Posted by
ryuzaki0
on from the chips-ain't-easy dept.
aftk2 writes "According to news.com, chip maker Transmeta - current home of Linux creator Linus Torvalds, has canned 40% (200 people) of its work force, and has shifted its goals toward obtaining profitability in 2003. No word on whether there were any penguins seen leaving the building."
Re:Not surprising...
by
Coplan
·
· Score: 4, Insightful
I agree: None of their ideas were "insanely great" as you so well put it. The problem as I see it was too many promises, late releases, and the simple fact that they bragged a bit too much too early.
They should base their business model on some company that is well liked as opposed to a company like Microsoft. Notice similar business tactics? Difference is, Microsoft is big enough to pull it off.
Seriously though, the ideas they had could very well be worthwhile. While I would hate to see Transmeta fold, at least the ideas and the technology are out there, and would likely be sold should the company fail completely. Supposing a company like AMD got ahold of Transmeta's research and knowledge base...a veteran company might be able to market such a product better than Transmeta has.
The way I see it, Transmeta will either pull through, or the technology will get passed on. Sad to say, but it's win/win for the industry...and only hard times ahead for Transmeta.
Re:Not surprising...
by
Com2Kid
·
· Score: 4, Insightful
Granted their code-morphing and use of VLIW had some interesting concepts, and their power consumption was perfect for laptops, but there just wasn't much of a market for what they developed.
Nah, there just wasn't much of a market for how they where selling what they had developed.
Now if instead they had, say, concentrated on making development platforms . . . . heh.
Can you imagine sitting down at a machine that is a Sun, PowerPC, and x86 all in one?
That is (was?) the true promise of Transmeta and shoving the chips into laptops was just plain silly. Bleck.
(actualy I just just kind of hoping for an uber emulation machine myself, hehe. Think the next Generation of MAME.:-D )
Re:Goals of the company
by
Planesdragon
·
· Score: 4, Insightful
Most could give a rip about battery life or heat.
You, sir, use a laptop as a portable, not a laptop.
Battery life is *more* important than processor speed, to me. Were I in the market for a new lappytop, I'd want something that I could use for a several hour stretch in the park, in the car, or just wherever the feng shui is best for writing.
Once it can run the word processor and MP3 player at once, at a speed I don't cringe at, I'm happy & the rest is just gravy.
Transmeta - the Power Management Company
by
Animats
·
· Score: 5, Insightful
The trouble with Transmeta was that the feature people wanted was fine-grained power management, not software translation into microcode. Transmeta was first with fine-grained power management, but as soon as it became clear that people cared about that, everybody else (i.e. Intel and AMD) started doing it, and Transmeta lost the only advantage it had.
Transmeta's "code morphing" turned out to be another Really Neat Computer Architecture Idea that Doesn't Matter. It goes to the graveyard with stack machines, tagged-word machines, capability machines, dataflow machines, single-instruction multiple-datastream machines, hypercube machines, and Forth machines. Each of those has been made to work, built, and sold. Few people have ever seen any of then, but they all did exist as working commercial hardware at one time or another. None of them had enough of an advantage over vanilla architecture to survive.
The same thing will probably happen to Intel's Itanium, which, even within Intel, is considered a marginal idea.
In a way, it's sad. We're stuck with vanilla architecture like x86 and vanilla languages like C. There are many better approaches, but none better enough that the pain of conversion is worth it.
Re:Transmeta - the Power Management Company
by
iabervon
·
· Score: 3, Insightful
Stack machines... you mean, like the Java Virtual Machine? Or PostScript printers? Nope, nobody using that idea any more. Actually, a number of the neat old ideas in computers turn out to be great for something somewhat different, or somewhat later. Stack machines are a great idea if you don't know how many registers you have. With real machines running compiled software, this is stupid; but for virtual machine or for document-formatting instructions, it's great.
In any case, it's not neat ideas that sell machines, it's solved problems. Code morphing is a great idea, and it'll be really big as soon as someone wants to do something that it's good for.
Re:Transmeta - the Power Management Company
by
hoggy
·
· Score: 3, Insightful
Stack machines... you mean, like the Java Virtual Machine? Or PostScript printers? Nope, nobody using that idea any more.
No, I think the OP meant that nobody is building actual stack-based processors anymore. The JVM is a virtual machine, and PostScript printers contain an interpreter running on a conventional processor (usually a RISC chip). Other than Sun's brief fling with the Java processor, stack machines have pretty much died.
Code morphing is a great idea, and it'll be really big as soon as someone wants to do something that it's good for.
Code morphing is a great idea indeed. But it already is in use in any JIT emulator or virtual machine. Crusoe is basically a very power-efficient processor running an x86 JIT emulator.
The big unanswered question is whether VLIW was a good idea or not...
I would like to point out that if the workers owned the means of production, this wouldn't have happened.
-- [o]_O
Re:Code Morphing
by
Ian+Bicking
·
· Score: 4, Insightful
My impression when Crusoe was announced, was that the x86 instruction set was very important to them -- it's a much easier thing to emulate efficiently. A RISC instruction set (as in PowerPC) is much more difficult -- since now you're translating from RISC to RISC (since the internal instruction set is more-or-less RISC as well). You can decompose CISC instructions efficiently, but there's nothing to decompose with RISC -- the instructions are already simple.
The other potential seemed to be that they'd create different cores with different optimizations -- the first one, Crusoe, being power-efficient, another one could be optimized towards floating point, another to integer operations, etc. But that hasn't happened.
