Hacker News Reader

Comment by jandrese | Hacker News Reader

I wish Apple would make a sort of Raspberry Pi like board with an A series chip, ethernet, HDMI, GPIO, SD Card, etc... for $50-$100 that runs Linux.

I would also like a flying unicorn.

Seriously, a high power SBC at a reasonable price point would be insanely useful for projects like portable MAME cabinets, video transcoders, scene displays, etc... I love the Pi but it's so old tech, especially the video cores, that it's annoyingly limited for some applications. Worse, the competitors with faster cpus often have worse video acceleration support, making them worse for this than the underpowered Pi.

View on HN

Not a flying unicorn, but still a winged equine: the Mac Nano.

A tiny white puck with a single usb-c plug for power and data. Inside, an A12 running OS X.

Education and “enthusiast” market. GPIO as a dongle, of course.

They can release the A13 version in colors the next year :)

Apple has been targeting the general public filling their needs with easy to use device/software stack and while part of it is starting to fail (dropping AirMac series), I would consider them losing focus if they start to build tiny machines only targeted at geeks.

But I do feel that without extending their business what they're going to do with their massive cash just sitting around doing nothing.

Doing nothing is far better than making a dumb acquisition, and most acquisitions are dumb. Just buy back stock, they have plenitude of cash flow to fund reasonable R&D.

I think you mean Airport

It’s called AirMac in Japan

Take my money

Have you tried developing apps for the Apple TV? It's basically like that; just too low-powered to actually run macOS.

It most definitely is not. Not anymore. It's just as capable as recent Mac minis. A10X Fusion is a very decent chip.

I'm confused as to how you have a computer with only a single port... what do you imagine plugging the one port into?

A USB-C Thunderbolt Display that delivers power over the cable and has a USB hub built into it. These exist, though they're expensive.

Or, an eGPU enclosure, which also delivers power over the cable and has a USB hub built into it. (You just need to plug a monitor into it.) These exist, and are surprisingly cheap if you buy them bundled with a graphics card.

Not all Mac USB-C ports are Thunderbolt, though...for instance, in the MacBooks.

Previous MacBook 12” user here. Either power, or peripherals. Just not at the same time.

Macbooks have large batteries. Wouldn't make sense to have one in this if you're going for as small as possible form factor.

For most of the day my Macbook Pro is basically a computer with one port plugged into a dock with its case shut since I like using my LG 21x9 screen more than the built in screen.

Do you have a link to a good guide on how you can use a MacBook Pro with the lid closed all the time? (Obviously with external display[s].)

I tried 2 times -- granted they weren't very honest attempts, I hoped to have it done in 5 minutes, but failed.

I am looking for a no-BS quick guide if you have one lying around.

1. Make sure an external keyboard and mouse are available, wired or wireless is fine

2. Make sure your MacBook is plugged into power and an external display

3. Close lid

4. Your main desktop should show up on the external display. If your computer goes to sleep you can wake it up by clicking the mouse or typing on the keyboard.

Thanks.

Will that wake the internal display though? That would warm it up and that can be pretty bad when the lid is closed.

If the external display wakes up as your main display with the same desktop picture and icons as your laptop it means the internal display is asleep.

If your mouse is laggy and display looks fuzzy it means your external is mirroring your internal display and that your internal display is on. Turn mirroring off if that happens.

If you’re worried about heat (lid shut mode blocks the vent that runs the width of the keyboard) you can leave your lid open and use your external display normally.

I have not noticed a heat problem using my 2013 MBP Retina in lid shut mode after my two year old threw a phone at the screen (kids! Lol)

Right but your Macbook has a large battery. Something it wouldn't really make sense to have if you were going for a very small form factor.

Remove the battery and screen, you get the Mac equivalent of the Intel Stick. Depending on price, I could very much see buying a lot of them for the students (with a USB-C Digital AV Multiport Adapter or equivalent for their home use).

[Edit]I would then stock the labs with docking ports like something from CalDigit and LG displays.

Power I'd guess, since he didn't specify a battery. The display would be wirelessly beamed directly into your brain.

Apple chips famously have Imagination GPUs that have zero open-source support, a far ways off even from Nvidia. They are the absolute worst.

That's before you consider that all the stuff Apple did to that poor ARM probably means you can't even boot Linux into a terminal on it today.

>Apple chips famously have Imagination GPUs

Not anymore, Apple switched to in-house GPUs recently.

That makes it worse? The Imagination crap is shipped on Android phones, too, so there are at least vendor kernel drops to work from.

The odds of Apple releasing a development ARM board is probably roughly equal to that of them releasing complete documentation and linux kernel driver support to go along with it.

