The Inquirer reports that Intel will begin using the Centrino system for the desktop.
I use the term “Centrino system” because the terminology used by Intel is a little confusing. The CPU used in Centrino is not called a Centrino; it’s called a Pentium M (which is not to be confused with the earlier Pentium 4-M. Centrino is actually a term used to describe a CPU/mobo combination which together provide a set of features.
If you’re interested in news you can use, this is actually a bigger story with more immediate impact than Intel saying anything about x86-64 being included in next year’s processors.
Hmmm. We’re selling server chips as gamer chips and notebook chips as desktop chips. I guess anything’s better than selling the new desktop chips as desktop chips. 🙂
Anyway, why is this interesting? Well, a Pentium-M runs faster clock-for-clock than a regular Pentium IV. More importantly for these times, a 1.7GHz P-M chews up about 30 watts max (page 25 of the datasheet), which is barely enough to edge the edges of a Prescott warm.
This is great news for the noise-challenged.
The Quiet Crew
Over the past 12-18 months, a niche of computer users interested primarily in quiet machines has emerged.
Some are people who need as quiet a machine as possible for their work. Musicians tend to want this; no one has ever liked the beat of a buzzing fan.
Others are people who need as quiet a machine as possible in order to work. Some have always been very sensitive to noise. Others are victims of PTSD (Post-Traumatic Seven-k fan Disorder).
The term is semi-facetious, but the problem is no joke. PTSD victims bought high speed fans to cool their CPUs, couldn’t hear themselves think, and found themselves sensitized to noise afterwards.
A desktop Pentium M means a lot of computing power for not much electrical power, and can be throttled down a lot to make it even quieter.
The Pentium M is certainly not “beat the FX” material, but properly gussied up with current desktop mobo technology (i.e. 800MHz FSB, dual-channel RAM) ought to perform perfectly well for all but the most demanding tasks.
Whether they’ll run or not on current mobos is unknown. Given past results with the earlier Pentium 4-M, it is likely to run, but not with all the flexible power features.
We’ll probably see modified desktop mobos equipped to handle Centrino’s extra features shortly.
I looked around for any reports of overclocking, but couldn’t find any. The Pentium 4-M could overclock a lot, but the Pentium M represent a rather different design, and probably won’t do as well. I wouldn’t even think about getting one to run at 3GHz.
They are fairly expensive, with the lowest speed (1.4GHz) Pentium M costing a bit over $200, but when you’re out to minimize the bang (OK, buzz) for the buck; cost can’t be the highest priority.
It will be a new front in the high-performance war, not major, but interesting.
One Intriguing Tidbit
The Inquirer article also mentions the possibility that Intel might take four of these chips and put them together on one die next year.
While you probably shouldn’t take that statement to the bank (and the Inquirer doesn’t suggest that you should), it does indicate a probable future trend in CPUs away from single high-speed processors and towards banks of low-speed processors working together.
If you had a bank of four 1.7GHz Pentium Ms working together today, with software optimized for that (and that’s the real problem with this route, getting software developers to do that); it would blow the doors off anything currently out there. Then again, it would do poorly running software that isn’t (or can’t) be properly optimized for such operation.
Presuming this would be done at 90nm (and assuming Intel gets its act together and can get some power reduction from the process shrink); the total power consumption would be no more than what we have with a current Northwood.
Could AMD pull the same trick? Most likely.
This probably the next major advance in CPU technology. It may be a bit much (or more likely, the first ones will cost too much) to expect at 90nm, but multiple cores ought to be affordable and fairly practical when Intel and AMD go to 65nm.
In short, expect to see and buy one in about two years.