On Raspberry Pi's 10th Anniversary, Eben Upton Ponders the Prospects for a RISC-V Pi

On the 10th anniversary of the Raspberry Pi, the Register takes a look backwards — and forwards — in a new interview Eben Upton: As for the future, a Raspberry Pi 5 is inevitable but, in a blow to those hoping for something even more exotic than the usual Broadcom Arm SoC, RISC-V isn't on the cards, certainly not in the timeframe for the next generation of Pi. "I can't go out today and license a RISC-V core," says Upton, citing the current state of the ecosystem, "which is even as good as the core of the Raspberry Pi 4."

"Hopefully I will in the future," he adds.

The implosion of the Nvidia/Arm deal has also reduced the chances of a RISC-V Pi (in the 2030 timeframe), down from 20 per cent to 10 per cent, according to Upton. "The ownership of the core of the de-facto standard low-power architecture by an organisation that's also a player was potentially very destabilising... and it was always going to be to the benefit of the obvious alternative architecture."
Upton answered questions from Slashdot's readers before the first boards were even ready in 2012. But he already knew that the device would be wildly popular. Upton tells the Register that months before, its Debian-based operating system got 50,000 downloads, "And all you could do was run it in QEMU (there was a QEMU config you could build that was similar enough to a Raspberry Pi).

"50,000 people downloaded this really primitive operating system for a machine that you couldn't buy!"

Finally a real CPU?

[gweihir]

Instead of that barely functional and very limited Broadcom crap-design? That would be something.

Re: Finally a real CPU?

[RazorSharp]

The value in the Pi is that it is economical. I do not know what you mean by a real CPU. RISC-V certainly cannot provide something better for the same cost. I would even argue the Pi is overpowered for what they are often used for. You can do a lot with just a Pi Zero.

Re:

[dskoll]

Also, the software ecosystem for the Pi is second to none. While there are lots of SBCs out there, the Pi software ecosystem and community blows them all out of the water.

Re:

[gweihir]

Also, the software ecosystem for the Pi is second to none. While there are lots of SBCs out there, the Pi software ecosystem and community blows them all out of the water.

You mean all the big-ego-small-skill cretins? Well, second to none is correct but not in a positive sense.

Re:

[dskoll]

Huh? What are you referring to?

Re:

[dfghjk]

He's saying your opinion doesn't matter if you can't roll your own distribution. There is a point there somewhere but it gets lost in the assholishness.

The problem is that the Pi isn't about the computer architecture, it's about enabling a different group of people, those "big-ego-small-skill cretins", to be able to apply computing to solve problems without concerning themselves with whether a Broadcom processor is good or not.

Re:

[Shirley Marquez]

Exactly so; that's been a key to the success of the Raspberry Pi. The organization supports the custom OS for the Pi well, and it pretty much Just Works. That's not something that can be said for most of the Pi alternative SBCs.(Sometimes there are bugs in Raspbian/Raspberry Pi OS, and they were slow at getting out a 64-bit build, so things aren't perfect, but compared to the competition they're great.) There are also lots of add-on HATs and other Pi accessories, which again mostly work with minimal effort.

Re:

[Ostracus]

For when CPUs become fodder for ideology. Are you using the right CPU comrade?

Re:

[Registered Coward v2]

For when CPUs become fodder for ideology.

As soon as the second CPU design was launched...

Are you using the right CPU comrade?

Of course. Are you?

Re:

[Powercntrl]

The value in the Pi is that it is economical.

Well, they were economical before the whole "everything shortage" started. The 4GB Pi4 seems to be going for around $190 lately, if you need one immediately.

