Slashdot Mirror
Transmeta Goes Embedded
quinticent writes: "An article at CNet talks about Transmeta's entrance into the embedded market. CEO Mark Allen is quoted as saying, "By this time next year, it could equal the notebook market." Wow, when can I get my hands on a cheap embedded Transmeta system to play around with?"
Wow, when can I get my hands on a cheap embedded Transmeta system to play around with?
Wow, that's a good question... considering I seem to remember hearing Transmeta say, in the very beginning, that they wanted to make inexpensive, powerful, efficient laptops, etc., available for everyone.
Where are those computers? The Transmeta laptops certainly aren't any less expensive than your typical Intel/AMD machines. This is unfortunate.
Transmeta may provide the solution to this equation.
Money for nothing, pix for free
They have pretty much failed to capture any significant chunk of the notebook market?
Even in the server arena those new low power 0.13u pentium III tualatin's compete very nicely in terms of performance and power consumption.
So has the hype finally been laid to rest?
My understanding is that embedded systems need more customization than your typical laptop. That's why companies like Tensilica can sell their design cores. Having a chip that relies so much on sw can be a boon for customization, provided that it is open enough (I'm assuming Linus took care of that). However, one of the reasons that so far companies have customized their chips is that they wanted to minimize the sw part for efficiency reasons. I wonder if this doesn't contradict the vision of cheap embedded systems for all...
I've been eagerly awaiting one of the Single Board Computer (SBC) venders to make a device with a Crusoe on it. I work for a company that makes embedded devices, and currently we are using a Pentium/MMX based system at 200 MHz. Even that CPU board sucks down about 25 to 50 watts, which is not good when you are trying to make a system that needs to run from a lighter socket (ca. 120 watts maximum, more like 80 under real conditions.) Plus, you have to get rid of that heat, which adds to the cooling burden, and the @#()$& CPU fans keep crapping out on us. This is on an industrial SBC costing about US$700.
I'd love to go to a Crusoe based system, pick up a few MIPS, and cut a few watts out of the power budget. Add to this the fact that the Crusoe has the North Bridge built-in, which reduces the size of the board, and you have a great win all around for us embedded types.
However, unlike the laptop market, x86 compatibility isn't as great a deal for us embedded weenies, and therefor the StrongArm XScale is an attractive option too. Transmeta had better make this a very compelling option.
www.eFax.com are spammers
When you look at the specs of (or the hype, if you hold that opinion) the crusoe chips, it boggles the mind why they didn't push this this sooner. 'Runs on one AA battery', goes the sales pitch - perfect for the embedded systems which need a half-decent chip but need to conserve battery life (i.e. practically all of them)
A friend of mine was so hoping to use a Crusoe in his embedded web browser tablet thesis project back in 2000. Pity it was released too late for him :)
"Einstein argued that [...] God is not capricious or arbitrary. No such faith comforts the software engineer." ~ Brooks
Maybe you don't blink at paying an extra $100 for a fancier CPU for your home machine, but in embedded systems price is a huge part of the deal. If you can shave $1 off of a CPU then that that's a big difference when you're selling a $50 appliance. Transmeta's chips are even *more* expensive than offerings from other companies, so it isn't at all clear how they're going to go up against processors like the 8051, ARM, MIPS, and so on.
The biggest problem with the ARM core is cost.
Buying the chips might be cost-effective if all you need is a basic ARM, but licensing an ARM core to incorporate into a custom design requires a $0.50 royalty on every every chip sold. This is fine for low-volume devices, but when you plan on selling 10 million chips, its cheaper to fork over the million $ for a MIPS design and sell the chips royalty free. Until ARM changes their licensing agreement they will be more or less locked out of the high-volume custom chip market.
Tim
"By this time next year, it could equal the notebook market," Mark Allen, Transmeta's CEO, said of the company's prospects in the market for embedded chips.
$280mm of cash and near cash
burn rate of $70mm per quarter
capital markets dried up entirely
macro environment going down the toilet
Transmetta will not be around in its current form a "this time next year".
As a stock holder of TMTA (ugh, lately :), I am a little concerned at their changing strategy. Generally speaking, changing business strategy so early in the game for Trasmeta speaks of trouble. It is especially difficult to come to market and try to play catch up with a product that's similar to what already exists, i.e. notebook cpu market. It is even more difficult to switch business strategies and succeed.
Having stated that, my opinion is that this company will end up succeeding. I have trouble believing that such talented individuals (Ditzel, Torvalds, Taylor, etc.) will end up falling on their faces. As a stock-holder (I bought in @$3), my worst-case scenario is that their technology gets bought by one of the bigger players (AMD, Intel). Best-case scenario is that they finally find a product that utilizes the competitive advantages of the Crusoe processor - whatever that may end up being.
Praying for the end of your wide-awake nightmare.
IMHO, the embedded market isn't as big as they claim. And if it is, why would anyone want x86 compatibility? The only use I can see for their current design is what it is doing now, Low power x86 compatible PC chip.
Personally, I would love to see how well the Crusoe goes running native code. The whole super-RISC VLIW architecture sounds a lot better than the x86 or even the PPC. Maybe they should try native BIOS AND a native Linux kernel. In fact, have everything possible native and leave the whole x86 'code morphing' bit for binaries you don't have the source for.
Another idea, since they are emulating/binary translating the code anyway, how big a job do you think it would be to have multiple virtual machines? Naturally they'd all need their own RAM but RAM is cheap. It could give dual booting linux with windows a new meaning!
Also, how easy would it be to add other architectures? I can see a PDA that not only runs Windows CE software but PalmOS stuff as well. Just some ideas, any of them feasible?
Can a Transmeta chip emulate a Java VM and run Java byte code natively. Or is Java byte code not low level enough.
Around a $1.30.
As high $50 last November.