The vendor kernels are worthless, tainted, piles of crap. Vendors rarely, if ever, maintain them once they are posted publicly. The GPU still requires proprietary blobs to be even remotely useful.

Of course. But they contain non-zero amounts of information on the chip so you can do a proper implementation or, much less ambitious, update it for newer kernel versions.

Not in my experience. The code in the kernel is usually just a shim for loading the blobs that do the real work.

All the reverse-engineered GPU drivers we have now (freedreno, etnaviv, nouveau, lima) started out with intercepting kernel ioctls, being able to observe the blobs at work. Many of them at one point ran with the vendor kernel interface but open-source userspace.

In many ways the kernel part of a GPU is arguably the least important part and it's the userland where the shader compiler sits and produces the magic, but if you start from nothing it's very useful to have a working kernel interface that initializes the hardware and accepts command streams to run.

This is marketing hype.

Apple’s commits to the LLVM (where the shader gets compiled into GPU executable code) states the A11 GPU target is “PowerVR Rogue” ISA. Which is identical to how it compiles for the A10.

So it’s still using foreign IP for the instruction set. Maybe Apple moved things around/optimized internals? Get better path optimization, or improve IPC?

But it’s weird they didn’t add optimizations specifically for that A11. So an A10/A11 GPU just run the same opcodes.

Citation for those commits?

Apple's GPU is designed just down the road from Imaginations HQ. Guess where the designers previously worked...

It's a completely locked down system - does it matter if the GPU isn't open source?

Maybe I misunderstood "portable MAME cabinets, video transcoders, scene displays" and the comparison to the RPi but surely the intention is to run your own code on it from the bootloader up?

Ok, but Apple devotes considerable resources to make all of that impossible. Why pick on the GPU specifically?

Worse is the array of smart devices that are forced to use Android type hardware / Android OS. Looking at the Peloton cycle I have specifically which clearly runs Android OS. Could be so much better if it was on faster hardware / could link up to Apple Watch for stats.

Make a version of iOS that companies can build off of; maintain the OS patch cycles but let them put their own app / front end on it.

Tons of smart devices exist now a days; TVs, refrigerators, bikes, etc... There's a market for an Apple based system for sure.

But yes - I'll taking a flying unicorn also in the interim.

That's an interestingly spun perspective. So you're asking for Apple to produce an open source ecosystem around their hardware and OS, while simultaneously refusing to use the products of the company that did?

No devices are "forced" to use "Android type hardware / Android OS". They use it because it's free. That's... important. Right?

Android isn’t “free” in the FOSS or Beer sense of the word.

The important bits (hardware drivers and core applications) aren’t open source (AOSP doesn’t include the new default apps or standard libraries.

Also, the important stuff is tied to google services that spy on you.

I think it’s completely reasonable to reject that stack, and lament that no one offers a high quality, supported alternative software stack at a reasonable price.

I hope and I wish the market and Android will force Apple to make this decision within the next 3 to 4 years.

They are still not making Apple TV set.

With WatchOS and the Apple S3 chip already includes 90%+ of the features needed. I bet the S3 cost less then $20, and Apple could have sold it at $50 along with WatchOS to specfic market that Apple do not intend to compete. Fridge, Fan, Lamp, bikes, TV set, radio, kitchen appliance or what so ever.

Apple currently charges $1 for the controller of Lightning Cables regardless of volume. They could have just offer similar program.

I would love this. They could also push swift with it, so I'm surprised they haven't done this already.

Why would that be good for Apple?

It would be good (and bad) for the same reasons it was for microsoft + intel back in the ‘80s.

> Raspberry Pi like board with an A series chip, ethernet, HDMI, GPIO, SD Card, etc... for $50-$100 that runs Linux.

The upboard fits that description perfectly - http://www.up-board.org/up/

The CPU is an Intel Atom which turns out to perform very well. The internal storage is blazingly fast eMMC. It has an Intel GPU which means complete open source driver support. (I can't comment on the GPU performance as I use the serial console.)

They also put a 40 pin connector on it that is hardware compatible with the raspberry pi - ie the same pins, voltages, and functions. You do need kernel patches for full support of the pins because the pins are specced at 3.3V while the Atom is 1.8V. There is level shifter hardware that fixes that, which the patches control. The patches are public from Emutex (in github) and there is also a PPA if you just want to play with Ubuntu and not worry about it.

It uses an AMI UEFI BIOS. You can compile the kernel with efi stub support and the upboard can boot straight into your code.

Your best bet would be a SBC with i.MX8M - the same SoC that will be in Librem 5. That's because it's one of the best supported SoCs by mainline Linux kernel.

With Etnaviv drivers for GPU you can be pretty much covered. And I think it will get better.

