Status on Loongson processor?
- Inicie sesión o regístrese para enviar comentarios
This is a topic about free hardware. I started digging news about Loongson and a few days ago I found the prototype for their new yeeloong machine at this link. It is listed in their official chinese channel. The processor is a 1ghz quad core and consumes only 10 watt. Wikipedia said it would launch at August last year. Lemote says it will be available "soon" but no date. But things got more interesting.
Loongson 3B is an HPC processor which this link provides benchmarks for. Assuming it is telling the truth, this little processor outperforms an i7 while consuming only 40w. Sadly there is no price of the laptop or statements of plans for fuloong or lynloong.
I haven't seen English text with detailed specifications of it before.
It's confusing, since Lemote's website mentions it having AMD integrated
GPU, kernel sources for it contain related blobs, while it's said that
it will not need any nonfree software.
I'm a contributor to Parabola mips64el, supporting Loongson 2F (as used
in 8.9 or 10.1 inch YeeLoongs). Porting there consists mostly of
building packages (this takes days for a KDE update, days for a GCC
update, shorter time for typical packages), fixing some problems (mostly
not patched upstream, due to quality of patches or lack of time), and
not having Java, Haskell, LibreOffice and WebKit-based browsers due to
difficult changes or build needed (e.g. LibreOffice needs three days of
building before it fails for many unrelated reasons, WebKit debug build
uses more RAM than it can with the ABI used, running Midori fails in
difficult to reproduce ways).
I believe making a Trisquel port optimized for Loongson 2F or 3A
machines would not be easier (Debian mipsel supports this machine, but
big speed improvements are done by using a different ABI which needs
patching some packages like WebKit).
I am a member!
I too have been waiting on the new system listed on their Chinese site for a while too. I keep having a Chinese friend of mine peek at it to see if it gets released.
I am a member!
I have gotten the impression that these systems aren't as free as claimed? or they think? Having a free BIOS and even largely free software compatible chipsets doesn't make the system 100% free. There are other bits. I have been told some aspects of the system(s) they are designing are free software compatible which won't be possible to free in an x86 design. AMD and Intel hold certain keys that they will never release. My guess is that these keys some how keep out competitor from the market. I know there are agreements between the two companies on patents related to the x86 architecture and similar.
The AMD GPU I'm pretty confident is going to be a problem with this new system too. I believe they aren't going to and haven't actually released a free software system either. They released a system that was partly compatible with free software which is how Richard Stallman can run mostly free software on it (and a few others, anyone else know who those might be?).
To a degree they have the same problems we have. There are certain things that just can't be done without significantly more resources. There are projects to fix some of these issues. The problem is they are *just* projects. There is no commercialization of these projects and things progress very slowly. Commercialization brings with it resources. It may not always be desirable although commecial aspects aren't in themselves neccessarily bad.
I also just want to be clear- I'm not saying non-free software is ok. There are other ways to commercialise on free software without closing the code.
> I have gotten the impression that these systems aren't as free as
> claimed? or they think? Having a free BIOS and even largely free
> software compatible chipsets doesn't make the system 100% free. There
> are other bits. I have been told some aspects of the system(s) they
> are designing can be free'd or are free software compatible which
> aren't in x86 systems.
YeeLoongs with Loongson 2F have SMI712 graphics chip, which doesn't have
3d acceleration and has available documentation and fully free driver
(although not as well written as popular chips). The EC and hard disk
have nonfree firmware (like all such devices).
Lack of uploadable CPU microcode might be a difference with x86 systems.
Was a bit excited this morning when I received a reply from someone in sales at Lemote.
Part of the email:
Mon, Feb 6, 2012 at 10:10 PM
"Yeeloong 8133 notebook would release next month, and we would update any news on it for you."
Also, it seems the page for the laptop was updated recently. *Unless I'm mistaken* [1]
I am a member!
Unless something changed is this not still heavily dependent on non-free code? I'm mainly referring to the graphics acceleration.
I am a translator!
You need non-free firmware for 3D acceleration, but you don't have to enable 3D acceleration and can actually run the thing without non-free code. You just have to give up 3D acceleration, which is basically useless anyway, since there are approximately 0 games for this arch. I guess it will also not render video as easily as it would with non-free firmware for the graphicscard.
