Slashdot Mirror
Transmeta Confirms Recall
jbischof was the first to tell us that Transmeta has confirmed that they are recalling Crusoe, as we
mentioned earlier. The statements says it's fewer then 300 NEC laptops, so it's not that huge of a deal after all. Of course the egg-on-face factor is still high.
You do realize that the reason that it is called 'BogoMIPS' is that it is a bogus measure of system or processor speed. It is only used for internal timing.
I noticed the article mentioned that Intel now has mobile chips that consume that same amount of power as the Transmeta chip. This seems to put Transmeta in a bad position. If they create software for their chip so it will translate for other architectures (powerpc, sparc, alpha, etc.), they will be in an excellent position. I just can't see them going very far right now without reducing power consumption more, or getting into other architectures.
Need Free Juniper/NetScreen Support? JuniperForum
It also bears stressing that all modern CPU's, and that does include Intel, have extensive "microcode" layer to actually drive the chip. It's perhaps something of a simplification, perhaps a bit nit-picky, but "uses elaborate software instructions rather than hardware" is quite misleading.
They BOTH use elaborative software instructions on top of hardware, but the hardware on a Crusoe chip is more advanced allowing many tricks the Intel chips couldn't dream of. And Intel provides a kludgy way to fix the CPU microcode through BIOS-loaded "patches" of which they've released quite a few to BIOS-authors already.
This could ofcourse be just the reporter not being that technologically knowledgeable and buying the Transmeta hype about "software processor" instead of bias, as well.
He could be out of a job if Transmeta doesn't turn things around.
:)
He could always go to work at Intel
Is this a foundry issue? This is one of the several weaknesses of Transmeta: that they had such a hard time getting a foundry to manufacture their chips, earlier this year, so perhaps this is indicitive of less than top-notch quality.
The P3 1.13 GHz recall affected only 200 parts (i.e. less than this Transmeta recall), but that doesn't stop AMD and TMTA stock-holding slashbots from bringing it up on a daily basis.
Of course it's brought up on a day-to-day basis. It was STUPID. Listen, Transmeta has a flaw in a few hundred chips. The *FIRST* chips Transmeta has mass-marketed. What, they get no slack? They're not doing too bad, considering they've come up with a chip that is competitive(on some grounds) with the likes of Intel. The Giant. That's nothing to sneeze at; they're doing a good job.
Now, the reason why everyone got sooo pissed at Intel is that they reached too far. They just COULDN'T stand having a slower-clocked processor than their rival, AMD. So, they did absolutely everything they could to make sure they had the highest clock.
These things included:
- Increasing voltage, a common overclocker's trick - not something chip companies should have to do on a regular basis.
- Disabling those parts of the PIII that didn't work at a high voltage(which parts escape me).
- "releasing" a processor, even though they could, at MOST, deliver a few hundred at press-time. I mean, this is the worst of it. They couldn't beat AMD on a clock-speed basis, so they rounded up a few hundred PIIIs that with a lot of work(and luck) were supposed to run at 1.13GHz, just so they could say "yeah, we're faster, go suck a pole".
Arg.
Dave
Barclay family motto:
Aut agere aut mori.
(Either action or death.)
Barclay family motto:
Aut agere aut mori.
(Either action or death.)
Everybody immediately thought design flaw. That would be bad bad bad. A manufacturing glitch is not such a problem.
:)
Hey! Who manufactures these chips for Transmeta anyhow? I see a potential scapegoat here...
I honestly don't think it makes Transmeta look particularly good to mention that certain companies are notorious for treating their customers very poorly indeed when problems arise. If you think that it's OK for a company to foist defective products on the public and then not take any responsibility for them, then I wish you everything you expect.
And the brethren went away edified.
Are you serious or just trolling? Intel bought StrongARM from DS over three years ago.
By the way, why is it only the NEC laptops with the problem? Shouldn't it affect all or none of the Crusoe's?
Sheepdot: Open Source good, Closed Source baaaaaaad!
If you can't differentiate between the Transmeta twirlie and the Debian curl, you're a moron.