Re:

[Chelloveck]

Tell me about it. A couple years ago I had a bunch of services running on my NAS. The NAS was getting old and cranky and hard to upgrade, and doing so meant downtime for *everything*. So I though, RPis are cheap. I'm move each service off onto its own RPi. This will eliminate a lot of interdependency, I can upgrade piecemeal without disrupting the whole system, it'll be great. So I did. Fast forward to today and I want to add a new service. I'll just go grab another RPi and... holy shit! The price has *tr

Re:

[Shirley Marquez]

Supply of the boards is very limited. But you can still walk into Micro Center and buy a Raspberry Pi 400 for its normal price of $70. That's essentially a 4 GB RPi 4 plus a keyboard and case packaged together in a single unit. If you don't need that Pi for embedding it's a good alternative, or if you already have a Pi board that you're using as a desktop you can replace it with the RPi 400 and free your existing board for an embedded application.

Re:

[ncc74656]

But you can still walk into Micro Center

Walk into a what? We don't have those here. We had a Fry's for about 15 years, but the Chinkvirus was the last straw for them.

There is a site that's popped up lately that might be useful for tracking down a Raspberry Pi at non-nosebleed prices:

rpilocator.com [rpilocator.com]

I managed to snag a Pi Zero 2 recently that way...it's making its way here from Blighty as I write this.

Re:

[xwin]

What so limited about this CPU? It is more or less standard Cortex-A53 core. It is not Broadcom's design and in fact an ARM design. There may be some other components that are Broadcom designed but the CPU is not one of them.

Re:

[AmiMoJo]

The GP is talking out of their arse. The Pi is extremely functional and was a real game changer when it came out. The issues it had were mostly because the Pi people screwed up.

Re:

[gweihir]

The GP is talking out of their arse. The Pi is extremely functional and was a real game changer when it came out. The issues it had were mostly because the Pi people screwed up.

Nope. The GP is an actual engineer and computer scientist and can recognize a half-arsed bad design when he sees it. Both the Pi and the CPU used are pretty screwed up. The only thing it had going was marketing and luck to come out at the right time. Like VHS tapes, for example. Really bad technologically, but a big commercial success.

Re:

[account_deleted]

Comment removed based on user account deletion

Re:

[AmiMoJo]

Can you be specific about what is screwed up?

Re:

[serviscope_minor]

He's just got a bee in his bonnet. He thinks he knows something special that shows the Pi is terrible, and his hatred of it demonstrates he's smarter than all the millions of happy customers.

Re:

[gweihir]

Nope. That is just you demonstrating cluelessness.

Re:

[serviscope_minor]

Cool story bro!

Tell you what, why don't you list actual, concrete complaints, and then we can shoot them down. Don't worry, it won't make you look more of a prat (that isn't really possible).

Re:

[gweihir]

Tell you what, why don't you list actual, concrete complaints, and then we can shoot them down.

Ahahaha, no. They are well-documented and solid. You are in denial. Not my job to fix that. I can ridicule you through. Well, I know, most actual experts simply stay out of discussions with bloody amateurs. I should probably too.

Re:

[serviscope_minor]

Ahahaha, no.

In other words, you're blowing smoke.

I can ridicule you through.

I invite you to tell me how the things I made that worked didn't in fact work. This should be good...

Re:

[dgatwood]

What so limited about this CPU? It is more or less standard Cortex-A53 core.

Actually, the Raspberry Pi 4 (2019) uses a Cortex A72 [geekbench.com]. A Cortex A53 was used in the Raspberry Pi 3.

It is not Broadcom's design and in fact an ARM design. There may be some other components that are Broadcom designed but the CPU is not one of them.

Ah, but Broadcom decided which core to use, and they chose poorly, IMO. Broadcom released that SoC in Q2 2019, but it was based on a core from 2015. It was fully four years old when the SoC shipped, and even older when the Pi 4 actually shipped.

For most of the stuff I want to use an embedded board for (NDI video), the Pi 4 is just barely able to cut it reliably (even with direct console access, i.e. no X1

Re:

[xwin]

You do understand that it takes time to produce a chip? Broadcom can't take the latest core from ARM because of many reasons. Cost is just one of them. It takes over a year to produce a chip. Then it takes another year to make a board with that chip. Also the chip in RPI is not exclusively for them. So somehow the price performance balance is selected and the chip is made. The chip must satisfy many customers on the price performance ratio.
People like to complain but to me it is amazing that the credit car

Re:

[dgatwood]

You do understand that it takes time to produce a chip? Broadcom can't take the latest core from ARM because of many reasons. Cost is just one of them. It takes over a year to produce a chip. Then it takes another year to make a board with that chip.

