Slashdot Mirror
Does Transmeta Live Up To The Hype?
onion2k writes: "In this article on VNUnet, Toshiba are saying that the Transmeta chip doesn't quite live up to its hype. Bit of a strange thing to do considering Toshiba were one of the original investors, but hey, thats corps for you ..." Talks mostly about the power consumption of the chips. If you're following Transmeta, this is worth a read.
they are slashdotted, so here it is :)
(posting as anonymous coward, since this is
obviously karma whoring
Toshiba questions Transmeta power claims
Toshiba has poured cold water on Transmeta's claims about the low power consumption and heat generation of its Crusoe chip, despite the fact that it is an investor in the chip startup.
Transmeta claims that Crusoe increases battery life in lightweight notebooks to eight hours, more than doubling the two to four hours provided by equivalent Intel chips. It also claims that notebooks running the chips are quieter as Crusoe does not need noisy cooling fans.
However, Steve Crawley, Toshiba UK's product marketing manager, said that the company had no plans to introduce Crusoe into future Toshiba products.
"[Crusoe] does give a reasonable increase in battery life, but nothing like Transmeta's publicity is claiming. The back light consumes a lot of power - one quarter of the power is used pushing light out. Realistically, in sub-notebooks it gives a 30 to 40 per cent increase in battery life," he said.
He added that Toshiba currently has prototypes of ultra-light notebooks with eight hour battery life using Intel rather than Transmeta chips. "This can be done with a standard Intel box," he said.
"It is an interesting technology but at the moment we are not convinced it offers the user what is required. It will be very interesting to see if it can add any significant value to the end user in terms of battery life or thinness," he said.
But Transmeta contested Toshiba's claim, saying its eight hour notebook was too heavy to be classed as an ultra-light device.
Ed McKernan, director of marketing at Transmeta, said: "Toshiba's eight hour battery life today requires a 2.2lbs battery attached to the base of their Portege 3440 and 3480 notebooks. This means that [it] ends up weighing 5.61lbs - which is outside the ultra-light category of 2lbs to 4lbs."
"Transmeta's Crusoe processor is today being designed into products that will arrive in the fourth quarter and first quarter [of next year] with all day battery life. In addition, it is providing relief to original equipment manufacturers and designers that must deal with the heat caused by the hot processors - even Intel's 'one watt' processor," he added.
Toshiba agreed a licensing deal with Transmeta in February 1998, following the signing of a similar deal with IBM in December 1997. The deals, which gave Transmeta access to IBM and Toshiba technology in return for the right to use Transmeta technology in x86 products, provided the startup with much of its early revenue.
Transmeta subsequently reacquired the rights granted to IBM and Toshiba to manufacture and market x86 compatible products. It agreed to pay IBM a total of $33m over the next four years and issued 600,000 shares of common stock to Toshiba. IBM and Toshiba retain a licence to manufacture, market and sell non-x86 compatible products incorporating the licensed technology.
Intel and AMD are likely to take other approaches, and improve power consumption. Perhaps they will not get as favorable results, but "close enough" is good here, much as with horseshoes and nuclear hand grenades.
Having a CPU that consumes ten times less power is of limited value if the hard drive and LCD display still suck (power).
When Transmeta pushes that the CPU consumes vastly less power just makes Toshiba look bad if Toshiba can't make all the other components consume less power.
"Politically," this is probably what they're pushing back at...
If you're not part of the solution, you're part of the precipitate.
One of the primary reasons for the lower power consumption claimed by Transmeta is that the chip area is much smaller - Pentium compatibility is achieved in software instead of transistors.
Wait a second... doesn't a smaller chip area mean that it's supposed to be significantly cheaper, too?
I guess it could have been much cheaper than a Pentium, but they have a huge investment to cover so they are going for a high-margin market by targeting manufacturers that desperately need something other than price - lower power consumption.
So what do you do if the CPU isn't the only power hog around?
----
Stop worrying about the risks of nuclear power and start worrying about the risks of not using nuclear power.
So Toshiba's backlight is a power hog and that's Transmeta's fault?
You don't understand. The question is not one of "fault" -- we are not trying to find somebody to blame. The question is whether it makes sense to make long and expensive efforts in order to reduce the CPU power consumption when it may not be all that important. Especially given that it involves performance trade-off.
Kaa
Kaa
Kaa's Law: In any sufficiently large group of people most are idiots.
