Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

I regret building this $3000 Pi AI cluster

jeffgeerling.com

Post reply on HN
Print view View original

321–330 of 377 posts

Re: I regret building this $3000 Pi AI cluster

#321

Post by jaredhallen »

greenavocado wrote: post #314
michaelt wrote: post #265

Earlier quoted context omitted.

Then you should take a good hard look at older, much cheaper Raspberry Pis. Then look at Apple’s ARM offerings, and AWS Graviton if you need ARM with raw power. If you need embedded/GPIO you should consider an Arduino, or clone. If you need GPIOs and Internet connectivity, look at an ESP32. GPIOs, ARM and wired ethernet? Consdier the the STM32H. Robotics/machine vision applications, needing IO and lots of compute pow…

N100s absolutely curb stomp Pis and have very good video hardware encoders/decoders to boot

A different thread on here the other day seemed to come to the consensus that if you're considering an N100, you may as well go with a low end Ryzen. And for much the same reasoning being used here. Seems to be a little bit of a slippery slope.

Re: I regret building this $3000 Pi AI cluster

#322

Post by motorest »

globular-toast wrote: post #276
densh wrote: post #238

For anyone interested in playing with distributed systems, I'd really recommend getting a single machine with latest 16-core CPU from AMD and just running 8 virtual machines on it. 8 virtual machines, with 4 hyper threads pinned per machine, and 1/8 of total RAM per machine. Create a network between them virtually within your virtualization software of choice (such as Proxmox). And suddenly you can start playing with…

I've been saying this for years. When the last Raspberry Pi shortage happened people were scrambling to get them for building these toy clusters and it's such a shame. The Pi was made for paedogogy but I feel like most of them are wasted. I run a K8s "cluster" on a single xcp-ng instance, but you don't even really have to go that far. Docker Machine could easily spin up docker hosts with a single command, but I see t…

> I've been saying this for years. When the last Raspberry Pi shortage happened people were scrambling to get them for building these toy clusters and it's such a shame. The Pi was made for paedogogy but I feel like most of them are wasted.

You're describing people using RPis to learn distributed systems, and you conclude that these RPis are wasted because RPis were made for paedogogy?

> I run a K8s "cluster" on a single xcp-ng instance, but you don't even really have to go that far.

That's perfectly fine. You do what works for you, just like everyone else. How would you handle someone else accusing your computer resourcss of being wasted?

Re: I regret building this $3000 Pi AI cluster

#323

Post by kldg »

jacobr1 wrote: post #60
phoronixrly wrote: post #9

Earlier quoted context omitted.

If they were cost competitive for ... anything at all really...

They are competitive for hobbyist use cases. Limited home servers, or embedded applications that overlap with arduino.

I picked up some 1GB Rock-2F boards while available in the US for ~$10/ea. Seems they aren't shipping the 1GB boards to the US anymore though. Before this, I had a couple Raspberry Pis; one I fried, and the other acted as a web server for a few years.

My realization in ordering the Rock-2Fs is I really only need an MMU (that is, an SBC instead of something like an ESP32) when I'm running something with a graphical desktop, which is, outside my workstation, never (except for kiosks, which I use Android tablets for). -OR when I want to plug something into a bloated SBC board which saves me from having to solder a connector on, which is sometimes.

I use one for running a timelapse camera (camera is USB) while another is a portable mp3 player I can put in shirt pocket and which has aux port (tho its aux line is noisy). -So that's two of the four Rock-2F boards in use.... but it took me far less time to think up uses and deploy 25/25 of seeedstudio's ESP32C3 boards I ordered a couple years ago, and have used ~5/25 of the ESP32C6s I ordered early this year. They're so cheap, and use so much less energy than ARM boards, that it's difficult to justify using the SBCs anymore.

I think they're asking $50 for a base 2GB Pi4B, now -- that's 10 ESP32C3 boards (with integrated WiFi and BMS, btw!) -- and the Pi5 is even less competitive except in what I'd characterize as a very unusual scenario where you need high compute at edge (where it's both needed AND the latency of computing at the edge is lower than sending it to central server for processing), OR you need the security of protected memory, OR you have no central server and an ESP32 isn't going to cut it (I'll say, though, that one can run a thermostat with multiple WiFi-connected thermometers, and run a web server interface just fine.).

Re: I regret building this $3000 Pi AI cluster

#324

Post by MathMonkeyMan »

le-mark wrote: post #318
bee_rider wrote: post #264

Earlier quoted context omitted.

Tangentially related: I really expected running old MPI programs on stuff like the AMD multi-chip workstation packages to become a bigger thing.

I actually worked with some MPI code way back. What MPI programs are you referring to?

