Hacker News Reader

Hacker News

Comment by schappim | Hacker News Reader

Parent Full thread

schappim

What's the same:

- Same form factor as the model B+ (your enclosures and daughter boards should still fit).

- Same full size HDMI port

- Same 10/100 Ethernet port

- Same CSI camera port and DSI display ports

- Same micro USB power supply connection

What has changed:

- A new turbocharged Broadcom BCM2836 900MHz quad-core system-on-chip with performance at least 6x that of the B+.

- 1GB of RAM

Source: http://raspberry.piaustralia.com.au/products/raspberry-pi-2-...

View on HN

Under the "Same" column, it looks like you need to add "Same price - $35".

I stocked up on B+s over Christmas - oh well...

The "at least 6x" performance claim should be "at least 3x" by the way, and even that is an estimate, not a guarantee. The 3x is on a single thread comparison, the 6x is on a somewhat synthetic multicore benchmark test.

The Cortex-A7 is also faster per-core, per-clock than the old ARM11 processor, and the clock is higher. So with 4x the amount of cores, some multiplier for the per-core, per-clock stats, and 1.28x the clock, 6x makes some (marketing) sense.

You might get 6x on some multicore benchmarks but for most applications in real life it's safe to say you're unlikely to see that.

> I stocked up on B+s over Christmas - oh well...

Got one for Christmas. Bought a case and a second one since. I knew it'd get outspecc'd someday, just wasn't expecting it THIS soon...

I bought on on Dec-31. Three weeks later, the bundle that I bought dropped in price.

Amazon price-matched it and then less than two weeks later...

Can you return them? If its less than 30 day...?

In the EU, the consumer rights directive allows you to return most goods bought online up to 14 days after delivery.

What's the same: Broadcom is still the non-datasheet-releasing, our-chips-implementing-standard-phy-spec functionality are subject to manhattan-project secrecy, assholes.

Except that Broadcom released "full documentation for the VideoCore IV graphics core, and a complete source release of the graphics stack under a 3-clause BSD license" [1], enough to enable a blob-free graphics driver. Code has already been committed to Mesa [2].

That is a lot more than you can say about the PowerVR core in the Beaglebone advocated by another response to this post.

[1] http://www.raspberrypi.org/a-birthday-present-from-broadcom/ [2] http://en.wikipedia.org/wiki/Free_and_open-source_graphics_d...

Beaglebone can be at least booted without non-free blobs, so it can be 100% free if you don't need to use 3D acceleration. Raspberry Pi cannot - even now with free GPU drivers, you still cannot get Linux working without providing non-free firmware blob to it first.

Of course you're referring to me. Arguing the whole openness of the video driver is a dead end [1]. If you want an open video driver use Intel, or AMD and be done with it.

Give me a hand-cranked universal Turing machine and I've basically got the same device as a RPi. The only difference, it's cheap, and I don't have to get a workout hand-cranking it. Video is just a (semi-)fixed functional unit. Sure, it'd be nice to have open drivers, but so would having the RTL of the processor, and the netlist and gerber of the board. Then they could actually talk about teaching kids something they actually couldn't learn with just having access to a hand-cranked universal Turing machine.

The fact that the foundation and Broadcom get their free goodwill from their claim of "for-the-sake-of-the-children" is the real source of my ire. Look at how many references to "industry" the article quotes from representatives of the foundation. Like the embedded SOM has anything at all to do with education, or the continuance of production of older models. Please...

[1] http://m.youtube.com/watch?v=eZd0IYJ7J40

Do you think it is feasible to release a similarly priced, similarly capable device with completely open hardware?

The RPi has sold immensely to home users, but also in education. The foundation is also open to criticism and has shown in the past that they are willing to change their production when given the opportunity. They have moved most of their production to the UK for example, whereas the first boards were all manufactured off shore.

It would be difficult in the near term. With lowRISC (http://www.lowrisc.org), some of us are working towards a complete open source SoC (i.e. the RTL is available under a permissive open source license). Although our longer term goals include being an acceptable SoC for a phone or tablet, the first iteration won't feature a GPU.

