Slashdot Mirror
Transmeta Mulls Exit From Processor Market
chill writes "C-Net is reporting that CPU upstart Transmeta, once the employer of Linus Torvalds and maker of 'Code Morphing' processors, is contemplating leaving the chip manufacturing business. Already their IP licensing revenue exceeds that of their microprocessor sales, though both are dwarfed by their recurring quarterly losses."
And yet they're going to the CES in Lost Wages. (Booth 36235, LVCC)
[Hello! My name is ARTHUR SWIFT] "Hi, these are our microproceesor products, which cost more to make than we sell them for. We're thinking about breaking into the game console market next. Losing money seems to be working for the X Box!"
A feeling of having made the same mistake before: Deja Foobar
"...Transmeta is reporting a further reduction in power requirements by 44% and sees the laptop and sub-laptop markert as the primary markets for their new CPU. Intel and AMD claim to be catching up with the Transmeta chips in terms of power requirements..." Yup, that worked out well.
Virtually every field nowadays seems to be ruthlessly dominated by one or two (if you're lucky, three) titanic competitors. Trying to break into an existing market is tantamount to financial suicide. Not because newcomers have bad ideas or make bad products-- but because the "mindshare" of the unwashed masses is so stuck on the existing titans..
.... well, Microsoft for operating systems.
McDonald's and Burger King for burgers.
Coke and Pepsi for cola.
Nike and Reebok for sneakers.
Microsoft and
Dell and HP/Compaq for x86 computers.
ATI and nVidia for graphics cards.
And... Intel and AMD for x86 CPUs...
Honey, I shrunk the Cygwin
TM always avoided industry standard benchmarks. That is always a bad sign.
Instead, they always had some loser from marketing spout about efficency, blah, blah, efficency. I still have those emails and they are very funny.
While I worked at a major OEM developing blade servers, we evaluated their processors and the performance was very weak. De-clocking existing proven designs was a better alternative.
As is often the case with weak products, non-disclosure agreements precluded benchmark publication and disclosure of evaluation results.
RIP TM.
Quarterly losses? Whew! At least they arent sufferent from monthly losses. Or even weekly ones.
Invoicing, Time Tracking, Reporting
Eating a Wendys burger with a can of RC Cola, Wearing British Knighs Snearkers, Running OS X, on a Power PC Processor, but I have a nVidia graphics card... Damn! I guess I am just a Puppet of the Man!
Actually they are 3 Mega Corps but the #3 is usually far behind, but still close enough to get good Profit.
If something is so important that you feel the need to post it on the internet... It probably isn't that important.
All that development and hype, yet now they are getting out of the market. Seems they should have been well positioned to dominate in the handheld and portable market. Bad business practices? The EE Times also has a good article on this.
Having done so much with so little for so long, I now can do anything with nothing at all.
"Code morphing" would have been more useful if the instruction set to be emulated was less well matched to a hardware implementation. The VAX instruction set comes to mind. That instruction set was hard to make run fast. Individual instructions had too many sequential steps. DEC struggled with that for years. But few need a fast VAX any more.
The only reason that Transmeta had any success at all was that they built a chip with good on-chip subsystem-level power management. That's something which Intel and AMD had previously not considered too important, having focused on desktops first and laptops second. But it's not hard to do, and Intel then started doing it.
I always wondered: with their "code morphing" technology, why didn't they turn it into a Java processor, with the ability to execute Java code natively? Yeah, I am aware of Sun's past efforts in this direction; but imagine if you will: Java apps running natively at the same time as Linux apps. Even if the processor is 3x slower than a x86, Java running natively would be comparable to Java running under VM on an x86 (please, I don't want to start a flamewar about Java's speed here).
Sad to say that, but Transmeta marketing is to blame. Face it, Transmeta could have had the market via now has with their C3 and Intel partially holds with their Pentium-M Centrino. The main problem was, that Transmeta went from day 1 to the manufacturers only, and were leaving the early adopters hardcore system builders out. Add to that an emerging home theatre pc market which Transmeta failed to cover that way (nobody bought at the early stages a HTPC from a manufacturer and most people still dont do due to DRM and other nastyness)
and a VIA which just said to the people, we are not fast, but they can handle the stuff you want to do with your HTPCs self made routers, firewalls, fileservers (you name it), we are cheap you can buy our stuff from the next vendor on the net and we will support you, and Transmeta was on a losing ground.
On one hand there was ARM which only sold cores and they did need less power, on the other hand there was VIA with the mentality you can buy our stuff even as a private person, and on the Notebook computer segment there were the Heavyweights Intel and AMD crushing Transmeta left and right.
So where did Transmeta stand there, basically nowhere because they refused people (and there were thousands who wanted to buy that stuff at an affordable price) the hardware, by selling only reference designs and not having others selling decent boards to an affordable price. Add to that that in Europa and other markets you basically could not get the stuff and that interested people were complaining in forums about that situation for years and you have a company doomed from day 1.
Now they want to concentrate on the core selling business, I wish them good luck they will need it, between a very good ARM on one side and VIA which still also sells boards to people if they need them on the other side and an Intel with a very good low to medium power solution on the server/notebook corner of things. Also IBM is in the business or at least other companies selling cores on the based PowerPC design.
Guess it is time to say to Transmeta, goodbye it was nice knowing you. (Hopefully not but there is a high chance)
Low level specs are great but I have a real 1Ghz transmeta chip on a sub 2 pound Sharp notebook (OK 2.5 with big battery). It lasts 9 hours like it says (tested on a long flight) and runs more than fast enough with XP for Powerpoint, Word and not bad even with a pig like CorelDraw. The power cord accidentally got unplugged while it was connected to the network once and it still had half the battery left after nearly a week on intermittent sleep
The price performance thing is pretty meaningless as long as it is fast enough to do what you need. Not everyone uses their laptop as a primary machine, or for video processing. My main need was something that didn't weigh like a lead brick and could let me do real work on a long flight or a meeting without having to plug in somewhere. The Sharp/Transmeta does that admirably.
As for the Centrino - it may be great - I don't know but I wouldn't go by spec-sheet alone (Xeons are the fastest chip right?). I'm curious if anyone here has *real world* experience with the Centrino based Sony? My understanding is that it has about half the battery life of the Sharp from the user reviews and I certainly don't discount that this might be because of different power management schemes that don't relate to the chip. But as a end-user consumer the Sharp notebook was a lot cheaper than the Sony last I checked and is far from being a sub-par product.