dothan tdp

[someaznguy]

i've been doing some research on the dothan and turion. i know there's 2 flavors of turions, 25w and 35w. dothans have a tdp of 27w.

now, according to this review, the dothan's 27w tdp is "maximum power dissipated at 75% of the maximum power for a given frequency."

i was wondering if someone could explain that in english or if anyone knew the tdp of a dothan while being stressed to its max, so i may compare it to the turion a little easier.

 

[pxc]

now, according to this review, the dothan's 27w tdp is "maximum power dissipated at 75% of the maximum power for a given frequency."

That statement is based on a very wrong assumption. It's lazy, too.

Intel defines TDP as:

The processor’s power is specified as Thermal Design Power (TDP) for thermal solution design. TDP is defined as the worst-case power dissipated by the processor while executing publicly available software under normal operating conditions, at nominal voltages that meet the load line specifications. The TDP definition is synonymous with the Thermal Design Power (typical) specification referred to in previous Intel datasheets. The Intel TDP specification is a recommended design point and is not representative of the absolute maximum power the processor may dissipate under worst case conditions.

Cutting through the highlighted statement:
"worst-case power dissipated" - clear enough
"while executing publicly available software" - if someone knew how to write a power virus, the processor could use more than TDP power. Still this does not mean the processor would use the simplistic Idd_max * Vdd_vid (see below).
"under normal operating conditions" - Tcase < Tcasemax (the processor is not throttling)
"at nominal voltages" - Vdd <= Vdd_Vid (the voltage is normal)
"that meet the load line specifications" - Idd = Idd_max (the chip is getting the required current)

The last 3 items are the *exact* same conditions that AMD measures TDP. But AMD does use "theoretical maximum power" in their datasheet (on Athlon 64, the core + I/O power).

Getting back to why someone would make a claim of "75% of the max power". The answer: they didn't read the power design guide. Idd_max * Vdd_vid does not accurately reflect how much power the chip uses for 2 reasons: 1) the chip won't ever use the theoretical max power due to real world software restrictions, and 2) the power design guide clearly shows Vdd decreasing as Idd increases [IOW, the chip may use either Vdd_vid (Idd << Idd_max) OR Idd_max (Vdd < Vdd_vid)].

 

[savantu]

Just to add so pxc's excellent post , Intel's TDP= 95% typically of max power.

Aka TDP=129w , max power 136w ( single core Xeon MP 3.33 ) or TDP=165w and max 173w ( dual core Xeon 3Ghz )

It is impossible to reach TDP level using commercial avaiable SW.Only Linpack Benchmark manages to hit TDP level while to go over it you need a thermal virus. ( and CPU-Burn / Prime95 are close to such a thing from what I've heard )

 

[pxc]

I just wanted to add what a "power virus" is:

linpack, prime95, etc are not power viruses. Prime95 and other CPU stress programs do load the processor at 100%, but a true power virus manages to use nearly every resource of the CPU at once (I/O, all execution/decoding units, heavy cache flush/reload, etc). There is no meaningful software that does that. In fact, there are no publicly released "power viruses" (not a virus in the same way as a self-replicating/infecting program). CPU manufacturers have probably managed to construct such things for testing, but they don't exist in real software.

 

[widefault]

The other "issue" is that TDP is the same for an entire clock range of chips based on core and bus speed. 400MHz bus B1-core Dothans are 21 watt, 533MHz C0-bus chips are 27 watts regardless of the actual clock speed. Doesn't take much to realize that while the 2.26Ghz C0 obviously has a higher TDP than a 1.6GHz in either core/bus, the 1.6GHz C0 and B1 can't possibly be 6 watts apart. It's a guideline, period. Throw in EIST and it means even less to me.

BTW, the Celeron-Ms with a Dothan core are listed as a TDP of 21 watts without regard to stepping, so that article's claim that the revised core caused the jump is misguided at best.

 

[someaznguy]

thanks for the explanations.

so would a dothan and a 25w turion be fairly similiar in heat output?

 

[omega-x]

short answer, the dothan's have speedstep. the turions have whatever AMD does.

a dothan at 600mhz uses damn near NO power (3-4? watts, more or less)

 

[SLee]

thanks for the explanations.

so would a dothan and a 25w turion be fairly similiar in heat output?

This review would say no:
babelfish.altavista.com/babelfish/trurl_pagecontent?lp=fr_en&trurl=http://www.matbe.com/articles/lire/251/le-turion-dans-les-desktops/page4.php

A P-M 2.13 has a difference of 17w between full load and stepped down idle. The 25w TDP 2.2 Turion MT has a 26w difference, the 2.2 Turion ML has a 42w difference.