Personally, I'd still prefer a x86 machine, just because x86 is the defacto standard nowadays. The Lemote Yeelong machines are more like a dream of a distant future, where x86 isn't the only choice around and manufacturers of hardware actually care about your freedom and the freedom of developers. I'd like to have one just to play around with it and see how much of my everyday computing can be done with it.
Since I'm playing videogames a lot (wine works wonders once you've ruled out the problems with the video drivers), I assume it won't work out for me as the primary system.
I also need to program stuff and I would like to be able to compile code to a x86 binary. Not sure how well crosscompiling works, but I have a feeling it doesn't work as good as compiling on a real x86 machine.
> You need non-free firmware for 3D acceleration, but you don't have to
> enable 3D acceleration and can actually run the thing without non-free
> code.
This is probably true, unless it needs a nonfree video BIOS (I've seen
kernel sources with it included as a blob). Although there are many
Radeon-related problems reported by users of Trisquel, often due to
drivers being unstable without the firmware.
> You just have to give up 3D acceleration, which is basically
> useless anyway, since there are approximately 0 games for this arch.
There are many free software games (Trisquel repos certainly include
them) and nongame programs like Blender which use 3d acceleration.
> Since I'm playing videogames a lot (wine works wonders once you've
> ruled out the problems with the video drivers), I assume it won't work
> out for me as the primary system.
Users of Lemote YeeLoongs play free software games, there are no
problems in running FreeDink or Wesnoth there. With a GPU not needing
nonfree software, it would be probably possible to play games using 3d
acceleration there.
> I also need to program stuff and I would like to be able to compile
> code to a x86 binary.
I do some programming using a YeeLoong and not building x86 binaries.
There is not reason to build them, unless distributing them to x86
machines which often build such binaries quicker. Distributing binaries
between various x86 machines/distros is difficult (e.g. due to different
shared library versions), building locally is better also without using
other architectures.
> Not sure how well crosscompiling works, but I
> have a feeling it doesn't work as good as compiling on a real x86
> machine.
It works for well-written packages, i.e. it shouldn't be a problem
unless you are a distro developer.
I am a translator!
Thanks for clearing things up. I'm aware that there are a number of free software games and I play some of them too. However I'm mostly interested in games written for Windows, or maybe DOS. They are non-free software, but I make an exception for games.
That's a clever solution, building the binary on x86 and only compile it on the Yeeloong. Didn't think about that. I find those little netbooks more appealing now, but the lack of 3D acceleration is, of course, annoying.
The main reason for choosing MIPS/ARM over x86 is the RISC design. CISC computers are power hungry and overheat a lot. The whole Genesi Smarttop[1] has 5w power consumption and does not have a cooling system. The CuBox[2] has another ARM processor, can decode 1080p and has 3w power consumption. Another one is the Wyse T50[3] thin client which can connect to dual monitors over Full HD and consumes no more than 7.2 watts and does not operate over 40°C. All of this is cheaper to produce than CISC counterparts. My current processor (Phenom II X2 550) alone consumes 80w incorporated in a giant motherboard which is inside another giant case that does not even fit in the desk spot designed for it. Stuff like the Fuloong or T50 would save me a lot of space and electricity money.
[1] https://www.genesi-usa.com/store/details/11
[2] http://www.solid-run.com/products/cubox
[3] http://www.wyse.com/products/hardware/thinclients/T50/
I am a member!
@Horgeon yes, all of those systems use less power. It is because they were designed to. The true measure of performance and power consumption is work per watt. As an example, imagine building Gentoo on 2 machines(hypothetically both architectures are supported etc). Sure a hexa-core desktop will draw much much more power than the smaller system. My desktop pulls about 300 watts with a full CPU load(no GPU loading). Another system may only consume 30 watts from the wall, which is 10% of my consumption. In the end my system MAY use less power because I could finish the compile and either put the system to sleep or shut it down entirely. Until performance/watt can be measured on systems that are as diverse as Loongsoon and x86, and measured reliably, it's too soon to say for sure which uses less total power.
- Inicie sesión o regístrese para enviar comentarios