Eben Upton: Introducing the New Raspberry Pi B+!

I like the Raspberry Pi because it provides adequate processing power for most things.  There is enough horsepower and memory that a new user is not immediately plunged into stripping down unneeded facilities of Linux to make room.  Most of all, I like that the foundation has kept a strict lid on the price, and keeping the price down is what makes it reasonable to have one per student and reasonable to mess with the hardware in ways that carry some risk of damage.  For the whiners who are always crying for more power and more memory, I'd say write better programs.  If you can't fit your code in 512 meg, maybe you belong in a Windows environment where the OS can't seem to run in less than 8 Gig.  Put another way, would doubling the power really satisfy you?  Could you live in twice the memory?  Or would you just be whining again that it's dual core, why isn't it quad core?  It's got a gig, why doesn't it have 2 gig?  If you want more memory and power, there are other solutions; other more expensive solutions.

condew , I disagree with some of your statements...

If you can't fit your code in 512 meg, maybe you belong in a Windows environment where the OS can't seem to run in less than 8 Gig

Unfortunately the 512mb of RAM is shared between the systems processes and video RAM. This means that it isn't, in its entireity available to programs coded for the Raspberry Pi. Further more, unless you're coding 'bare metal' then the operating system will use up a good portion of that RAM, unless you roll your own Linux distribution for it - and I bet the majority will be running Raspbian.

Or would you just be whining again that it's dual core, why isn't it quad core?

Being at least dual core would be an acceptable trade-off to cost, if you're running an operating system on the Raspberry Pi alongside your code then you have the issues of not only interrupts raised by components of hardware on the board wanting attention but the management of the operating system itself using that processing speed. If you have at least dual core then one core can handle the operating system tasks while the other can run any code you create.

A stock installation of Raspbian for example, is slow and sluggish; though this is in part because of lack of graphical acceleration because of a lack of a suitable driver for it at present.

There are more expensive solutions, and I'm all for using an appropriate hardware solution for a relevant problem but frankly, it just isn't the same as using a Pi.

There isn't really any way to beef up the MHz, RAM or number of cores while still keeping it compatible.

The chip they have chosen is ARMv6 and it represents just about as far as ARMv6 ever went, if you want more processing power you have to go to ARMv7.

You'd certainly have to go to ARMv7 to get dual core.

As I understand it the current chip they use won't take more than 512MB RAM.

The Pi is limited to 515M because nobody makes a compatible 1G, and we won't be changing SOC because there's a lot invested in code that's been proven on this SOC.  I'm fine with that, but it you read the comments on the Raspberry Pi blog about the B+, there were so many griping because they didn't get a pony.  The Raspberry Pi is a lot of machine for $35, and the B+ means you don't need a USB hub to go with it for the nearly standard keyboard, mouse, and WiFi dongle; plus lots of other nice corrections.  Economies of scale could bring the price down, but I'm glad to pay the same for a better product, and help the foundation fund their education-related efforts.  The only hope I have for the near future is that the A+ will have 512MB and not 256 MB.  If an application is tested and works on a B, but doesn't need the ethernet or more than one USB, if would be nice to use an A to put it into use without worrying that it might not work as well with half the memory.

Too many programmers I talk to now think efficiency is of no consequence, and so we have programs that can't perform reliably even on mid-level machines even with more than 4 gig of memory.  But their employers provide them with the fastest machines to develop and test, and they have the arrogance to believe that anybody who wants to use their product should just go out and buy the fastest machine, regardless of the cost.  Starting on a low power machine should help make young programmers aware of efficiencies of time and space as a consideration in any good programming solution.

Evan Phillips wrote:

 

Too many programmers I talk to now think efficiency is of no consequence, and so we have programs that can't perform reliably even on mid-level machines even with more than 4 gig of memory...

Amen.  One of the reasons I was an early RasPi supporter is that I had hoped that RasPi's limited memory and processing power would encourage new programmers to care about efficiency.  In fact, I felt that going to 512MB was a step in the wrong direction, but it may be that 512MB is only marginally more expensive than 256MB.

I've found that a 256MB RasPi runs my 60K-line GUI application quite well.  OTOH, it's nowhere near enough for a wasteful resource hog like most browsers.

I prefer to develop on limited-performance hardware, since it exposes inefficiencies in my code earlier on.  But then I got started in the punched card era, which pretty much limited your programs to 2000 lines (if you wanted to carry it under one arm) or 4000 lines (if you wanted to lug around two boxes).  Program size was very real back then.

Evan Phillips wrote:

 

Too many programmers I talk to now think efficiency is of no consequence, and so we have programs that can't perform reliably even on mid-level machines even with more than 4 gig of memory...

Amen.  One of the reasons I was an early RasPi supporter is that I had hoped that RasPi's limited memory and processing power would encourage new programmers to care about efficiency.  In fact, I felt that going to 512MB was a step in the wrong direction, but it may be that 512MB is only marginally more expensive than 256MB.

I've found that a 256MB RasPi runs my 60K-line GUI application quite well.  OTOH, it's nowhere near enough for a wasteful resource hog like most browsers.

I prefer to develop on limited-performance hardware, since it exposes inefficiencies in my code earlier on.  But then I got started in the punched card era, which pretty much limited your programs to 2000 lines (if you wanted to carry it under one arm) or 4000 lines (if you wanted to lug around two boxes).  Program size was very real back then.

But then I got started in the punched card era

Pity someone never really expanded on the 'hole' part for the program information ... much easier to carry holes around.

There is certainly something to be said for limited size, but then we do live in a graphical world.

Mark