"Ancillary does not mean you get to rule the world." --U.S. Circuit Judge Harry Edwards, speaking to the FCC's lawyer
But this is a recall which required less power and required fewer resources than other recalls...
It could have been worse. If Transmeta were more like M$ they'd have shipped several million units before announcing the problem, and then charged you for an "upgrade" to fix the bug. If they were more like Intel they'd just deny that anyone should care, and you'd have to show that you had a good reason for caring before they'd replace the chip.
And the brethren went away edified.
The P3 1.13 GHz recall affected only 200 parts (i.e. less than this Transmeta recall), but that doesn't stop AMD and TMTA stock-holding slashbots from bringing it up on a daily basis.
Plus, consider that this is TMTA's first product: their batting average is 0.000, but Intel's is about 0.950. IMHO, this is a major blemish for TMTA which will take years to overcome, and will greatly dissuade fence-sitting OEM's who were considering using TMTA parts.
A manufacturing defect is still Transmeta's problem: it's a quality control problem. The designer of the CPU writes test vectors to test the chips. If it was indeed a manufacturing problem it means they didn't properly engineer test vectors to screen the chips. Intel's DPM is ludicrously low (500, IIRC), and with 300 defects, Transmeta would have had to ship 600,000 parts to reach Intel's quailty level (somehow I doubt they shipped 600,000 parts).
Sure, recalling 300 units wouldn't be a big deal for a nuts-and-bolts company like Fairchild. But for a company that's relies on buzz as much as Transmeta does, egg on the face is a major deal.
(Thinking that over a little, maybe I'm being too harsh -- I mean, Transmeta is trying to leapfrog Intel, not start yet another Internet pet food company, and they deserve credit for thinking big. Still, they profit tremendously from their high public profile and that's going to hurt them here.)
Completely off topic: Does it seem to other people that Slashdot is shrinking back to its pre-Columbine (pre-post-Columbine?) scale? Except for flamebait articles about the election, most stories are down to ~100-200 +1 posts. It's nice not feeling like if you don't post on a story in the first minute there's no point bothering.
Come to think of it, the continuous real-time dick measuring contest called Advogato.org has turned into a desert, at least in the posting area, and LinuxWorld (which I think is underrated) and Linux Today forums are mostly empty, too. Are we seeing a trend here?
Normally I sae this sort of stuff
Nov. 29, 10:24 AM:
from the screwing-with-carniore dept.
Nov. 29, 3:19 PM:
from the now-thats-a-wierd-thing-to-do dept.
Nov. 29, 4:00 PM:
Possible Crusoe and Recall? (uh, yeah)
And now, Nov. 29, 9:37 PM:
.. to tell us that Transmeta has confirmed that we they are recalling Crusoe .. The statements says its fewer then 300 NEC laptops, so its not that huge of a deal after all ..
This calls for the creation of the CmdrTaco Quality Seal! The first one is given to JeffK! Only awarded to sites with a minimum of spelling and grammar errors.
You might, however, also employ a grammar & spelling nazi.
He could be out of a job if Transmeta doesn't turn things around.
My sister is trying to get out of being a social worker and she can't get another job because she doesn't have Windows or Office 2000 experience.
Let's face it - Linus doesn't either.
I'm willing to put up $500 bucks to help pay the $6000 cost of MCSE training in the event Transmeta folds - how about you guys?
C'mon, we can't leave Linus out in the cold after all he's done for the open source community.
--Shoeboy
String corporateMistake( Corporation corp )
{
if ( OpenSourceCompanies.contains( corp ) ||
corp.employeesLinusT()
)
return "Things will get better";
else
return "Ha Ha " + corp.getName() + " sucks just like MS,"
" this incident is proof positive";
}
The press release makes it sound like a manufacturing defect (it only affected a single batch of chips) rather than a design defect. It doesn't sound like a huge problem. All the companies produce defective chips; they should never make it past quality control.
Compared with Intel's famous fdiv bug which was a design fault and affected all pentiums at that time, this is relatively minor.
I'd say that the biggest loss in this case would be due to the bad publicity that Transmeta has received.
Chris