...

[wombatman]

I can second that distro. The best part is, you can email the creator directly and they will help you out with any issues you're having.

It has been rock-solid for me so far, and setting up my Pis to work with CADO-NFS is super easy.