None of those numbers seem plausible to me. The chip was released in 2019, and the Raspberry Pi 4 was released in June of 2019, so unless Raspberry Pi has enough pull to get prerelease samples at least half a year before the chip became available, it likely took much less than a year to do the board design.

Either way, it doesn't typically take a year to release a chip based on an ARM core. The Huawei Mate 8 used A72 cores, and was released in November of 2015, just nine months after the announcement of th

Re:

[hattig]

The A72 is a very cheap core to licence. The cutting edge cores are not.

It was a surprise the Pi 4 actually got A72s in my opinion, but they saved on silicon area with the shrink and kept a tiny GPU.

Pi 5 will likely get A76s, which appears to be the next 'cheap ARM big core'.

Remember the cost of the SoC has to be under ~$10 for the Pi. You can't put 4 $1 CPU cores on it.

Re:

[Osgeld]

the arm cpu is more like the traffic cop, almost everything else is handled by the broadcom ch

Re:

[serviscope_minor]

Are you a wannabe hipster or just a contrarian.

There's a certain kind of person who just loves shitting on the raspberry Pi. If you're expecting a Ryzen 5950 running off USB-A for $30, then it's not the designers of the Pi that are fools.

The Pi family are an amazingly good family of products for a variety of niches. You can get a range from single core 700MHz with half a gig of RAM to a quad core 1.5GHz out of order A73 with 8G of RAM. They're amazingly useful for a variety of tasks, very long lifetime (the

Re:

[gweihir]

There are also people that have used Pi alternatives and the Pi itself and found the alternatives to be much better designed, better documented and having far superior interfaces in the CPU, etc. Some of these people are even PhD level engineers that have a clue what they are talking about, unlike the wannabee "makers" that are so in love with the RPi.

Re:

[account_deleted]

Comment removed based on user account deletion

Re:

[serviscope_minor]

Perhaps these people should list some of these "much better designed" and "better documented" alternatives, then.

It's a weird sort of contrarian anti-hipster hipsterism. It's because the Pi is popular, they have to find some usecase where it's not the best choice just to prove they're smarter than all the people using raspberry Pi's happily. As if that's a thing...

Note: they never actually have that usecase. It's all armchair engineering.

Re:

[account_deleted]

Comment removed based on user account deletion

Re:

[gweihir]

It's because the Pi is popular, they have to find some usecase where it's not the best choice

Alas, they didn't even do that. They never explained any specifics as to what they think is so badly designed and why or presented any sort of actual usecase, but rather just resorted to making some really big claims and then attempted to pre-empt any disagreeing voices by claiming to be a highly-skilled engineer (something that I have not seen any evidence of yet, either!) and therefore superior.

Not needed. It is blatantly obvious to anybody with actual skills. It is completely useless to explain to the others, because they are hardcore reality-deniers.

Re:

[account_deleted]

Comment removed based on user account deletion

Re:

[gweihir]

Not needed. It is blatantly obvious to anybody with actual skills. It is completely useless to explain to the others, because they are hardcore reality-deniers.

Ah, so you don't actually have an actual argument. Got it.

Thanks for making my point. No, I do not expect you to understand.

Re:

[gweihir]

Note: they never actually have that usecase. It's all armchair engineering.

Wrong.

Re:

[dskoll]

Which Pi alternatives, specifically? And how much do they cost?

Re:

[gweihir]

Don't be lazy. There is a whole range out there. All better than the RPi because of no need to use substandard Broadcom hardware and design by actual experienced EEs.

Re:

[account_deleted]

Comment removed based on user account deletion

Re:

[gweihir]

Don't be lazy. There is a whole range out there.

*You* are the one who made the claim that there are plenty of far better alternatives, no one else. You are the one being lazy and dishonest here.