As much as they are "open to criticism", I very strongly doubt that they are going to use anything other than Broadcom.

Do you think it is feasible to release a similarly priced, similarly capable device with completely open hardware?

https://www.olimex.com/Products/OLinuXino/A20/A20-OLinuXino-...

Certainly not as widely known as the RPi, but I'd attribute that more to a general lack of marketing/"presence" in education.

"... and dual-core Mali 400 GPU"

Do you really claim that the Mali 400 GPU is more open than the VideoCore IV?

That doesn't meet the (absurdly high) openness standard of "RTL of the processor" mentioned by throwaway000002.

My, overly-impassioned, exaggerated point is that the RPi is no better that many other competing devices on the market and should me measured solely on technical/business merit. The popularity of the device has spawned a product category of itself. However, at this point in time, claiming something about education is total rubbish. Take the old Pi, run ARMv7 QEMU on it et voila you have the new Pi. There is nothing new to learn, to grow, to probe questions about. The GPIO of the device could just as easily be be provided by a USB breakout board, etc, etc. Cypress had a $4 PSOC prototype board and if anyone (especially Cypress) bothered to develop open-source tooling for it, it would have be a far more useful device to learn from.

(As an aside: thank you to HNer asb and colleagues for the lowRISC project. Also, thank you to people like "bunnie" and "xobs" for Novena. These are real people, with real projects working with the best of intentions.)

I think you've missed what layer the education is targeted at: software, especially Python but also Scratch, SonicPi, and so on. The hardware stuff is at the "turn on LED with GPIO" super-introductory level.

Edit: Eben Upton is also a real person?

Eben Upton and the foundation should rightly be lauded for the RPi Model A. That was 3 years ago. What purpose does this update serve? If it's the "developing world", then you haven't made the device any cheaper, and more functional competitors at similar price-points have existed for at least a year. As for the "developed world", I could have bought a Chromebox for $150 on Black Friday. I'm sure a school could easily negotiate that price and have a device that easily has >6x the functionality (and far more open thanks to tireless work of the coreboot team and Googlers working with Intel on minimizing the closed bits).

What again is the point of the RPi2? What new thing does it bring to the table that an at most $5 set of GPIO pins, in the physical form of the Pi connector, off a USB plug with an API can't bring to a basic PC?

The core target is unchanged - kids in the UK. The exact functionality level isn't nearly as critical as the standardisation of the system and its community support. Updating it keeps it competitive so people don't fret about being stuck on old technology.

What a Pi brings over a PC is ownership, stability, and freedom from the usual risks of breaking a PC. You can fiddle with any of the software on a Pi and always get it back by replacing the SD card. PCs are vulnerable to tinkering and inadvisibly installed software.

This conversation reminds me of http://slashdot.org/story/01/10/23/1816257/apple-releases-ip... : "No wireless. Less space than a nomad. Lame."

It pleases me to see you mention the PSoC 4. I think it's a highly underrated educational tool due to the software stack used to develop on it. If they were to open up the platform I know several places that would love to use it when teaching kids about hardware and software development.

To respond to jvdh, yes it is possible. Others have mentioned the more capable Allwinner based Olimex devices.

And if this really is an issue of "Buy Britain" why isn't the device based on Bristol based Xmos silicon. Now there's something that's a really neat platform.

In short, the RPi is an inexpensive enough device that a lot of hobbists can buy on a whim and proceed to install and reinstall various flavours of linux on to marginally satisfy their sense of somehow being "hackers". As I said in my initial post to this announcement, the work of the foundation in making a cheap computer a reality has been totally ground-breaking.

Now everything has been turned into at best a marketing exercise. Now there's news of RPi2 as a Windows IoT platform, with requisite Windows machine for dev work. Will wonders cease to amaze...

