Who would pay $120 for a Raspberry Pi?

Raspberry Pi's 16 GB Pi 5 completes their lineup: from 2 GB ($50) to 16 GB ($120).

Raspberry Pi sent me the Pi 5 used in this video for testing and review. They did not pay me to make this video, nor have any editorial input or review before the video was live.

Mentioned in this video:

Contents:

00:00 - Are they out of their minds?
01:10 - Not too much, not too little
02:27 - N100: a better deal?
02:57 - Cosmetic improvements
03:53 - Keeping up with the Joneses
04:30 - LLMs and AI
05:12 - Better performance
06:46 - Consolidation
07:50 - Fun and 4K Gaming
09:04 - Should you buy it?

[Atom]

I'd rather want a Raspberry Pi Zero 3W with 4GB of RAM or at least 2GB. It's long due now.

[ExplainingComputers]

Greetings Jeff! :) Fantastic presentation.

[gmtoomey]

I use Raspberry Pi for two things - PiAware & free Mathematica. For anything else I'll be moving to x86 minipcs, the Pi value proposition is not what it was 5 years ago.

[robinmaurer]

Where did my 20€ mini tinkering pc go :'c

[very-ann]

I like that the conversation around developers adding bloat on top of bloat every generation is happening. Hopefully more industry leaders and influencers talk about it.

[kritikusi-]

"Llama increased from 1.2 to 1.3" LMAO. I am sold. Thanks Jeff!

[kelvin]

Why does it feel RPi have left their "lets get a cheap computer out to encourage kids to program" ethos. I mean it is $50 for the board, plus PSU, HDMI cable(s), storage.

[misteragony]

120$ without proper heatsink, powersupply, case, and with crappy usb storage... I moved away from Pi's for all my home automation, energy monitoring/management, ... stuff. Especially the usb storage gave me stability issues and it was too "fiddly" for me. Now I run unraid on a mini PC with 6 sata disks as NAS and 'non critical' containers. And a rock5b board (16Gb ram) with 1TB nvme storage for everything home automation related, running a dozen or so containers. It's been rock stable this far, but if the latter would break down, I will replace it with a mini pc as well and move away from ARM SBCs. For anything that requires gpio's, I tend to try to use something from the ESP32-family microcontrollers.

[otak_]

For the Steam CPU usage it's not only decompression, it's also decryption. This is CPU intensive for any CPU, even if there's instruction support for AES-NI etc.

[iroh]

The equation on the graph at 0:27 took me a minute to understand. I was trying to plug in 32 GB to see what a hypothetical Pi with that much ram would cost, only to get an absurd value over $7000. The equation seems to be treating 2 GB as x=1, 4 GB as x=2, 8 GB as x=3, and 16 GB as x=4. If I follow the pattern and use 32 GB as x=5, it comes out to $172.50 instead. However, that got me wondering what a normal linear fit using the actual GB values would look like, and it turns out that y=5x+40 exactly (r squared = 1). In that case, a 32 GB Pi would come out to 5*32+40 = $200.

[yetson]

I paid $99 Australian dollars ($62 USD after conversion) for a KAMRUI N100 mini PC with 16GB of RAM and a 512GB SSD. Was very happy when I scored that deal. The only thing I didn't like about it is that it shipped with Windows and the seller had thrown in a spyware (yay, Freebie) inside the operating system image, but I detected it early and replaced the OS with Linux.

[digiscream]

The trouble is...none of these use cases are better on a Pi than on an old i5 NUC. The only advantage that the Pi has is the accessible pins, but that's not really of any use to somebody trying to train AI models etc. Hell, even running it as a NAS, the network and CPU performance are far more of a bottleneck than the RAM capacity.

[stan-bihl]

08:40 Jeff, no offense, but this is a baseless tangent. We have retail Steam games with data discs around here, or at least had them until very recently, apparently far longer than USA wanted them. The decompression of those was pretty much as CPU intensive as Steam downloads at the time, for similar reasons - the DVD drives aren't fast, more uncompressed data means more DVDs (and more swaps), and the CPU is presumably not doing anything else when gamer wants a new game, so might as well use it (especially back when they couldn't throttle). And this wasn't specific to Steam games, the installers of older PC games tried their best to do the same thing. So do PlayStation ones since they started installing as a standard, in spite of those dedicated decompression circuits in PS5. Discs have advantages sometimes, but they do nothing to prevent intense emulated installation decompression computing.

[jagc]

About the steam download speed, most gaming PCs have really slow internet and really fast CPU, so it makes sense to optimize for this. Very few people have fast internet and a slow CPU like you, so it makes total sense to use intense compression for game downloads

[PuntoHowto]

I don’t need a 16Gb Pi but it would be nice if Raspberry Pi would give us who use them for everything, a POE hat for the Pi 5

[marksmithcollins]

The problem is not inflation.
The real problem is that RPi is getting itself out of segment between MCU and Arm/x86 mini PCs, and fighting against well established cost effective solutions.
RPi was exceptional for its price/target. now it is no more at the price once it were, also no more at the power once it had.

[nikobellic]

+1 on praising the armor lite v5 cooler. A hidden gem, until this video! I'm glad i ordered a couple more yesterday

[jmr]

9:27 I keep watching to see who he will be next time.

[rednassie]

I'm personally against "just upping all the specs for no reasons" because it makes developers not take power usage into consideration anymore and makes everything less efficient. Not saying that we should code everything into assembly or something, just that a calculator app doesn't need a full gig of RAM just because phones these days can handle it, and it's cheaper to develop with

[zeberto]

The piece about software bloat is so true. Most modern games these days are made for consoles first and thus are developed for systems that have 12-16gb shared system and vram. There is little to no insentive for the devs to streamline the code to work well with less ram. Contrast that with the way say Satisfactory was developed which was PC first and has had tons of optimisations to work on less powerful hardware.