I don't know, but when I was playing with finite difference code as an undergrad in Physics, all of the docs I could find (it was a while ago, though) assumed that I was going to use MPI to run a distributed workload across the university's supercomputer. My needs were less, so I just ran my Boost.Thread code on the four cores of one node.

What if you had a single server with a zillion cores in it? Maybe you could take some 15 year old MPI code and run it locally -- it'd be like a mini supercomputer with an impossibly fast network.

Re: I regret building this $3000 Pi AI cluster

#325

Post by 0xDE7EC71V3 »

motorest wrote: post #322
globular-toast wrote: post #276

Earlier quoted context omitted.

I've been saying this for years. When the last Raspberry Pi shortage happened people were scrambling to get them for building these toy clusters and it's such a shame. The Pi was made for paedogogy but I feel like most of them are wasted. I run a K8s "cluster" on a single xcp-ng instance, but you don't even really have to go that far. Docker Machine could easily spin up docker hosts with a single command, but I see t…

> I've been saying this for years. When the last Raspberry Pi shortage happened people were scrambling to get them for building these toy clusters and it's such a shame. The Pi was made for paedogogy but I feel like most of them are wasted. You're describing people using RPis to learn distributed systems, and you conclude that these RPis are wasted because RPis were made for paedogogy? > I run a K8s "cluster" on a si…

I’ve learned so much setting up a pi cluster. There is something so cool about seeing code run across different pieces of hardware.

Re: I regret building this $3000 Pi AI cluster

#326

Post by Symbiote »

uncircle wrote: post #297
shermantanktop wrote: post #286

Earlier quoted context omitted.

Presumably the non-economist following them picked up the twenty, unencumbered by theory.

I've never seen a $20 bill on the ground in my entire life, so I guess the economists are actually right.

I found 500 DKK on the ground outside once, which is about $77.

Re: I regret building this $3000 Pi AI cluster

#327

Post by AnthonyMouse »

subscribed wrote: post #302
qmr wrote: post #282

Earlier quoted context omitted.

No need for so much CPU power, any old quad core would work.

Old quad core won't have all the virtualisation extensions.

> Old quad core won't have all the virtualisation extensions.

Intel's first quad core was Kentsfield in 2006. It supports VT-x. AMD's first quad core likewise supports AMD-V. The newer virtualization extensions mostly just improve performance a little or do things you probably won't use anyway like SR-IOV.

Re: I regret building this $3000 Pi AI cluster

#328

Post by wltr »

Waraqa wrote: post #135
wltr wrote: post #48

Earlier quoted context omitted.

Well I have a Pi as a home server, and it’s very energy efficient, while doing what I want. Since I don’t need latest and greatest (I don’t see any difference with a modern PC for my use case), it’s very competitive for me. No need for any cooling is bonus.

>very energy efficient If your server has a lot of idle time, ARM will always win.

My home server mostly not used. It triggers some simple bash scripts each hour, or each night. It serves some simple personal web pages. I access them a couple of times a day, each time randomly. I have a Raspberry Pi 1B, and it looks like despite being massively underpowered, it’s the most energy efficient. And in my use-case, I think of using it for the task.

One day my primary Raspberry Pi broke (turned out to be a PSU issue), and I thought of having an old laptop running 24/7 as a home server. While being not very power hungry, it’s still wants much more energy (plus it has fans). For a casual usage (I forgot to mention Pi-Hole) it feels like an overkill. So, while a Raspberry Pi isn’t the best, it has its niche, and I’m happy of having one (actually, a few).

Re: I regret building this $3000 Pi AI cluster

#329

Post by Symbiote »

yjftsjthsd-h wrote: post #73
blobbers wrote: post #55

Earlier quoted context omitted.

The internet says 100W idle, so maybe more like $40-50 electricity, depending on where you live could be cheaper could be more expensive. Makes me wonder if I should unplug more stuff when on vacation.

> Makes me wonder if I should unplug more stuff when on vacation. What's the margin on unplugging vs just powering off?

That also depends on the country you live.

The EU (and maybe China?) have been regulating standby power consumption, so most of my appliances either have a physical off switch (usually as the only switch) or should have very low standby power draw.

I don't have the equipment to measure this myself.

Re: I regret building this $3000 Pi AI cluster

#330

Post by AnthonyMouse »

anaganisk wrote: post #320
qmr wrote: post #282

Earlier quoted context omitted.

No need for so much CPU power, any old quad core would work.

Aren’t newer CPUs especially AMDs more energy efficient?

Newer CPUs have significantly better performance per watt under load, essentially by being a lot faster while using a similar amount of power. Idle CPU power consumption hasn't changed much in 10+ years simply because by that point it was already a single digit number of watts.

The thing that matters more than the CPU for idle power consumption is how efficient the system's power supply is under light loads. The variance between them is large and newer power supplies aren't all inherently better at it.

Post reply on HN