IBM Wary of Crusoe?

Angus writes "VNUnet have just posted a story that IBM is being cautious about the future of Transmeta's Crusoe in production machines. Suggestion is that Intel is still the player for the future of portables." An interesting comment at the end: 'All Intel has to do is cut prices to squeeze transmeta out of the market'

Re:Hmmm, IBM wary of their own product?

Sing after me: The laptop division isn't connected to the chip division, the chip division isn't connected to the software division, ...

Re:The road for Crusoe isn't easy

That is correct!

I think one issue Transmeta may face is the fact that Intel's latest Celeron mobile CPU's are already using far less power consumption than their predecessors.

Once Intel goes to the 0.13 micron process to make the Celeron mobile CPU, the chances are good this future Celeron variant will have the same power consumption as the Transmeta CPU.

Remember, for business applications and Web browsing, the Celeron CPU with its on-die 128 KB L2 cache is more than sufficient for most users, even in corporate environments. It's only in environments where CPU power is at a premium (high-end games, illustration programs, image-processing programs and CAD/CAM programs) where the Pentium III and Athlon CPU's really become userful.

Re:Gotta love IDC Analyst

You're so right. Here's what I've picked up from the latest c't magazine cpu-news column:

Intel has only been able to claim lower than 1 Watt consumption on their new SpeedStep CPUs trough a trick. They measure what they call "Average Power", which is supposed to simulate a realistic usage pattern. In reality, this means their cpus spend 80 to 95 percent of the time effectively turned of. Maximum consumption at 500 MHz and 1.1 Volt is still 8.1 Watt.
Now compare this to the Crusoe which is supposed to burn only 1 Watt while decoding an MP3 (1.8 for DVD decoding), and add to this that the Northbridge is _integrated_ into the Crusoe. An Intel BX chipset typically burns about 2 Watts, which would put the Intel cpu at around 6 Watt for something like mp3 decoding.

So, everything else being equal, the Crusoe is a long way ahead in terms of power consumption. Overall, since Crusoe integrates part of the mainboard chipset, it should allow for simpler mainboard layouts, which again translates into size and consumption savings (not to mention cheaper manufacturing).

I think Transmeta have a damn fine product, and Intel should really watch out if they don't want to lose a big part of the mobile market (hmmm, only the paranoid survive anyone?)...

Re:The road for Crusoe isn't easy

You're right - it does sound real close to a laptop. I think that real "post-PC devices" will take advantage of flash memory, improved LCD technology (not necessarily in the full-color, XGA-class form factor), and a mixture of chicklet keyboards (when applicable) and stylus input for navigation and editing. The other thing that can help these devices is some of the newer lithium battery technology where you can pretty much mold the battery to the available space.

The advantage to leveraging software doesn't come in just recycling Win32 cruft (or Unix cruft, for that matter). Photoshop on a tablet is probably a waste of a tablet and a waste of Photoshop as well. But a simplified sketch application that can re-use some Photoshop code on top of Unix API's might be a hit. Likewise, I'm not suggesting that Word (or StarOffice, for that matter) would be a program that should or could be ported. A tablet application would probably be far simpler, and a "terminal appliance" that fits into a PC form factor but uses far less power (and generates far less heat) might use a more robust app. But good development can allow a vendor to re-use a lot of code in between.

I don't see us running today's x86 software on these devices as is - but there is some potential, more so than if you have to write from scratch. Despite the elegance and simplicity of the Palm OS (and I'm addicted to my Palm Vx), there is very little potential to bring software over to the platform from other systems. It's not easy at all, though it is do-able. Mobile Linux at least has the potential to be better than that - code reuse doesn't necessarily mean application reuse, though. Thank heavens.

- -Josh Turiel

The road for Crusoe isn't easy

With Crusoe, Transmeta is playing in the second-most cutthroat market next to DRAM - the x86 CPU business. At one point, there were a plethora of x86 CPU's on the market. Now there is basically just AMD and Intel (I know Cyrix still exists through Via, but it's in a small niche until proven otherwise). And as much power as mobile chips burn, the other components burn as much or more power in operation, meaning that there will have to be some pretty darn compelling advantages to Crusoe for it to get design wins in traditional laptop systems. This is where Transmeta wants to play.

The caveat to this gloom is that Transmeta's chips can also be used for newer, non-laptop devices that can use different components and lower power budgets than traditional PC laptops. Mobile Linux is a factor here, too - the chips usually deployed in this product space (Motorola Dragonball - Palm, Intel StrongARM - the late Newton, WinCE, and others, Hitachi SH3 - WinCE, and a few other chips, too) have excellent performance and are generally light on power draw, but do not offer x86 compatibility. Transmeta brings that to the table with much less power usage than the AMD or Intel alternatives. This will allow companies to leverage the existing x86 software base and Unix software much easier than other platforms do. The "post-PC" market is where Transmeta will live or die.

The other thing to remember in this market is the cost factor. Intel and AMD already charge low prices for their mobile chipsets. Transmeta isn't trying to compete on a "bang-for-the-buck" basis, because they'll get slaughtered that way. Transmeta can't afford to go up against Intel directly like that, nor can they even afford to take on AMD. The answer is to "hit 'em where they ain't", to paraphrase Wee Willie Keeler, and play for the wireless, appliance, and PDA markets against those specialty chips.

It's definitely do-able, and Transmeta may get a few design wins in the laptop market as well - but any laptop chip sales are almost gravy on top of the appliance market. There's room for another player there - why not Transmeta?

- -Josh Turiel

Re:Something's not right...

