Slashdot Mirror
Build Your Own Crusoe-Powered Computer
jonmason00 writes "Just checked the Transmeta webpage, and discovered that they are now offering a Crusoe TM5800 System Development Kit. It's a bit expensive ($995) and you gotta register before you can buy one, but they need your support." How about an Astro development kit instead? :)
Listen, I love Transmeta because they had some cool ideas when they started out. And hey, Linus Torvalds, the creator of Linux, works there as well. So their geek factor is really high and they seem like nice people as well (I visited back in 2000).
I was just let-go by my company last Friday after 10 years of service. Bills are lying around waiting to be paid and I'm trying to send out resumes and find some work before my wife and kid leave me. I love them to death and would probably "end it all" if I lost them.
So why should I spend my meager (non-existant, now) salary to support a company? They're in a business, and the business of business involves profiting.
If you can't profit, you lose the business. So Transmeta, enough with this puppy-eyed cutesy appeal to the geek masses for financial support. Many of us are unemployed as it is and risking losing our lives from insanity.
Yeah, it's stupid to say that a for-profit company needs our support, but this thing is not meant to be your next desktop machine. Transmeta knows you can get a barebones x86 box for much less with far greater performance--they're not as stupid as many of you would like to believe. They're selling a development kit, i.e. for someone wanting to prototype, say, a stereo component or set-top box for resale.
We in the Java world were creaming all over ourselves at this prospect.
Why didn't they persue the embedded device executing bytecode natively path?
Additionally, if you're looking for higher end right now, choose one of the many mini-itx configurations available. http://www.mini-itx.com is a wonderful site based in the UK. Buy directly from them or use one of the vendors they recommend.
Sorry Linus , but people developing for tiny platforms can't afford to spend an extra $400-$500 for a Transmeta solution.
It's very nice that there could be an alternative too Intel/AMD, but this product is simply to expensive.
Transmeta is like the Green Party. In 2000, neither the presidential candidate from either of the two major parties was terribly attractive. Nader, on the other hand, had some good ideas. Then again, was it really worth voting for him? It was impossible for him to get elected.
I voted for him...
pirates
What would really be cool is if they had a kit that allowed one to right their own "code morphing" code on top of their vliw core. I'm sure someone is/has work/ed/ing on a jvm for it. But imagine a linux kernel on core. I know that stuff like this has been tried before and failed miserably, but hey, that's what stuff like OS is about, just doing and making it work, and saying "now isn't that cool".
Or you could even come up with your own custom extensions to the x86 IS, implement game logic, whatever.
That's a fuckload more than transmeta has done for linux. (Hiring Linus was a marketing move and little more).
Call it flamebait. Personally, I wished the transmeta had succeeded. The idea was (and is) incredible. But they're basically just an expensive, low-speed celeron. wank-wank.
Do you even lift?
These aren't the 'roids you're looking for.
AMD and Intel are pushing for integrated DRM in all systems. Using Transmetta products might be a way of avoiding that - if enough people boycott AMD and Intel and are vocal about their reasons we might be able to get compulsory DRM at bay.
However I did notice that they use an ATI video card. Bad move if you're wanting to use that under Linux. Their video cards are all tied up in patents. I have been trying for 18 months to get an answer on why ATI asked people to cease development of TV-out support on my Radeon. That was one of the reasons I bought it, and it WAS supported and worked well at the time. Now however it only works with 18-month old drivers that don't really sit well with X. Damned ATI. Oh and Damned nVidia also. Their driver lock my system every 20 mins without fail - the other reason I chose to buy a Radeon. Maybe I should just redirect my console to my canon bubblejet...
totally agree - every time there is a stupid quiet pc or case moding story on slashdot I post to mini-itx.com. don't really know why people arent' more pysched, this is the board everyone claims they've been wanting - cheap, quiet, low power, mid range performance. perfect for web and multimedia.
Sure, I know they're a business. But they're going out of their way to support linux and that's something I'm not ashamed to support. You scratch my back, I'll scratch yours as they say.
$ host -t NS kernel.orgkernel.org name server ns2.transmeta.com
kernel.org name server zenii.linux.org.uk
kernel.org name server ns.vger.kernel.org
kernel.org name server ns1.kernel.org
kernel.org name server ns1.transmeta.com
kernel.org name server ns2.kernel.org
Multiple clocking. Some chips are clocked to rising edge, some to falling edge, and some to both. Clocking to both doubles the EFFECTIVE clock speed.
Differently specced clocks. The AMD Elan I'm working with now can use either 33.000MHz or 33.333MHz as its base clock chip, leading to slight variations in the final frequency, but that's rounding error.
Split clock, anyone? Horrible to work with if you've done any LSI or VLSI, but running the chip on multiple multiples/divisors of the base clock is actually relatively common.
Finally, asynchronous designs, lol, although neither of the systems mentioned are asynchronous. Clockspeed is no longer a true concept anymore.
Hardware, software, and blinking lights!
The more I read about the Crusoe chips, the more I start to think about how they should just drop the strict x86 compat, and start people developing directly for the core of the chip, in it's native tongue.
... make the Crusoe a 64 bit chip, and codemorph SPARC or MIPS.
The Crusoe is a chip that runs comparable to a similar Intel/AMD chip and yet does it all through software emulation. Imagine what it can do with it's straight instruction set! Developers could start programming applications that take advantage of this, but then if they required a backward compatible app that has been compiled for x86, then the chip could also run those too. A distro based on the "native" Crusoe instruction set, could run x86 compiled applications.
Why stop there? If they are able to codemorph x86, why not PPC or
Your mom always said, a PB&J is better than nothing, and God is nothing, is a PB&J better than God?
Or, build your own Intel- or AMD-based computer that's cheaper, faster, and uses less power.
Not according to the guys who know what they're talking about.
Your AMD or Intel machine will not get anything close to the mips/watt ratio that the Transmeta does. The LANL people go on to conclude that the Transmeta is cheaper in the long run as well, because of power (including cooling) and space savings. Faster, I'll grant you, but one out of three is pretty poor batting, certainly not worth a moderation of 5.
For a cluster, the faster argument goes out the window as well, because the performance equation comes down to mips/watt, mips/cubic foot and mips/$$$, in all of which Transmeta leads AMD and Intel.
Now I don't know about you, but I find the monolithic, nuclear reactor core kind of box is getting less and less interesting as time goes by, and what I really want is a box full of much more efficient processors, all dirt-cheap of course. I'll admit that that there's no way for the typical home user to get into this kind of system for a price that competes with a single, Athlon or P4, but that's this year. Check again next year.
Have you got your LWN subscription yet?
Considering Torvalds' flirting remark with X86-64 half a yea ago, will there be a Crusoe supporting x86-64? Anyone heard any rumours?!