Why would Toshiba diss a company they have invested in before the IPO if their claims do not have substance?
Unlike movie stars, stocks usually are affected negatively by bad publicity and react well to good publicity. For Toshiba to damage potential payoffs from a successful IPO by badmouthing Transmeta must indicate that all is not kosher with the Intel killer.
Hanlon's Razor
I mean take a look at Intel and RDRAM.. RDRAM is 8x more expensive than PC133 SDRAM and yet performes only marginally better (from what i have seen through numerous benchmarks). It's pretty obvious that RDRAM is a sub-par tech that Intel invested in and Intel seems to be QUITELY moving away from RDRAM... however u don't see Intel denouncing RDRAM by officially saying "RDRAM sux, we trying to phase out the usage of it, but we are trying to save face".
Assuming what i say is true, what reason would Toshiba have to do this right before TM's IPO? Now i'm not a big fan of consiracy theories, but it looks like there might have been a falling out between Toshiba and TM. Something like "Ok TM u fuc*ed us over, now we are going to fuc* u over by screwing up your IPO".
To me, i think that BECAUSE Toshiba is an investor in TM, it lends them LESS credibility to say things that will give TM (and therefore themselves) bad PR.
Of course in the end the technology will have to stand on its own two feet... only a released TM chip can disprove what Toshiba has said.
Ceres
So Toshiba's backlight is a power hog and that's Transmeta's fault?
-
-
Give me liberty or give me something of equal or lesser value from your glossy 32-page catalog.
The problem with Transmeta is that their entire business plan depends on being bought up by Microsoft. The bait is Linus. They want to use him so M$ will buy up the company and make him work 2 days a week at a minimum wage job so he and his family starve to death in the valley. This is why minimum wage should be $80 / hour, so all those people raising their families working only a couple days a week can afford to live in dignity.
How do I know all this, well let's just say that I have nice paychecks coming from two of the companies in my comment, but I won't say which.
I think what Toshiba may be discovering is that the semi-software solution to running CPU registers may not be the best solution to run operating systems reasonably fast.
Besides, note that Intel recently has made massive strides in reducing the battery drain of their Celeron/Pentium III CPU's designed for laptop operation. And AMD is heading towards that direction by going to a 0.13-micron process for their mobile Duron/Athlon CPU's to be released within the next 18 months.
Transmeta will have to crank up the speed of their CPU's substantially if they are to stay competitive against these new generations of AMD and Intel mobile CPU's.
Raymond in Mountain View, CA
Maybe Transmeta can get out of the processor business and invent a technology that prevents sites from getting slashdotted. Oh, wait, there allready is such a technology, its called mirroring!
Geoff
I think people should probably reserve judgment until units actually start to ship. Only by observing the real world performance, both at the CPU/throughput level and the power management level, will we really know if Crusoe will live up to the hype.
;-)
I for one can't wait to get my hands on a unit to evalutate. Hey, a good battery test would be to run the Q3 Quaver demo over and over and over....hey, if the Crusoe really does optimize on the fly, then my framerates should get higher with each interation!
Bit of a strange thing to do considering Toshiba were one of the original investors, but hey, thats corps for you
"Hey thats corps" what is this supposed to mean. Toshiba have one of the best hardware R&D arms around and were as stated one of the people behind Transmeta.
How about this for an idea....
Transmeta is good, but not that good, its not a revolution its just pretty good. The hype accorded to Transmeta is way out of kilter with its proven ability. Look at ARM, years of deployment from StrongARM 64 bit in servers down to hand-helds. So what have Transmeta actually DONE ? Very little indeed and yet judging by the hype hear on Slashdot they are the successor to Intel, they will triumph, their technology is much better than anyone elses.
Now look at how Microsoft hyped Win2K before and after its release. Do you see a difference ? The only one I see is that MS have delivered something.
An Eye for an Eye will make the whole world blind - Gandhi
Maybe Toshiba are just annoyed it didn't work out for them...
Strong data typing is for those with weak minds.
Strong data typing is for those with weak minds.
Never knock on Death's door:
The Anti-Blog
So what is Transmeta for now?
8 hours? Bah. A MacPortable got 4-6. And it's carrying case would hold a second batterry.
