Heat on in Athlon-Pentium 4 power chart

OUR INBOX IS SIZZLING after Eva Glass published a piece of competitive analysis AMD is offering its customers to diss Intel.

We find that matters of heat often cause more friction than you'd think propelled by fans arguing the for and agins.

So here's a brief selection of the letters Eva's and the INQ have received since we posted the story, which reproduced faithfully AMD's own chart.

Include Tcase
hi,
i think you need to add Tcase (Theatspreader) for the intel parts for comparison. amd parts don't have a heat spreader, so system designers must design for Tdie for AMD, but they design to Theatspreader for Intel parts.

Thermal specs for Northwood
Iust read the article on The Inquirer about AMD vs. Intel thermals. I've been researching this lately, and I've found the page below that lists all the thermal specs for all P4, with both pin arrays, including the new Northwood CPU. Just thought i'd pass it along. Intel URL

You state "we'd be interested in seeing how things compare on the Northwood 478 pin shrink". Some of this info isn't too hard to find:

This URL has the Thermal Design Power and Tcase (max) for the Northwood P4s. What I can't find is maximum power numbers. Note that the column labelled "Tdie" in your table for the socket 423 P4s exactly matches the figures given by Intel for Tcase at this URL.

Anyway: Northwood 2000: Design Power: 52.4 W Tcase: 68 C

Northwood 2200: Design Power: 55.1 W Tcase: 69 C

Eva has major problem
There is a MAJOR problem with your AMD power article. The figures you quote are max power consumption for AMD, and average power for Intel. They are about 33% off. Go to sandpile and check it out, the difference is about 30w in some cases, no laughing matter. If you did get the #s from AMD, give the guy my address, or me his, and I'll set him straight. Either way, someone bought the intel FUD.

See Sandpile

The P4 #s you quote are from the 'typical power' use section,. which you can find here.

The Athlon #s are from the 'maximum power' section. on an AMD 1600/1900+, the difference is ~7w, no huge deal. On the P4, the difference is ~23w, or quite a bit. you are looking at 69/92w for intel, 61/68w for AMD. Even if you take the lower of the 2 sets (typical consumption), AMD wins by 10%.

AMD sleight of hand
Eva

First off, Northwood's power consumption is WAY below that of Willamette, so you are correct - those numbers would be totally different. 35W at 1.6 GHz is anticipated for the mobile parts - close to half of what AMD's claiming here.

Secondly, this chart is pretty funny sleight-of-hand on AMD's part. Since they are kicking around “PR ratings” instead of MHz, it looks at first glance that AMD has an advantage in power - until, that is, one realizes that performance ratings are pretend, made-up numbers. Look at the real numbers:

The Athlon XP “1900” is actually a 1.6 GHz part - so it should be compared to Intel's 1.6 GHz P4. That means 68W for AMD, 61W for Intel. Same for the 1.533 GHz “1800” - 66W AMD vs. 58W (at 1.50 GHz) Intel.

Why use typical vs. maximum power for P4? The maximum thermal power from Intel is applicable only during a power virus - which manages to switch on most of the 42 million transistors on-die. Typical power is not mentioned by AMD since they have no way to guarantee that the CPU power will remain at that level by use of a thermal diode, sliding clock frequency - as the fires on Tom's Hardware Page demonstrate.

During normal operation, regulating features of the P4 can hold power output to practically any level you like. These features work - they don't depend on external 3rd party support, nor the PR department's excuses. How else can a ~70W processor run only 85 degrees with no heat sink at all? µ