Slashdot Mirror
Compaq Holds Off On Crusoe
scott1853 writes "Accroding to this article on ZDNet, based on this article at G2News, Compaq has decided to put off use of Crusoe in favor of Intel's P3. Unlike IBM, Compaq isn't claiming to put the processor on hold, they appear to have made their final decision not to use it. Could this be Intel flexing it's influential muscles, or is Transmeta not being competitive enough price-wise?"
Could it be that well established companies don't want to take a risk and use a new product from a company nobody knows anything about? Could it be that without Linus Torvalds, nobody would even bother to look at Transmeta? Could it be that the power savings from the CPU are hardly noticable in the large scheme of things when you have LCD backlights and hard drives to power?
The Crusoe chips in their current incarnations are simply x86 wannabees. They emphasise power thriftiness over performance. Therefore if Intel and AMD come out with equally power efficient chips, their advantage would be gone. Besides, as has been pointed out numerous times, the CPU is only one part in the power budget of a portable, and not even the largest one.
.NET architecture. Or develop a hybrid chip that can switch between x86 and PPC instruction sets dynamically, allowing you to run Windows and MacOS for example on the same machine at the same time. There are lots of issues that would make the latter pretty hard beside the CPU, but my point is to take advantage of your strengths to do something your competitors can't.
Transmeta should instead apply the code morphing technology in fields without ready competitors. Maybe something like a Java chip, or something fitting into Microsoft's
The problem with today's benchmarks are that they don't simulate real honest-to-goodness usage patterns.
;)
Before the Crusoe, they came close enough. Running the same program on a processor multiple times yielded the same results, so long as the program wasn't competing for resources. Those processors were in a limited way "stateless," which is to say that they don't remember anything from the last time the program ran.
Since benchmarkers didn't need to worry to much about it, the benchmarks generally run a program once, then run a different one, then a different one, etc., etc., etc..
In real life, however, a person is likely to open Word, then maybe open Internet Explorer, close Word, open Excel(which shares a lot of code with Word), close Excel, open Internet Explorer, and so on. That is *real life* usage, and Crusoe performs fairly well at it - certainly not as bad as traditional benchmarks would suggest.
When was the last time you played Quake III for two minutes? Yeah, I thought so. Fact is, when you're running on a Crusoe, those first two minutes will be much slower than the subsequent hour
So, as you can understand, Crusoe doesn't need "rigged" benchmarks - it just needs *real* benchmarks. The problem isn't that Crusoe will only perform well if the benchmarks are rigged, the problem is that Crusoe *won't* perform well unless the benchmarks put a computer under the same stresses and loads(or lack thereof) that a computer will go through on a daily basis.
Dave
'Round the firewall,
Out the modem,
Through the router,
Down the wire,
Barclay family motto:
Aut agere aut mori.
(Either action or death.)
Just like Linux is having/has had a hard time overcoming the Windows mind-share, so will Transmeta have a similar fate with regard to Intel/AMD.
You can't really expect such a big splash right away with relatively unproven technology (market wise). Just be glad someone (Sony) has given Transmeta/Crusoe a chance and let the performance of their product dictate their success. Look how long AMD took to really get a foothold in the CPU market. After all their K6's, it wasn't really until the Thunderbird Athlon that they earned their place in the market with a fantastic product.
Praying for the end of your wide-awake nightmare.