I started out when you had to pay hundreds of dollars to get even a C compiler (for MS-DOS). I amused myself by hex editing COM files (because I figured out that for some reason it worked better than EXEs when randomly changing stuff) and using "debug" to assemble simple programs. I wanted to make EXEs and I didn't know how, and MS BASIC didn't make them for reasons I didn't understand. I didn't know where to get MASM that the few public library books I could find were on about. Then I got on a BBS with a program called Telix and got something called Turbo Pascal, and my life kinda changed. Writing directly to segment B800...

And then a few later Linux came on the scene and I discovered the magic of having GCC just there waiting for you. But this had to happen at school, after school hours because I had a 8088 at home, and I needed a 386 which thankfully my teacher allowed me to set up a dual boot for...

I will _never_ _ever_ allow computing to be made inaccessible to my future generations to understand and learn from. I don't want to drone on about this issue, but proprietary is fine. Making money is fantastic, necessary, and a beautiful indication of the value you bring.

But putting people in a prison of no understanding is not at all okay.

The irony is that the Pi comes from someone who has exactly that kind of childhood hacking background who wants to replicate it for everyone else. It's the opposite of a "prison of no understanding". Holding out for free-to-RTL hardware is allowing the best to be the enemy of the good.

They may be changing a little bit, slowly. They're not used to this sort of thing, at all.

This is actually a big step: this wasn't just SoC repurposed for the Pi, this SoC was created for the Pi2.

There are a lot of details missing about the VC4, that's true. They could still publish a lot more of those than they have. We should be given enough to start the ARMs, warm the caches, talk to the devices and manage, start a kernel, even without video. The architecture means that the GPU plays the host role, however, so Linux is pretty much postboxing things to it asking it to do things when it comes to the devices - this complicates things, because that means the GPU firmware is probably going to have to be a bit more than a simple shim to get by. They can't just open-source all their side of the firmware either - it's based on a commercial RTOS, ThreadX. That's why they hired a guy to do a rewrite.

Arguably still better than NVIDIA. <g> Also better than MediaTek and the like - the usual way you get anything out of them is someone steals it (back).

Open or not... I want a working DSI or I'll go with something else with eDP.

Why are they required to release those datasheets?

They're not, but it's been common practice in their industry for decades, and it's generally seen as beneficial to OEMs and users to have access to datasheets before purchase. This is why it's also commonplace for most chip manufacturers to offer free samples, even of chips upwards of $20 list price. Most of them even publish enough information to develop hardware and software to interact with these chips even without one on hand.

That Broadcom doesn't do this is very atypical.

Ok. I find it rather rude to call a company 'assholes' for doing something that is 100% within their rights. They're releasing a chip that powers a relatively cheap fairly powerful computer, have supported the effort in as many ways as they could without giving up their competitive edge (when you're in a leadship position it is sometimes quite risky to open up too far), nobody forces anybody to buy the product and lastly you can of course always fire up the disassembler and do it the hard way.

I think aswanson is simply expressing his feelings. Broadcom has garnered goodwill from being involved with the Raspberry Pi foundation. Goodwill, I must add, from a broad swath of people who are not too technically inclined. Yes, the computer (or device if you're lacking generosity) is a god-send for those without the capacity to buy an Intel/AMD-based PC, but let's not kid ourselves: the device is less open than a PC was even 20 years ago. Which begs the question, in what capacity is the foundation moving computing forward? That's where I lose goodwill even for the foundation. Sure, they did something special, that really made compute affordable. But that was years ago. Now they need to make it open, so anyone, not just students, can answer probing questions about how the device works, and in so doing be able to make a Raspberry Pi device, or something totally different, for a whole new generation.

All this said, if you want cheap and software open, go with the Odroid C1 (China really is the future, and in this case a Korean firm is acting as the intermediary to Chinese silicon.) And if you want fully open, TI's Beaglebone. Good, solid, open, American.

> All this said, if you want cheap and software open, go with the Odroid C1

According to http://www.hardkernel.com/main/products/prdt_info.php the Odroid C1 uses a Mali-450 MP2 GPU. This is at least as closed as the VideoCore IV used in the Raspberry Pi.