And 2.2 lbs? so what. The batteries were 5 or 6 lbs--quite handy as a security measure: if someone tried to steal it from you, you pulled the battery out of the case, and let them run off. Cassually follow them until they tire from teh weight, and then hit them over the head with the spare battery.
:)
hawk
Would they? Given the Linus' standing, would we care as much about Transmeta if he wasn't involved? Is a lot of the interest in Transmeta because of Linus works there or is it because there's this huge demand for low power CPUs? The way I read Transmeta stories/postings/etc its as if we'd never seen a low-power CPU before and we were all running computers that needed two-phase 208v and air conditioning.
You'd think if it was just low power CPUs people would be into all kinds of StrongARM solutions (not that some aren't, mind you, but there's no Linus mystique). I know the emulation, which is only x86 as of yet anyway, part is cool but at the same time I can't help but think that its 50% cool tech, 50% Linus worship.
If I was to buy a K6-II+ notebook and a Transmeta notebook (assuming that the transmeta didn't cost twice as much which is likely will, from what I hear) which one would be faster? If I buy a Sony VAIO with a Celeron or P3 in it and a Transmeta notebook, which will be faster? Are these supposed to be Cadillac notebook chips? Or "working man's" notebook chips?
Next, if I was to treat the Transmeta part like an embedded chip. How cheap can a wireless webpad with Linux and mozilla on it be? How long do the batteries last? Can I plug my Nokia PCS phone in to it? Yeah, these are all product specific questions. The big one is "how cheap?" I'm not going to buy a $1000 web pad when I will be able to buy a $1000 K6-II notebook ($800-$1200 is what they are aiming for with the k6-II+.) The TM chips are supposed to cost in the $70 range in quantity so I'm guessing you're not going to see many $300 TM based products.
There is a fair amount of information on the processor but I still feel like the important real world information is missing. I also feel like TM has been kind of deceptive with it, if they had a cheaper, lower power process that delivers the same performance they would be banging that gong to no end. They don't, it's either more expensive or it's slow, or both. It also kind of feels like they are moving slow. I know it's a first rev of their part but since that first press conference it has almost been long enough for rev 2 to be nearing the end of the pipeline. Something else that they should be hyping, especially if their first rev is an underachiever, that's expected but it's what you follow up with that makes or breaks you.
It's all very cool none the less, I hope they do well. More competition is better.
This is my signature. There are many signatures like it but this one is mine..
The big criticism seems to be performance. I see this as a complete non-issue, but then again I'm not a compulsive upgrader.
Just two years ago, I was doing serious day-job work with a 200MHz Pentium running Windows NT. I had no speed complaints whatsoever. I'm currently doing software development for a _large_ project using a 400MHz Pentium II. It's plenty fast for my needs (which includes lots of heavy compiling). I also have a 333MHz Pentium II at home, which I use for running Lisp, doing 3D modelling, compiling, and graphic arts work. No speed complaints from me. It's a very snappy system, even running Windows 95. (I'd switch to Linux if I could find a desktop environment that didn't drag performance down to sub-Windows levels. Even so, I still have a 2M Linux partition that gets the occasional use.)
Now, most of the Transmeta detractors are complaining because Intel and AMD are beating them in the raw speed department. But then I realize that even Transmeta's low-end chips are running at outrageous clock speeds, giving performance better than any machine I typically use. That's enough for me. More speed than I know what to do with. Very low power consumption. Please, get these into inexpensive notebooks ASAP.
Apple's G3 laptops, on the other hand, have battery lives that are two to three times longer than Wintel laptops. That suggests that Crusoe and the old G3 are similar in computation per unit power. Of course, Apple/Motorola has gone the other direction with G4. It has a huge die to accommodate the SIMD instructions in the Altivec. Power consumption has skyrocketed. Still, Apple's MPEG DVD playing demonstration with the G3 is a great benchmark for a computer.
It's clear that Transmeta began as an academic exercise in exploring a new kind of computer chip that converted the instructions on the fly. They were entranced with the possibilities of doing pre-processing of instructions at the time of execution. This gives them some neat benefits, but they're not huge ones. After a bit, the Transmeta folks gave up singing that their processor was going to go faster and started hyping the power consumption.
It's not clear to me that Transmeta hasn't used any techniques that can't be easily ported to the latest model of the Pentium. So we'll see what happens.
I would also take Toshiba's decision with a grain of salt. There are plenty of other big companies with good engineeer who are still behind Transmeta.