Here’s the deal. Transmeta, the most secretive chip designer on earth, has finally released its mobile processor family. This is not your father’s Oldsmobile. This is not even your Oldsmobile.
The Crusoe runs its own native instruction set, but you’ll never see it or hear of it. Why? The Crusoe’s “Man Friday” is the decoder, a special invisible program layer that turns x86 instructions into Crusoe code. As if that weren't neat enough, the decoder also optimizes the code the more often it sees something. In other words, the longer you run an application, the faster it gets.
Reason number one
The decoder can be switched out via ROMs, so you could swap a chip and go from an Intel x86 compatible to a Mac-, SPARC-, or MIPS-compatible chip. The decoder could also, in theory, be loaded with hardware-specific software to enhance many common applications. Cool, no?
But it gets better. There are two chip ranges, the 3120 and the 5400. The 3120 is intended for small, embedded systems. It runs at a speed from 333-400 MHz on a 0.22 micron die. It has an integrated North Bridge and SDRAM memory controller. It has a 96 Kb L1 cache and no integrated L2 cache. Current estimates say the 3120 will be on par with a Pentium II 300, but that’s still a guesstimate. Costs on the 3120 are intended to be low, in the $65-$85 range.
Need more power?
The 5400 will be released midyear for use in “ultralight” laptops. It runs at a speedy 500-700 MHz and is equipped with 128 Kb L1 cache, 256 Kb L2 cache, a DDR memory controller, and North Bridge on a 0.18 micron die. The 5400 is also equipped with “Long Run” power management that alters power needs at a rapid pace: Expect it to switch power levels between keystrokes.
If it works, LongRun will prove to be far superior to Intel’s SpeedStep technology. According to folks at Transmeta, the 5400 will be equivalent to a 500 MHz PIII. Look for the 5400 to debut at around $120 midyear.
Reason number two
Don’t expect to see these on desktops. The Crusoe isn’t designed to have the raw horsepower of the desktop workstation because it isn’t intended to consume as much power. The Crusoes are low, low, low power devices with correspondingly low heat outputs. Remember, if the heat output is low, you don’t need a fan and you manage to further cut a system’s power requirements.
Here’s a quick comparison chart of various mobile or embedded processors competing with the Crusoe (the K62 Embedded is a new chip, just released on Jan. 19, so the information available is limited to the official press release until a datasheet is put online):
|3120/5400||333 MHz||500 MHz||400 MHz||350 MHz|
|Crusoe||Mobile PII||Mobile PIII||Mobile K6-3||K62 Embedded|
From this, you can see that a Crusoe chip should be able to increase battery life of portable devices by a factor of 4. In many cases, it might be greater, depending on the power savings achieved by reducing cooling needs and eliminating the requirement for a North Bridge chip. (Remember, it’s integrated on a Crusoe.)
Reason number three
Administrators should find their users enjoy the increased battery life. Additionally, admins using laptops for remote administration will enjoy the same benefits.
There are a few unknowns that might challenge Crusoe. Via has its Joshua chip, based on technology acquired from IDT and Cyrix, while National Semiconductors has its Geode chipset, which also integrates a video system. The StrongARM chips are also potential contenders, given their low power needs, but their non-x86 nature prevented their inclusion in this particular evaluation.
Regardless, we should soon see a crop of light, portable, long-lasting devices running a variety of relatively complex software. Maybe, just maybe, you’ll be able to play Quake on your cell phone while waiting for a plane.
And, since I know you’ll ask, Linus Torvalds (yes, that Linus Torvalds) spent his time at Transmeta developing Mobile Linux.
Reason number four
This Debian-based package will support the special needs of embedded devices, palmtops, and other ultra-light systems. You should see the source code released sometime in the next month or two as Linus gets a chance to work out bugs and complete the feature set of the final production versions.
Oh, and don’t expect to see a Debian IPO any time soon: Debian is a group of individuals, not a company, so Transmeta avoided any greed-based battles with its package choice. But don’t worry; since the code will be released, you can expect your favorite Linux flavor to have a Mobile variant soon enough.
James McPherson is a network administrator for a nationwide ISP.If you'd like to share your opinion, please post a comment below or send the editor an e-mail.