4 ms·
I wonder what the cost-effectiveness is in terms of compute power. I'd assume not great, but who knows? Raspberry PIs pack a lot of punch for their size and pri
by _bxg1 7y ago
I wonder what the cost-effectiveness is in terms of compute power. I'd assume not great, but who knows? Raspberry PIs pack a lot of punch for their size and price.
- chapium 7y agoDefinitely not great but thats one of the tradeoffs.
- jandrese 7y agoIf they can leverage the GPU on the Pis it could be quite a bit of compute power spread across that array. Even if they're purely CPU bound that is 4240 cores which is not nothing. Power wise each Pi going flat out burns around 5W, so all told you're burning ~6kW, which is quite modest for a big computing setup.
- willis936 7y agoIt looks like you can get 400 single precision MFLOPS for those 5W. So 80 MFLOPS/W +/- 20%. It isn’t the worst and it isn’t the best, but it is respectable. It would be a big hassle to make and maintain a cluster of thousands of these. https://www.arnabkumardas.com/platforms/nvidia/nvidia-jetson-nano-review-and-benchmark/ https://www.arnabkumardas.com/platforms/nvidia/nvidia-jetson... https://raspi.tv/2019/how-much-power-does-the-pi4b-use-power-measurements https://raspi.tv/2019/how-much-power-does-the-pi4b-use-power... https://en.wikipedia.org/wiki/Performance_per_watt https://en.wikipedia.org/wiki/Performance_per_watt
- ramses0 7y ago...but you could experiment with "disposable computing" (ie: similar to original google datacenter racks: never replace a node, only replace a rack at a time if 51% is failing).
- jacquesm 7y agoThat and interesting configurations (systolic arrays, message passing architectures, Erlang on a large number of nodes and so on). Lots of fun with stuff like this.
- willis936 7y agoOh jeez. Out of curiosity I did a comparison against a gaming GPU. A GTX 1080 uses 200W of power for 9 TFLOPS, which is roughly 500x better performance per watt than a raspberry pi. Outside of the fun of it, and maybe a cheap way to prototype larger systems, I can't see a serious reason to make a raspi cluster.
- asdfasgasdgasdg 7y agoWhen it comes to cluster-scale computing, size is not the main issue. The main issues are: compute/$ and compute/watt. There is also the matter of how much RAM and disk can be attached to each board, and the speed of the NIC. Gigabit ethernet is a little slow these days.
- willis936 7y agoGbE is slow, yes, but GbE for the amount of compute a raspi 4 can do is massive. Raspis don’t even have the oomph to max out GbE links. You hook 48 raspi 4 into a switch and using a 25 GbE link to a storage cluster. Edit: It is conceivable that a raspi has the oomph to max out a GbE link but is limited to 300 Mbps due to the use of a USB 2.0 bridge in between the ethernet interface and the SoC. My mistake.