best line from a meeting today

[Jonathan]

"I don't want to be mean to people with leprosy."

[quantus]

This just begs for the question: "How the fuck did leprosy come up in a meeting?" Sure, if you were Alan, and trying to go to Med school, then I'd say, sure, ok, it could come up. BUT, you work for Intel! Was there a lepor in the fab shedding skin through his bunny suit on wafers causing yield loss or something?!

editted to add "suit"

[Jonathan]

someone was saying that fixing [this problem] was like worrying about a hangnail on a guy with leprosy. he then followed that with an aside: the immortal line above. some other wit added, "but being mean to people with hangnails is ok."

[quantus]

The 90-nm Mobile Pentium-4 uses 123 watts at full tilt (1.6 GHz), CTO Ditzel said, while Transmeta's Efficeon consumes only 17 watts at the same clock rate.

was it this problem?

Dally compared power dissipation curves in modern microprocessors with those of ASIC devices. In raw energy, an ASIC might take 2 picojoules to complete a 32-bit operation. A PC or workstation processor takes 2,000 picojoules. The culprit, Dally said, is the interconnects the processor must drive in tasks like generating an address and fetching data from an array. ASICs have scaled gate lines much more aggressively than processors, Dally said, resulting in lower power consumption.

Take that microprocessors! ASICs forever! well, at least 'til structured ASICs take over...

[quantus]

Hey Dave. Maybe you should work on this. I bet Intel would hire you for a bunch of money if you came up with a cool idea that worked like this...

Stanford mechanical engineering professor Ken Goodson described new cooling technologies. "It's a great time to be a heat transfer specialist," he said. As much as $60 million in venture funding went into microfluidic and thermoelectric cooling projects in 2004, he said, as well as a variety of research activities funded by the Defense Advanced Research Projects Agency (Darpa).

Finned aluminum heat sinks and heat pipes now being utilized with 200-watt processors are often 3,000 times larger than the die itself. "The dream heat sink," Goodson found with Darpa funding is a micro-machined pump that suppresses hot spots on the CPU surface using fluid flows to lower the pressure.

Sorry for hijacking this topic

[Dave]

using E=MCMC you convert the energy generated by the die into pie. problem solved.

[Jonathan]

The 90-nm Mobile Pentium-4 uses 123 watts at full tilt (1.6 GHz), CTO Ditzel said, while Transmeta's Efficeon consumes only 17 watts at the same clock rate.

was it this problem?

No.
http://anandtech.com/cpuchipsets/showdo ... i=2342&p=3

Intel Pentium 4 520 (2.8GHz) 84W

I work for Intel, but I do not speak for Intel. My opinions are not necessarily the opinions of Intel Corporation.

[quantus]

I dare say that those are different numbers... One (the latter) is thermal power (TDP in the article) or energy converted to heat and the other is actual power drain from the power supply... Or I could just be wrong...

[Jonathan]

All energy used in computation is converted to heat. We're not creating mass here, or driving chemical changes.

The article is just wrong.

http://www.cpuheat.wz.cz/html/Mobile_Pentium4.txt

1.6 GHz P4 consumes between 32-39W depending on the model you're talking about. Also, there is no 1.6 GHz 90-nm Mobile Pentium-4: Prescott bottoms out at a bus ratio of 14, which at 133 MHz FSB (533 effective) comes out to 1.86 GHz. Also, the idea of the comparison is wrong-headed, because Intel positions the Centrino Pentium M processors as the low power x86 processors. A better comparison is between Efficeon at 1.6 GHz and 17W and Pentium M at 1.6 GHz and 30W (or Dothan at 1.7 GHz and 27W). And even that is somewhat silly; you should be comparing performance per watt.

[Jonathan]

http://www.cpuheat.wz.cz/html/IntelPowerConsumption.htm

A list of links to the various numbers I just spouted.

I am an Intel employee, but I do not speak for Intel. My opinions are my own and not necessarily the opinions of Intel Corporation.

[quantus]

We're not creating mass here, or driving chemical changes.

You should work on this. If the energy could be used to make gold instead of heat, that would be awesome.

[quantus]

All energy used in computation is converted to heat.

What about energy used to drive chip to chip interconnect?

[Jonathan]

Also winds up as heat, just not heat dissipated by the heatsink over the chip.