I'm hoping that someone will make a netbook with i.MX8M. Or maybe even Chromebook - it would be then easily repurposeable.

I'm eyeing WandPi 8M [0], armstone mx8m [1] and Nitrogen8M [2].

[0] https://www.wandboard.org/products/

[1] https://www.fs-net.de/en/products/armstone/armstonemx8m-with...

[2] https://boundarydevices.com/product/nitrogen8m-imx8/

Releasing a chip like that would be close to useless without also developing a user-friendly hobbyist API like the Raspberry PI and Arduino's already have.

Releasing the chip plus a nice users manual of all of its registers and assembly instructions might pose a security threat for device jail-breaks that they probably aren't interested in either.

Seems to take about 4-5 years from release for the depreciation&upgrade curves to bring an iPhone under $100. At that point, the main thing keeping them from being what you're asking for is:

1) No one has been particularly dedicated to developing an open-source OS for these things, despite the fact that used phones are steadily becoming some of the most common devices in the world, to the point where we have to treat them like waste. 2) Apple's treatment of ports of any kind as a blemish.

Solve these problems, and you have the board in a slick package. :)

I'm less than half kidding. #2 is obviously solvable, as there are lightning dongles/docks for just about everything on your list. #1 gets harder over time with security measures but it has to be possible, it's a question of resources.

So basically a PC with GPIO?

NVIDIA Jetson is the closest you can get in terms of performance, but the price is closer to $500.

DX2 specs:

GPU: NVIDIA Pascal™, 256 CUDA cores (1302MHz)

CPU: HMP Dual Denver 2/2 MB L2 + Quad ARM® A57/2 MB L2 (1.4Ghz/2Ghz)

Video: 4K x 2K 60 Hz Encode (HEVC) and 4K x 2K 60 Hz Decode (12-Bit Support)

Memory: 8 GB 128 bit LPDDR4 (59.7 GB/s)

Display: 2x DSI, 2x DP 1.2 / HDMI 2.0 / eDP 1.4

CSI: Up to 6 Cameras (2 Lane) CSI2 D-PHY 1.2 (2.5 Gbps/Lane)

PCIE: Gen 2 | 1x4 + 1x1 OR 2x1 + 1x2

Data Storage: 32 GB eMMC, SDIO, SATA

Other: CAN, UART, SPI, I2C, I2S, GPIOs

There is video story at all (except a serial console), but the apu2 amd-based boards are nice, stable targets, and much, much faster than a rpi for networking applications:

https://pcengines.ch/about.htm

*TX2

Also it has a student discount to only $300 which imo makes it much more compelling if you're eligible.

Still, only 1.5Tflops (roughly 750Ti level), which makes it insufficient for bleeding-edge robotics, but 8GB is nice for inference. NVidia has some high-performing board for automotive, but that one is 500W for a change...

I'm right there with you. There are some pilikes like Espresso Bin that are much more powerful, but they just don't have the community support that Pi does. http://espressobin.net/

Still at $49 it's cheap enough I'm going to see if it'll work as a wall-of-ceph appliance. Would love to have an SBC designed to be an OSD node but Western Digital never put theirs in the channel and there's not much else comparable AFAIK.

Apple is so in love with walled gardens. I can't imagine them ever making truly open hardware. Perhaps they are uniquely qualified to make something open because of the amount of effort they have put into making things that are closed.

It's not that they love walled gardens, it's that they have different priorities than other companies. Users and usability are their primary concerns. Everything else is done to support that.

They have open-source initiatives, they like to be open like that when it doesn't interfere with those mandates. It's just that usability is such a huge thing that it impacts every step of their process.

Maybe something targeting schools/education?

> Seriously, a high power SBC at a reasonable price point would be insanely useful for projects like portable MAME cabinets

Right now one can buy an Exynos5422, which CPU wise (I've never tested the GPU) is 4/5+ times as fast as an RPi 3b, for the price point mentioned (<100$, board only).

For me, this is effectively a "high power SBC at a reasonable price", however, at the same time, users interested in SBCs (or micro-PCs) seem to have a wildly different array of requirements (I've read virtually everything: (very) low price, GPU, kernel support, open source drivers, arch, ports, form factor, etc.etc.), so a desirable SBC seems like a pipe dream due to the fuzziness of the requirements.

(For reference, the CPU performance ratio is calculated on single-core performance of a generic RPi 3b core vs. a fast Exynos5422 core, on OpenVPN maximal bandwidth).

Link/source ?

Btw imho the worst part about RPi is that it has poor ethernet performance. This improved a bit with RPi 3+, but still meh.

