What Do You Want To See Out Of The Redesigned, Next-Gen Raspberry Pi?

[uid313]

Raspberry Pi is a big name, so its a big win for any contractor to win have their SoC on that that board. So I think they should shop around and see whats to offer, see if they can negotiate with Samsung who have Exynos and HiSilicon who have Kirin to see if they can offer them a free open source board with open schematics and open source firmware and no binary blobs, before they decide to stay with Broadcom.

They ought to consider RISC-V if they can yield something more open.

I would like to see UFS storage. Its what phones use nowadays, so its like a modern eMMC or SD I guess.

I would also like to see it being powered by a USB Type-C connector, with power delivery over USB. So it just uses standard USB for power, no weird proprietary power connector.

[Guest]

• Vulkan capable GPU
• 2-4GB RAM
• 8-core ARM CPU
• Don't care for all the ports that only make it big
• Don't care for the $35 price. Make it 50-70 if it takes to have these specs.

[baracca]

Basically Raspberry Pi has always been trying to imitate its competitor (Beagleboard, which came out two years earlier) and Pi clones.

Suggestion #1: stop, please, STOP with that moronic "squared" form factor.
You were able to shrink the Pi to Zero, then you can do it: a 1"x3" board, or at most a 1.2"x4", and all connectors on one side. C'mon, realize at once you will enable a huge lot of new projects with bizarre space constraints.

Suggestion #2: start saving on less used peripherals.
Raspberries could easily produce cheaper models featuring *unpopulated* camera/video outputs. Not everyone needs HDMI. Not every embedded project needs audio and extras. And -believe me- a very few projects need both ethernet and wifi/bluetooth.

Suggestion #3: stop demanding 5V input.
Some ten years ago, in the epic times of the first Beagleboards, I humbly asked for a board having some step-down converter to accept anything between 6 and 30V input. You can get a good 3A rated ones from Pololu and others for a reasonable price; I guess mass production would keep the price down and allow for a very low-ripple design. This would allow using any sort of battery packs - from 6 AAA to your 24V-rated large truck one - and a whole lot of 9V or 12V adapters/sources, plus solar panels, etc.
Countless times I unsuccessfully asked again.

Suggestion #4: power saving is more important than powerful CPUs.
With the exceptions of signal/video processing and a very few other applications, you won't notice a board being 20-30% faster than the previous model (also:how many currently published projects would actually benefit from a 64 bit ARM instead of a 32 bit one?).
Yet your battery-powered project will perform a lot better if the new board is designed to save 10-15% under the very same CPU load.

Suggestion #5: the faster the boot, the better.
Yes, there's quite a difference between a 30-seconds "power on to actually logging data" time and a 40-seconds one. A small NAND block could allow loading uBoot, kernel and DTBs as soon as possible, without having to wait for all USB and memorydevices to get ready and scannable. An on-board jumper would select "normal (debugging) boot" and "fast (pre-populated and pre-configured NAND things)".

TL;DR: stop being obsessed with filling a large "Specifications" sheet. Do a few things, do them very well. Don't waste milliamperes and don't make the user software wait a millisecond more than strictly needed.

I'd say the ideal boards should be somewhat like:

Model D:
- dual-core processor
- 512 Mb RAM
- one USB OTG port, possibly headers only
- on-board wifi + bluetooth
- microSD storage
- some NAND for quick boot (depending on pricing goals)
- no ethernet/video/audio/camera/hdmi/etc
- four LEDs (powered, heartbeat, disk activity, user) for diagnostics
- 6 to 30V input

Model E and E+:
- same as above, except:
- two USB 2.0 ports (500mA each)
- ethernet
- some selection of cameraport, hdmi, wifi+bt, depending on pricing goals

Model F and F+:
- same as above, except:
- quad-core processor
- 1Gb or more RAM
- four USB ports (including some USB 3.0 if possible)
- ethernet, wifi+bt, hdmi
- camera port, hdd port, power-on-ethernet, eSATA/mSATA, etc, depending on pricing goals and feasibility.

[Min1123]

Hardware:
4 core cortex A55
NVRAM + basic support for video, USB input, USB Mass Storage, and low-level bootloader (a simple BIOS)
2GiB RAM
USB-C power input
USB3 w/ backwards compatible 2/1.1 with independently switchable ports
GPU that has FOSS drivers + FOSS firmware
Gbe ethernet
Expansion header (way to attach daughterboards for additional RAM/(I/O))
RTC + Pins onboard to connect a battery for RTC
Drop HDMI, use Mini Displayport due to licensing costs and compatibility with HDMI and DP and form factor

Software:
UEFI 64-bit bootloader available (see Fedora's aarch64 for RPi3) that enumerates devices for OSes
Raspbian that uses aarch64

[pcxmac]

no more broadcom and a discrete graphics chip.

[L_A_G]

I assume that using a newer Broadcom BCM 7000-series with the VC5-graphics is already set in stone? Because if they can maintain that it should be a decent upgrade over the previous ones.

As for what else could/should be put on the board, the way I see it the most important thing is to provide an improvement without significantly increasing cost. Chasing new use cases by moving to x86, adding SATA IO, etc. may be what people with use cases that they want to use the rPi for, but any additional use cases should always be considered against use cases that are no longer possible or practical because of changes made. Because of that they should absolutely be ready to disregard a lot of any feedback relating to new hardware or very significant changes.

What I'd personally like to see is a move to a slightly more powerful SoC, like the BCM-7268 (currently only really used in Sky set-top boxes) and then removing some bottlenecks like moving at least two of the USB ports to USB 3.0 and the ethernet chip being connected more directly to the SoC rather than trough a USB 2.0 controller.

[leiptrstormr]

U-Boot support, 2GB RAM or greater, hell I don't even care about anything else.

[Unklejoe]

U-Boot support, 2GB RAM or greater, hell I don't even care about anything else.

For what It's worth, I think U-Boot is already supported. I know I have mainline U-Boot running on my Raspberry Pi 3.

[cb88]

I think at some point we will see RISC-V boards outstrip the RPi boards, feature and ease of use wise.

Already at $7.50 you get an 800Mhz RISC-V dual core + 500Top int 8 vector unit... add LPDDR ram support, Maybe A RISC-V core with GPU extensions , wifi/ethernet mac support, USB and HDMI and you have an RPi competitor... and most of that stuff is an IP block that would get dropped into a RISC-V design.

[pal666]

poe+ (802.3at)