Alternate architectures would be interesting -- at least PPC. In a Mac, it could allow efficient Windows emulation... but that seems like less and less of an issue, as portable applications usually mean web-based, and non-web applications usually have Mac alternatives. At least, I don't think Apple is enthusiastic about Windows emulation, and without Apple PPC is useless, since they won't have MacOS. Other non-x86 architectures don't seem important -- there's little software available for ARM or SPARC that won't be ported to x86 if there's demand.
It's like with languages -- if you know English, learning a second language is no longer that important. Transmeta started out learning the English of the instruction sets -- x86 -- and there's little incentive to learning other languages. Even if some programs started out with different machine languages, they all learn to speak x86 eventually.
Re:Intel Fires 4000 Employees 2002-07-16
by
buffy
·
· Score: 5, Insightful
Why didn't Slashdot report this news item, hhmmmmm?
Because Linus doesn't work for Intel, silly.
Re:Goals of the company
by
sql*kitten
·
· Score: 3, Insightful
Their morph-code technology wasn't as desired; look at how many people don't even flash their BIOS.
The real problem with code morphing is that why would anyone pay for morphing code to x86 when x86 processors are so widely available? Now if they'd had the ability to code morph x86 to native *and* the ability to code morph the instruction set of (say) the Java Virtual Machine to native, then maybe we'd be talking. All the advantages of Java, but executing at native speed, plus compatibility with all x86 applications, and maybe SPARC too. Then they could have simply sold the company to Sun for it's next attempt at thin-client desktops. But what was the point of code morphing to only one target? This is something the VCs should have asked before investing a single dollar.
Slashdot Mirror
They should base their business model on some company that is well liked as opposed to a company like Microsoft. Notice similar business tactics? Difference is, Microsoft is big enough to pull it off.
Seriously though, the ideas they had could very well be worthwhile. While I would hate to see Transmeta fold, at least the ideas and the technology are out there, and would likely be sold should the company fail completely. Supposing a company like AMD got ahold of Transmeta's research and knowledge base...a veteran company might be able to market such a product better than Transmeta has.
The way I see it, Transmeta will either pull through, or the technology will get passed on. Sad to say, but it's win/win for the industry...and only hard times ahead for Transmeta.
Granted their code-morphing and use of VLIW had some interesting concepts, and their power consumption was perfect for laptops, but there just wasn't much of a market for what they developed.
:-D )
Nah, there just wasn't much of a market for how they where selling what they had developed.
Now if instead they had, say, concentrated on making development platforms . . . . heh.
Can you imagine sitting down at a machine that is a Sun, PowerPC, and x86 all in one?
That is (was?) the true promise of Transmeta and shoving the chips into laptops was just plain silly. Bleck.
(actualy I just just kind of hoping for an uber emulation machine myself, hehe. Think the next Generation of MAME.
Need help treating your acne? Come here!
Most could give a rip about battery life or heat.
You, sir, use a laptop as a portable, not a laptop.
Battery life is *more* important than processor speed, to me. Were I in the market for a new lappytop, I'd want something that I could use for a several hour stretch in the park, in the car, or just wherever the feng shui is best for writing.
Once it can run the word processor and MP3 player at once, at a speed I don't cringe at, I'm happy & the rest is just gravy.
Transmeta's "code morphing" turned out to be another Really Neat Computer Architecture Idea that Doesn't Matter. It goes to the graveyard with stack machines, tagged-word machines, capability machines, dataflow machines, single-instruction multiple-datastream machines, hypercube machines, and Forth machines. Each of those has been made to work, built, and sold. Few people have ever seen any of then, but they all did exist as working commercial hardware at one time or another. None of them had enough of an advantage over vanilla architecture to survive.
The same thing will probably happen to Intel's Itanium, which, even within Intel, is considered a marginal idea.
In a way, it's sad. We're stuck with vanilla architecture like x86 and vanilla languages like C. There are many better approaches, but none better enough that the pain of conversion is worth it.
I would like to point out that if the workers owned the means of production, this wouldn't have happened.
[o]_O
The other potential seemed to be that they'd create different cores with different optimizations -- the first one, Crusoe, being power-efficient, another one could be optimized towards floating point, another to integer operations, etc. But that hasn't happened.
Alternate architectures would be interesting -- at least PPC. In a Mac, it could allow efficient Windows emulation... but that seems like less and less of an issue, as portable applications usually mean web-based, and non-web applications usually have Mac alternatives. At least, I don't think Apple is enthusiastic about Windows emulation, and without Apple PPC is useless, since they won't have MacOS. Other non-x86 architectures don't seem important -- there's little software available for ARM or SPARC that won't be ported to x86 if there's demand.
It's like with languages -- if you know English, learning a second language is no longer that important. Transmeta started out learning the English of the instruction sets -- x86 -- and there's little incentive to learning other languages. Even if some programs started out with different machine languages, they all learn to speak x86 eventually.
Because Linus doesn't work for Intel, silly.
Their morph-code technology wasn't as desired; look at how many people don't even flash their BIOS.
The real problem with code morphing is that why would anyone pay for morphing code to x86 when x86 processors are so widely available? Now if they'd had the ability to code morph x86 to native *and* the ability to code morph the instruction set of (say) the Java Virtual Machine to native, then maybe we'd be talking. All the advantages of Java, but executing at native speed, plus compatibility with all x86 applications, and maybe SPARC too. Then they could have simply sold the company to Sun for it's next attempt at thin-client desktops. But what was the point of code morphing to only one target? This is something the VCs should have asked before investing a single dollar.