So is it possible to get an Odroid C1 from Hardkernel now? I put in an order for one in December, and they canceled my purchase and said they stopped shipping to the US due to reports of missing packages. They said to order from ameridroid.com, which isn't served over HTTPS, and when you get to the checkout, you're taken to https://04622216-5b1d-11e4-8b60-14feb5da1938.mysimplestore.c....

"We designed our storefront to be indistinguishable from something resembling an sslstrip attack in action" leads to an "I guess I'll have to do without, then" response.

The Odroid C1 uses an SoC from Amlogic, which is also American.

Mea culpa. I see their SOCs on all these Chinese media-boxes that I just assumed they were Chinese. How did an American company break the Chinese market like that? Impressive! I should do research on who's running that ship...

Whatever they are or aren't doing on hardware openness, Broadcom are doing a lot for open software with the Pi platform. That's quite a lot there, anyway.

Dead on. Those pricks cost me about 5 weeks of progress at a startup because they didnt provide drivers for linux or a register set, much less a description. All because some clueless hardware guy used their chip because it was 'pin compatible'. Unless you are Apple or Dell, do NOT do a design with their chips.

You are 100% within your rights to not tip at restaurants in the US and yet I believe most people would consider you an asshole. If someone does something that benefits themselves at the expense of others when it's socially assumed to act in another way, many would call a person an 'asshole'. Seems like a similar situation if it is normal practice in the industry.

When somebody pulls a dick move that they're totally entitled to pull, they're being an asshole. When they're not entitled to, we call it a crime.

Maybe I'm missing something, but I'm not sure what a disassembler has to do with datasheets?

I think he means that if no datasheet is available, you can try to reverse engineer it by using, say, a disassembler.

But a dissembler doesn't produce the same information that a datasheets provides? Unless he means to physically disassemble, but that isn't very easy.

Indeed. But perhaps enough to write an (incomplete?) open source driver?

I can see how you might think that seeing as how a lot of manufacturers in the low cost and hobby markets are quite liberal with their datasheets . but it is not uncommon in regards to the higher end soc companies like marvel,qualcom or in this case broadcom.

"Low cost and hobby market" companies like Intel, TI, NXP etc..?

they aren't required, but without it, the few hobbyist trying to make Linux work on it will have to devote 10x hours instead of x. that is all fine for broadcom. but you're doing a disservice to those developers and to you by supporting and popularizing this board, because if you supported a open one you would now have 10 chips with Linux support instead. or some other feature you are waiting or something so cool and futuristic that you can't even imagine will never happen because those folks were doing the tedious task of supporting a closed platform that unfortunately became popular.

They're not required, but (thankfully) we're not required to use hardware from vendors that don't do that.

There might come a point where essentially nothing on the market is open anyway, however. Meaning we are required.

What about power consumption? I know it is the same power supply (5V, 1A), but this doesn't mean it uses the same amount of power...

The official power supply has 2A.

Will Mathematica still ship with it and will the 6x performance increase make it be actually usable? What's the per-core increase?

It will continue to be available, but I'm not sure what the performance increase is. There'll be a blog post on walkingrandomly.com coming in a few days - a guy who benchmarked it when it first came out. See http://www.walkingrandomly.com/?p=5220

Real-world max for single-thread speedup is probably the 3x quoted for the cache-busting SunSpider benchmark, with around 1.5x being more common. In terms of issue width, the new core is only a modest improvement as the Cortex-A7 is largely single-issue.

here's a comparison with the correct information - https://learn.adafruit.com/introducing-the-raspberry-pi-2-mo...

"... Best of all, the Pi 2 keeps the same shape, connectors and mounting holes as the Raspberry Pi B+ ..."

Phil, does this means the case (model A+) used in the 'Touch Pi' will be modified to fit the Pi2 at some stage?

cf: https://learn.adafruit.com/touch-pi-portable-raspberry-pi & https://learn.adafruit.com/3d-printed-raspberry-pi-b-plus-en...

Reply on news.ycombinator.com