Jump to content
The Inquirer-Home

AMD TDP jiggery-pokery exposed

How to make one plus two equal about one-and-a-half
Thursday, 6 December 2007, 17:54

A POSTER on a Yahoo finance board, has gone through a couple of AMD pdfs with a fine-toothed comb to unearth actual TDP figures for the chip upstart's latest selection of processors.

AMD brought in what it claims is a more representative measurement of power usage with its four-core K10 offerings. It dubs this Average CPU Power (ACP), or as the poster dubs it, "Fake-o-watts".

The firm no longer wants to talk about TDP in public as it considers the measurement irrelevant.

But it appears that in updating a a document entitled The Truth About Power Consumption Starts Here, AMD has let slip its TDP figures for the K10 core for the first time.

In what appears to be the original pdf here, AMD shows three TDP families for the chippery: 68W, 95W and 120W.

In what appears to be the revised pdf, here, AMD looks to have increased these values to: 79W, 115W and 137W.

Both claim the same 55W, 75W, and 105W Fake-o-Watt™ values. µ

See Also
Average power consumption is anything but
Readers respond to AMD power averages

Share this:

Comments
Soooo

Why make such a big deal about it now...this is old news...like 6month old.

Intel has done this for years...and i mean years... they have made unfair comparisons with amd claiming lower tdp when they were not because the "avg" was lower.

Unfortunately amd is now doing the same, i with they would have forced intel to come clean and give real numbers instead of both giving average max... 

not that the average max is a bad number to have but that it is given as tdp* 

I would rather they both give an avg tdp "max" load tdp with normal usage and MAX tdp load for 24+ hours.

Instead I now rely on independent tdp test for what the real max is and the real average is.

This article fails to say is Intel started it...
that their presshot 550j 3.4Ghz has a tdp* of 89 watts. but tends to use 150 watts on a regular basis. I have one trust me its hot for a reason. 150 is the real tdp...hmm fishy numbers

too bad amd started using them as well. at least they are closer to the real tdp. with thier independent power plane for each core.. the average can be lower...read the reviews.

dont attack one company for something it does second. especially when they are attacked on that same point... amd was almost pushed into it... if i were them I would have just made 2 numbers, avg and real then showed the world Intels true numbers...don't follow suit with number tricks.

posted by : Bryan, 06 December 2007 Complain about this comment
Standardization

The ISO or some similar standards body should produce a standard called
"measurement of electronic chips power" That way every one measures the same way. Personally I would suggest four measurement points: 100% load(TDP), 75%, 50%, 25%, and idle.

posted by : Mohammed Raei, 07 December 2007 Complain about this comment
Sooo?

Hey, this average power rating is NEW... so what is this referance to Intel using it in the past? stay with the program

posted by : Art, 07 December 2007 Complain about this comment
Merely adopting the Intel standard

Intel has been using "fake-o-watts" since before people even cared about watts. And worse: they use average values and call them "TDP" (which means manufacturers can't really rely on those numbers). AMD at least continues to supply (real) TDP numbers, even if they now add "fake-o-watts" to the "consumer" pamphlets...

posted by : Mark, 08 December 2007 Complain about this comment
Actually logical

TDP at AMD has basically meant the worst case you could possibly generate. 

That means not just pegging the CPU at 100%, but pegging the CPU at 100% with the combination of instructions that kept the CPU the most active. 

While this is a nice number to know for thermal design, it's not absolutely mandatory in the age of frequency throttling.

Intel long ago dumped this philosophy (if they ever had it) and they came up with a TDP that was for 100% CPU utilization under a typical load mix where parts of the CPU weren't fully utilized. 

It wasn't a bad thing, but it made it difficult to compare their power consumption with AMD, and, infamously, it resulted in OEM machines that would frequency throttle more often than they really should have due to under design of the thermal management.

AMD is simply following Intel's lead.

posted by : Colin, 08 December 2007 Complain about this comment
Advertisement
Subscribe to the INQ Newsletter
Sign-up for the INQBot weekly newsletter
Click here to sign up Existing user
Advertisement
INQ Poll

Christmas computer sales

Will you be buying a new computer this Christmas?