Hahaha, you wish. The RPi basically has created its own cult of people that think there is only one true god. And they leash out against anybody that calls them on their ignorance and stupidity. I just like to poke fun at morons like you from time to time. All the data is out there and easily accessible.

Incidentally, you do not even understand what an "appeal to authority" is. Here is a hint: It does not apply to the difference between experts and non-experts. Non-experts usually do not understand that, see

Re:

[serviscope_minor]

There are also people that have used Pi alternatives and the Pi itself and found the alternatives to be much better designed, better documented and having far superior interfaces in the CPU

Better designed? If you're going to grouch about some suboptimal piece of performance in an aspect I don't care about, then why should I care? Thing is, the RPi is well good enough for a huge variety of tasks, has a long supported life (the model A+ will be available until 2026), is widely available in the distribution ch

Re:

[gweihir]

that have a clue what they are talking about, unlike the wannabee "makers"

Oh wow are you ever sore about people who are better at making shit than you.

Hahaha, no. Cretins with big egos are a nuisance. They are not an object of envy. My designs work since I threw out the 3 RPis I bought. Made by amateurs for amateurs.

Re:

[serviscope_minor]

You don't have any designs. Stop pretending otherwise, no one believes you.

Re:

[gweihir]

You don't have any designs. Stop pretending otherwise, no one believes you.

I am not trying to convince you. That would be futile. I know what I have and what I do not have.

Re:

[serviscope_minor]

It would be futile because you have no idea what you are talking about.

I know what you've done too: the answer is nothing. I know that because you have refused to say anything remotely concrete. You're a basement dwelling armchair engineer with no actual skills or knowledge. You try and feel better about your lack of actual achievement by shitting on others.

Here's a free clue: that won't help. You won't feel better trying to drag others down. Go make something yourself instead. You'll feel better, learn som

Re:

[itzdandy]

I think this comment is fair in a certain context and unfair in another. The Pi4 sits right in the middle. It's way overpowered for many services that run on it, and it's way underpowered for graphical tasks. yes you can run a GUI on it, but it's not pleasant.

I would love to see the Pi 'A' vs 'B' type scenario be a bit more spread. Have the 'A' have identical port configuration but just less expensive components like 100M network and a low clocked ultra-cheap CPU. Then the Pi5 B/B+ whatever be a very m

Re:

[xwin]

Do you really understand what you are saying? You are comparing $35 retail computer with a $700 retail computer. The chip in RPI cost under $10. The M1 cost is probably $100 to $150. The PI has 32 bit or 64 bit bus and M1 has 128 bit bus if not 256 bit. The cost of the chip packaging is not even comparable.
The RPI chip is fine. It compares fine with the comparable SOCs from other manufactures. I did run benchmarks.
If mac mini is cheaper in your case, please get it. But your comparison is idiotic.

Re:

[itzdandy]

Do you read entire comments before posting? I literally said I wasn't comparing the platforms directly, just comparing same instruction set performance.

The Pi4's A72 broadcom is getting handily bested by $3-7 chips from Rockchip, allwinner, and Realtek and absolutely thrashed by an $11 chip from amlogic. The BCM2711 is basically at the back of the pack for ARM =2Ghz quad core chips. These are ALL ARMv8 chips.

Estimated BoM on the M1 is $64. ~6x the price of the BCM2711 for 10-100x the performance. Whole

Re:

[Shirley Marquez]

A new CPU for the Pi doesn't necessarily mean RISC-V. They could also use a different Arm SoC, including the possibility of designing their own as they did for the Pico microcontroller board. The Pi Foundation is already an Arm licensee for that design, though their current license may not cover the A-series cores that would go into an SoC for a Raspberry Pi 5.

Cooling Fan

[backslashdot]

Why does the raspberry pi 4 require a fan? That was a dumb decision. I hope the 5 does not require a cooling fan.

