I've been upgrading some of my cluster to Raspbian Buster. I have been re-imaging the SD card with buster-lite and then reinstalling BOINC on them. So far nothing much to report, for a headless Pi it looks the same as Raspbian Stretch
The two Pi4's that I ordered have an ETA of September.
All crunch fine although have had some problems with the Odroids on e@h. They are doing U@H and Seti until I get around to figuring out why they do not like E@H. Been busy with other problems and projects.
I like them all but kinda like the Asus TB a little more. They and the Nano are strong crunchers.
I suggested to Noctua last year they might want to look at shrinking one of their coolers to fit the Pi. Back then it was the 3B+ but now we have the 4B producing even more heat. I even got a reply back from them saying they weren’t interested. I think we’ll see more miniature CPU coolers for the Pi.
Hi there guys! Looking at setting up a Pi cluster myself... with the introduction of the Pi 4 I was wondering if it would be necessary to get the 4GB version of the Pi4 or if I would be able to go slightly cheaper and use the 2GB version? looking at getting a small cluster of 4 Pi's to start off with and then go from there...
It depends on what you're going to do with it. As a headless cruncher of Einstein@Home, my pi4's can easily run 4 tasks at a time and don't use half of the 4GB memory. I can't use more than 3 CPU's on pi3 versions - some only 2. If you are also needing desktop gui's or memory intensive distributed apps, stand in line and pay the extra $10 for 4 GB. In a cluster mix and match is easy. I am running everything from Zeros to pi4' s together.
Biggest issue is power and heat. I have gone to POE and fans. All my pi4's have the POE hat with a fan. The pi3's have a case with fans and are POE or a POE adaptor.
I run 4 concurrently on my 1gig ram RPi2s and RPi3, each task takes 202MB ram. They run headless and I increase the memory setting in Boinc to use 90% while computer in use.
It depends on what you're going to do with it. As a headless cruncher of Einstein@Home, my pi4's can easily run 4 tasks at a time and don't use half of the 4GB memory. I can't use more than 3 CPU's on pi3 versions - some only 2. If you are also needing desktop gui's or memory intensive distributed apps, stand in line and pay the extra $10 for 4 GB. In a cluster mix and match is easy. I am running everything from Zeros to pi4' s together.
Biggest issue is power and heat. I have gone to POE and fans. All my pi4's have the POE hat with a fan. The pi3's have a case with fans and are POE or a POE adaptor.
I’m looking at using it for solely number crunching on Einstein@Home, I need something that’s a bit more power efficient than my R7 2700X and GTX 1080... plus I think that having a Pi cluster would be great for continuous crunching seeing as I do tend to use my main PC a lot which means I suspend tasks and leave them sitting there doing nothing quite often.
I have 2 pi4's (4GB) on the
)
I have 2 pi4's (4GB) on the way. One has been delivered already. Both will be powered via poe as part of a 8 node cluster.
Edit: Up and running. So far wow. Will post benchmarks ASAP
Edit: Running BRP about 30% faster than pi3B+. Also can run all 4 CPU's because of the 4GB memory
Thanks for the benchmark link!
Raspberry Pi microSD card
)
Raspberry Pi microSD card performance comparison - 2019
https://www.jeffgeerling.com/blog/2019/raspberry-pi-microsd-card-performance-comparison-2019
4 different benchmarks plus RPI 3 to 4 speed comparison.
I've been upgrading some of
)
I've been upgrading some of my cluster to Raspbian Buster. I have been re-imaging the SD card with buster-lite and then reinstalling BOINC on them. So far nothing much to report, for a headless Pi it looks the same as Raspbian Stretch
The two Pi4's that I ordered have an ETA of September.
MarksRpiCluster
I have a mixed bunch of SBCs
)
I have a mixed bunch of SBCs now.
2x RPi2
2x RPi3
1x RPi3B+
2x Asus Tinker Board
1 Nvidia Jetson Nano
1 Odroid XU4
1 Odroid MC1 Solo
All crunch fine although have had some problems with the Odroids on e@h. They are doing U@H and Seti until I get around to figuring out why they do not like E@H. Been busy with other problems and projects.
I like them all but kinda like the Asus TB a little more. They and the Nano are strong crunchers.
Hope to get a RPi 4 when they are more available.
Cheers!
https://youtu.be/RyUXC3886Ic
)
https://youtu.be/RyUXC3886Ic
RPi 3, 3b and 4 tower cooler. Overkill but still pretty cool. May not be overkill in hot environment.
I suggested to Noctua last
)
I suggested to Noctua last year they might want to look at shrinking one of their coolers to fit the Pi. Back then it was the 3B+ but now we have the 4B producing even more heat. I even got a reply back from them saying they weren’t interested. I think we’ll see more miniature CPU coolers for the Pi.
MarksRpiCluster
Hi there guys! Looking at
)
Hi there guys! Looking at setting up a Pi cluster myself... with the introduction of the Pi 4 I was wondering if it would be necessary to get the 4GB version of the Pi4 or if I would be able to go slightly cheaper and use the 2GB version? looking at getting a small cluster of 4 Pi's to start off with and then go from there...
Cheers!
It depends on what you're
)
It depends on what you're going to do with it. As a headless cruncher of Einstein@Home, my pi4's can easily run 4 tasks at a time and don't use half of the 4GB memory. I can't use more than 3 CPU's on pi3 versions - some only 2. If you are also needing desktop gui's or memory intensive distributed apps, stand in line and pay the extra $10 for 4 GB. In a cluster mix and match is easy. I am running everything from Zeros to pi4' s together.
Biggest issue is power and heat. I have gone to POE and fans. All my pi4's have the POE hat with a fan. The pi3's have a case with fans and are POE or a POE adaptor.
I run 4 concurrently on my
)
I run 4 concurrently on my 1gig ram RPi2s and RPi3, each task takes 202MB ram. They run headless and I increase the memory setting in Boinc to use 90% while computer in use.
Chris put up another RPi4 cooling video.
https://youtu.be/AVfvhEJ9XD0
bowguy wrote:It depends on
)
I’m looking at using it for solely number crunching on Einstein@Home, I need something that’s a bit more power efficient than my R7 2700X and GTX 1080... plus I think that having a Pi cluster would be great for continuous crunching seeing as I do tend to use my main PC a lot which means I suspend tasks and leave them sitting there doing nothing quite often.