All Intel has to do is cut prices to squeeze transmeta out of the market

You beat me to it. This is true, except for the fact that Intel would become the DoJ's next illegal monopoly target. It doesn't matter if there are a few other competitors such as AMD. If Intel were to slash prices until transmeta died, and then raised prices very high, the DoJ would be all over them.

I like the idea the DoJ is starting to investigate many large US corporations for aggressive tactics to hurt their competition. Certainly Intel has been under investigation several times in the last few years, and each time they have promised to change their ways. And they've sincerely changed enough after getting their hand slapped. They don't want to get hauled to court and get engaged in a fight to the death like micros~1 is doing.

Also, the knowledge transmeta has collected would not instantly cease to exist. The knowledge would resurface inside another Intel competitor, better funded and ready to do battle. So it is not in Intel's best interests to obliterate small competitors like transmeta, just try to corral them.

the AC

IBM doesn't know how to utilize Transmeta

I had always been under the impression that the reason Crusoe is such hot shit is that its power consumption is so low. So it's not blazing fast - big deal. Do you need to do 3-d rendering during your AM commute? Of course not.

Besides, handhelds are where Transmeta is really going to clean up. It seems to me like Transmeta isn't really going head-to-head against Intel. It's the ARM and Dragonball folks that should watch out.

What ever happened to the Crusoe Linux kernel that Linus was working on. Any news on that? I'm wondering - does it run in the x86 emulation mode, or in native and therefore faster Crusoe mode?

-carl

Re:Intel Shoudl Just Buy Transmeta

Ok, suppose they do... then what? My bet is that if Intel buys Transmeta, Crusoe will be forgotten. Intel is a case study in NIH, that is, 'Not Invented Here.' ARM has been in the tank for years until just recently, despite the fact that it's a great platform. Why? Intel didn't invent it, therefore they're more than a little leary about marketting it. Things like the Pentium get out because Intel feels good about designing, producing, and marketting it in house. Maybe things are changing, but if the past is any indication (and it usually is) an Intel buyout of Transmeta will mean the death or severe beating of Crusoe.

Re:"Native" Crusoe mode?

IIRC there is a JVM interpreter for Crusoe already.

Also the Alpha interpreter is much more efficient that the x86,
according to this IEEE Spectrum article [ieee.org] which was posted on Slashdot a couple of months ago.

In the short run, I think Transmeta have a hard fight on their
hands just to survive. The Spectrum article hints that Transmeta were
disappointed at the results they reported at Crusoe's unveiling, that
they had expected a real showstopper. In the long run, I am convinced
that this is a much, much better way to design processors.

Re:taking the flamebait...

confusious say:
there's alot more to speed than megahertz (or gigahertz for that matter)

when the pentium processor was released in 1993, (so it's not really a decade, but it's been long) it was supposed to be real fast, and everyone was supposed to love it. It debuted at a whopping 60 MHz. What intel didn't tell you is that the DX/4 100, and AMD's 120 MHz 486 ran faster. The pentium was a flawed chip, but everybody still wanted one, because it was new. Ever notice how linux will run on a 386? that's not a mistake. All current x86 processors have their root in the 386, intel's first 32 bit mainstream processor. After the pentium was released, intel released the Pentium Pro. It was a fine chip in it's day, but it was basically a pentium with on chip L2 cache and higher clock speeds. Not much of an innovation. Then came the Pentium MMX. 57 new multimedia instructions for the Pentium core. Intel promised that this would make all of our lives better because code would run faster. If the code in question was MMX optimized. That's a big IF. So then came the Pentium II. The pentium 2 was a pentium pro with MMX and L2 cache that ran at half processor speed. The pentium 2 was a one step forward and two steps back from the pentium pro. But people liked it because it was cheaper! Then came the pentium 3. The pentium 3 should never have had it's own product name. It should have been called a pentium 2. Now, the pentium 4 does have potential to be a good chip, but Intel is just killing it with it's senseless branding techniques. And don't get me started on Merced. Intel has been blowing smoke up our asses about merced for the last 4 years. I remember when they were saying that the Merced would be a "64 bit processor with speeds of up to 600 MHz." Intel's roadmap doesn't leave any room for it's pissing contests.

Hmmm, IBM wary of their own product?

I copied this off the IBM intranet:

*****

Transmeta Unveils Copper Processor Manufactured by IBM
Crusoe(TM) Chip Poised to Revolutionize Mobile Internet Computing

January 19, 2000

Transmeta Corporation today ended four-and-a-half years of secrecy with the introduction of Crusoe(TM), the world's first family of smart microprocessors. In a foundry relationship with Transmeta, IBM is manufacturing the new Crusoe chip in copper at MD's Burlington, Vt., site.

Transmeta's Crusoe processor family is based on a breakthrough software approach that will deliver on the market's need for "all day computing" with a PC-compatible, high-performance solution with low power.

*****

So, IBM is wary of a product they produce? I don't think so.

Gotta love IDC Analyst

"Intel's Speedstep technology allows lower clock speed and lower wattage. Surely what Transmeta is offering is not radically different to that. It will all depend on pricing," he said."

This goes to show how much Mr. Brown knows about the technical aspect. He seems to make it sound like all Transmeta has done is make a slimmed-down version of a Pentium. He is oblivous to how the Cruose works, and why it runs at lower power consumption levels than an Intel chip.

---

This is outragous

I can't believe what I am seeing.

Transmeta never WANTED to be a "production machine". Crusoe was designed to be a low power, close to Intel speed processor. Made for Portable products. Not something that you'll keep on your desktop ..

People need to stick to what the crusoe was designed for.. quit trying to throw it in desktop machines..