Re:

[xwin]

I don't think it needs a fan. It may need a heat sink. The problem is that the CPU die is getting smaller and clocks are getting higher. It is hard to dissipate the heat with a small die. Fan would just allow you to use a smaller heat sink. The latest chip does not use stacked RAM package and that probably helps.

Re:

[dskoll]

The Pi 4 does tend to run hot, but it doesn't require a fan. I purchased a Pi4 case that's made of aluminum where the entire case acts as a big heat sink, and it runs fine.

Re:

[ArchieBunker]

Because so many people complain it isn't powerful enough.

Re:

[xwin]

If RPI is not powerful enough, people should use something else. I use ARM6 based computer with 256MB of ram as my media server and a backup server. It does not break any speed records but it does the job. People want to use Arduino to run a internet facing server and then complain that it is slow.

Re:

[Badoit]

I have a Pi4 8gb that is a Tor proxy, torrent box, dlna server, NAS, nzb downloader, and it is in an aluminum Flirc case. No fan. Normal temp around 48 C. Sometimes goes to 58-60. Also a Pi4 4GB that I bought when they were released in 2019. Because of what I read about heat, I bought a Pimoroni Fan Shim. The controlling software triggers at 65 C and the fan cuts out when the CPU is down to 55. Since the Pi4 firmware updates a couple of years ago it never runs. I see around 45 to 50 normally. As thermal th

Re:

[ncc74656]

Strictly speaking, it doesn't. Without some sort of additional cooling, though, it might throttle. That "additional cooling" could be a fan, or a decently-sized heatsink. There are even cases that double as heatsinks, which might work for your application.

Display tech

[backslashdot]

They should explore making ultra cheap display technology. Display is getting to be the most expensive thing.

Re:

[fuzzyfuzzyfungus]

I don't know what supply chain mysteries are at work; but displays seem to be more of an availability/packaging problem than a cost one: if you look into the murky depths of absolutely garbage tablets you start seeing genuinely useful displays(ie. better than 800x480 or 1024x600, which have some definite specialty applications but are pretty miserable as general-use screens) in the $50-70 range; and that's quantity one, for something that also includes a SoC of some description, a battery, some RAM and flas

Re:

[serviscope_minor]

You're not wrong. The official screen is one of the cheapest options, but it's become fashionable to prevent main application windows from resizing vertically below a certain amount, so a lot of programs won't work on the screen well because e.g. the buttons are off the bottom. There's really no need, but there you go.

Yes

[pele]

Please!

No RISC-V?

[srg33]

Above: "I can't go out today and license a RISC-V core which is even as good as the core of the Raspberry Pi 4." I just don't buy that. There seem to be RISC-V cores available: commercial and even BSD licenses. SiFive etc. RISC-V Exchange: Cores & SoCs [riscv.org] RISC-V might be too expensive to produce, but . . .

Re:

[_merlin]

None of them perform as well as the relatively mediocre ARM cores in the Raspberry Pi 4, especially in the same power envelope. The Pi 4 is already pushing the limits for power dissipation, requiring a fan. Requiring any active cooling is a big step backwards. Low power and natural air circulation cooling was a big benefit for the Pi lineup.

Re:

[wildstoo]

A fan isn't strictly necessary. Depends on your cooling setup and workload. If you have good passive cooling e.g. custom case with adequate heatsinking then it's generally fine in normal use. If you throw a Pi4 in the official plastic case in a warm room and then do multi-thread compiles non-stop then yeah, it'll probably throttle.

What Raspberri Pi's?

[BrendaEM]

They are $189 each on Amazon because they are in short supply.

Cheri RISC-V would be worth it

[carton]

RISC-V is interesting to the industry, but as the summary says there's no value beyond some day lower price / performance.

Cheri RISC-V [cam.ac.uk] can do things ARM can't. It can provide memory protection without an MMU or TLB flushing so it can be used within a process, for example to isolate an untrusted codec, parser, or web rendering library. It can provide bounds-checking to C-like languages in hardware so there's no instrumentation or associated performance penalty. It can enforce C 'const' qualifiers, for exa