This post is a response to the following blog entry by Jeffrey: infoblog.li (Those who don’t understand German may use an automatic translation tool…) It talks about whether IPFire will be running on the Raspberry Pi and that the RPi would be a possible replacement for the Alix boards which are quite weak.
I personally think that the Raspberry Pi board can not satisfy what users are expecting from it because of the following reasons:
Ancient SoC
The SoC that is working on the RPi board is old. I mean really ooooold. It is from the ARM11 (ARMv6) generation which was released in 2002. The computational power is not bad if you consider the power consumption, but the actual performance that is needed to run Quake 3 and some other demos is provided by additional components like the GPU which can decode high definition videos and some more fancy stuff.
Lots of people dream about tiny server systems that run web services at home or things like that but those do not make benefit of a fast GPU. They need CPU power and RAM which is not sufficient for modern software on the board.
Not enough NICs
Using the RPi board as a router is not a very common use case so I cannot blame the developers for not adding a second network interface to it. They really need a lot of physical space and have a high power consumption, even if they are unused.
But another big problem for us is that most of the SoCs do not have the controller for the NICs built in. The reason for that is that they all are targeted to run on mobile devices which do not need an Ethernet adapter. Thus, the LAN ports are connected to the USB bus which causes very poor performance.
The only exception is the Globalscale’s Dreamplug which has 2 GBit ports directly connected to the platform bus when the Pandaboard, Beagleboard and RPi connect an 100M port to USB.
To clarify your question, Jeffrey: Adding another USB adapter which adds a second NIC is totally fine and should work exactly like the onboard NIC.
The Raspberry Pi computer is currently not available
The echo of the so called launch was overwhelming. Most of the people who heard about it thought to give it a shot and tried order one on the release day. Shops broke down, you all know the story…
All that is comprehensible. The board is about 30€ and that would be worth a shot. I already mentioned that there were a lot of things one could do with such a little computer coming up soon. But until today, all have been waiting to get their hands on it. The launch completely failed because they could not satisfy the huge demand and have not been prepared at all it seems.
There were also a lot of problems since then. The CE sign to sell the board in Europe was missing for example. Until today, delivery was delayed which is very frustrating to both, the foundation as well as all interested hackers.
If you are building a distribution you will consequently think about adding support for a new device. In fact, the Raspberry Pi is interesting because it is cheap. It has not got very much power, but well… if people want to use it I am fine with it. But what is the point in doing all that work if nobody actually can buy such a device? It’s not worth doing it. In the IPFire ARM development group, we had the discussion which resulted in a clear: No – because of various reasons.
Why is the Raspberry Pi not eligible for other things?
If the Raspberry Pi computer is not the very best device to run IPFire on one could possibly use it for something else. I already mentioned a web server for personal use: Too less RAM and CPU performance.
It can decode 1080p videos, why don’t use it as a media player for my TV?
Here in Europe the DVB streams are using the MPEG2 codec which the GPU does not support. So are DVDs. However, I am not aware if the CPU can do that in real time but I reckon it can’t do that.
Workstation
Using the RPi computer as a workstation must be horrible. Even the Pandaboard which has got at least twice the CPU power is very very slow and hence there is no nice user experience. It’s like using your old P3 500 from last century again. Admittedly, the graphic would be nicer and the power consumption would be much less. But I don’t think anybody wants to do this.
Network storage device
It has got a 100M NIC. Seriously? There are boxes which are based on the Marvell Kirkwood SoC that have one or two GBit interfaces and are about the same price.
I could possibly think of more such things for which the RPi computer is not a good choice. However, I can not even think of a single one for which the RPi computer is eligible. Maybe you tell me?
Conclusion
Despite of the reasons above, I think there is a point in doing the Raspberry Pi project. The approach to bring cheap computers into schools that can not afford bigger ones is good. I cannot tell if the Raspberry Pi foundation is on the right way or is facing the same issues like the OLPC project did.
For all the hackers out there, the Raspberry Pi is just a toy. It is cheap enough to play around with it and maybe IPFire will be a part of this game at some point. But for that, somebody needs to convince me and lots of other people I talked to…