Here is a Phoronix test suite run:

https://openbenchmarking.org/result/1805276-FO-1703199RI52

It's hard to asses a generic ratio, since the performance increase is wildly variable. I think it can be reasonably assessed a "base" 3x performance ratio against the RPi 3B (with an outlier of 8x).

They'll fix it in the RPi 4 and then you can complain they don't have 10Gbit support, so at least you have those two things to look forward to.

> I wish Apple would make a sort of Raspberry Pi like board with an A series chip

Likely outcome of the unlikely discovery of some lost ancient document that somehow makes The Woz indisputable CEO for life.

Someone recommended in an article an ODRIOD-C2 over a Raspberry Pi. They said you get quite a bit of performance spending just a little more ($40). Here's a comparison:

https://www.phoronix.com/scan.php?page=article&item=raspberr...

I might buy one depending on what feedback I get on it. Anyone here tried those out for Linux or BSD projects?

Surely there's got to be something like that available already. I'm mostly dealing with low powered M0/M4 boards, but the selection in this range is massive.

There's an absolute ocean of old obsolete ARM cores clocked slowly. The difference in performance is so vast it would be hard to compare with an A11 chip, especially once you include graphics acceleration in the mix.

Apple has been doubling down every year on the "we can make it go faster", and the rest of the ARM chip makers have responded with a "meh". It's so bad that people get really excited for a Samsung chip that is no slower than a two generation old A series chip.

Performance is highly dependent on task and graphics is extra tricky to compare, but looking at geekbench, it seems the Samsung S9 is about halfway between iPhone 7 and 8/X. So half a generation behind is more accurate.

Well, the price is wrong, but nvidia's Jetson platform is very powerful compared to the *board SBCs:

https://www.phoronix.com/scan.php?page=article&item=march-20...

Jetson CPU performance is on par with chips like the RK3399 or the Exynos5422, which are 4/5+ times as fast as the RPi 3b (I owned RPi 2/3b series, and the Exynos).

Note that the first benchmark in the mentioned article is an outlier.

No, it's not. Having worked with the Jetson platform extensively, the TK1 was equivalent to a 5422 (because they're both built on the vanilla A15 architecture). However, the TX1 and TX2 ("Denver" cores) are closer to (though, still slighty behind) the current SD 835:

https://www.youtube.com/watch?v=l6y4atVl-mc

And when it comes to GPU compute, it's much more powerful than all the alternatives.

I'll correct to "comparable", as it's more appropriate.

However, it's not a "clearcut no" or "very powerful in comparison", due to the number of cores (no doubt about the GPU, I had made this explicit in the parent post).

I've run the Phoronix test suite (which uses a set of real-world applications)[⁰], and, excluding the outliars for both parts (Redis/OpenSSL), we're talking about 10 to 40% advantage for the TX1, and little more for the TX2; XU4 has even the edge in one case.

They're definitely faster for the majority of use cases, however, I wouldn't classify them as "very powerful in comparison", at least, when considering the performance of an RPi 3B.

[⁰] https://openbenchmarking.org/result/1805276-FO-1703199RI52

The TX2 is a dual-core chip (ignoring the 4 "littles" since they're unused in benchmarks). The 5422 is a quad core. And even then, it's outperforming the 5422 by a decent margin in most tests. Even c-ray, an intrinsically multithreaded application is close.

So you're correct, they're "comparable" in that if you take twice as many 40%-as-powerful cores, you might shrink the gap. But for any fundamentally single threaded tasks, a massive gap exists. And if/when you leverage the GPU (which can't be ignored, despite your handwaving), it becomes a canyon.

Let's call this product "iPhone mini". With a good name for it, I'm sure we can convince Apple to do it.

I find it's a pity there's no Mac mini anymore, at least not a powerful one, with dedicated GPU, a kind of NUC Skull Canyon.

I still run a 2008 Mini with Manjaro Linux..still good for Kodi

Have you looked at raspberry pi clones built around FPGAs?

I didn't even know these existed! From a googling I can only find hats/daughter boards for existing raspberry pis. Is that what you mean or are there ones that are all in one?

I own a parallela with a Zynq1020 fpga and a form factor similar to a raspberry pi.

Here is somehing that costs around a hundred dollars (a bit higher) called Zynqberry.

https://shop.trenz-electronic.de/en/TE0726-03M-ZynqBerry-Zyn...

http://papilio.cc/index.php?n=Papilio.Hardware

It's not a pi clone, but google snickerdoodle board.

Like Apple TV?

The rock64 and variants should do what you're looking for.

I said I wanted a working GPU.

It does have a working GPU. There is the RK3328 specific libmali: https://github.com/rockchip-linux/libmali

And the docs: http://opensource.rock-chips.com/wiki_Graphics

Odroid N1 is getting close.

Reply on news.ycombinator.com