Does anybody know how much power the 90nm K8 dissipates? The reason I
want to know is, there's a .37C/W HSF (heatsink/fan) available whose
noise level really is 21dBA.

A Thermaltake subsidiary(?), TR2TT, has a 21dBA 32.4CFM 2500rpm 80mm
fan that they use in 4 HSF models:

AMD K7 TR2-M3 (TT A4004D) .48C/W
P4-478 TR2-M5 (TT A4005) .48C/W
Prescott TR2-M12 (TT A4012) .42C/W copper
AMD K8 TR2-M6 (TT A4006D) .37C/W copper
[TT = Thermaltake]

The P4 model is hard to find. But Froogling the TR2TT part # or the
TT part # will reveal single-digit pricing (US$) on the others. And I
know of other single-digit prices that don't show up on Froogle.

A long time ago in a galaxy far away, I spent 8 years running a
company whose business was building environmental acoustic noise
monitors - sound level meters with a built-in microprocessor for
unattended statistical data gathering. I have some real-world
experience in measuring low noise levels.

So I didn't believe it when I saw HSFs speced at 21dBA. I bought one,
the M3, to disprove it. I was wrong. Late at night I turned off
everything in my house (not apt) and used a 12V lead-acid battery for
power. At 6 inches, I could barely detect the fan noise. Then the
refrigerator came on, 20' away around a wall, and I could no longer
hear the HSF even at 6" (quiet refrigerator).

I'm guessing that automated fan/rotor balancing is used to achieve
this low noise level. Other makers have 21, 20 and even 19 dBA
fan-based HSFs, so the technology (which is really fan technology) is
becoming widely available.

These fans don't move a whole lot of air, so they can't be used -
yet - to cool a 92W .13u Athlon64. But I think the 90nm version, now
available across the counter, can get by with a .37C/W HSF. Can
anybody confirm that with actual power levels?

Felger Carbon wrote:
> Does anybody know how much power the 90nm K8 dissipates?

That is a moderately broad question. How fast and how much L2?

By my own measurements, a 2 GHz Opteron 246, E4 stepping,
dissipates about 65 W and a 2.6 GHz Opty 252, E4, is about 77 W.

(My measurements method was to take the difference between the
power drawn by a 2P system with both CPUs running and the same
system with one CPU disabled. I assumed PSU efficiency of 65%.)

At the same speeds, the Athlon FX and the 1 MB L2 versions of
Athlon64 chips at the same as the Opteron thermally. 512 KB L2
versions of the Athlon 64 should use a little less.


There are also low voltage (and very expensive) "HE" versions
advertised as 55 W max for the Opty x46, and x48.


The reason I
> want to know is, there's a .37C/W HSF (heatsink/fan) available whose
> noise level really is 21dBA.

FWIW, I have tried the AlCu versions of the Zalman CNPS7000A and
the CNPS7700 in Opty dualies using the 246, 250 and 252, and in
single Opteron systems using the 148 and 150.

They do the job and run *very* quietly.

Beware of the large size and weight of the 7700. If you have any
doubt about your ability to squeeze one into your system, get the
smaller 7000 instead.

http://www.zalmanusa.com/usa/product...x=142&code=005
http://www.zalmanusa.com/usa/product...x=146&code=005

I have never tried the pure Cu versions of those fan/sinks:
their huge weights intimidate the hell out of me.

>
> A Thermaltake subsidiary(?), TR2TT, has a 21dBA 32.4CFM 2500rpm 80mm
> fan that they use in 4 HSF models:
>
> AMD K7 TR2-M3 (TT A4004D) .48C/W
> P4-478 TR2-M5 (TT A4005) .48C/W
> Prescott TR2-M12 (TT A4012) .42C/W copper
> AMD K8 TR2-M6 (TT A4006D) .37C/W copper
> [TT = Thermaltake]
>
> The P4 model is hard to find. But Froogling the TR2TT part # or the
> TT part # will reveal single-digit pricing (US$) on the others. And I
> know of other single-digit prices that don't show up on Froogle.
>
> A long time ago in a galaxy far away, I spent 8 years running a
> company whose business was building environmental acoustic noise
> monitors - sound level meters with a built-in microprocessor for
> unattended statistical data gathering. I have some real-world
> experience in measuring low noise levels.
>
> So I didn't believe it when I saw HSFs speced at 21dBA. I bought one,
> the M3, to disprove it. I was wrong. Late at night I turned off
> everything in my house (not apt) and used a 12V lead-acid battery for
> power. At 6 inches, I could barely detect the fan noise. Then the
> refrigerator came on, 20' away around a wall, and I could no longer
> hear the HSF even at 6" (quiet refrigerator).
>
> I'm guessing that automated fan/rotor balancing is used to achieve
> this low noise level. Other makers have 21, 20 and even 19 dBA
> fan-based HSFs, so the technology (which is really fan technology) is
> becoming widely available.
>
> These fans don't move a whole lot of air, so they can't be used -
> yet - to cool a 92W .13u Athlon64. But I think the 90nm version, now
> available across the counter, can get by with a .37C/W HSF. Can
> anybody confirm that with actual power levels?
>
>

"Rob Stow" <(E-Mail Removed)> wrote in message
news:3WV8e.1068088$6l.371871@pd7tw2no...
>
> FWIW, I have tried the AlCu versions of the Zalman CNPS7000A and
> the CNPS7700 in Opty dualies using the 246, 250 and 252, and in
> single Opteron systems using the 148 and 150.
>
> They do the job and run *very* quietly.

But they don't come with single-digit price tags, as I recall. ;-)


> Beware of the large size and weight of the 7700. If you have any
> doubt about your ability to squeeze one into your system, get the
> smaller 7000 instead.

The Zalmans are magnificent if you have lots of money and space. The
new, quiet HSFs are for folks who are limited in at least one of
those. ;-)

But only if the CPU power isn't too high.

On Mon, 18 Apr 2005 22:36:00 GMT, "Felger Carbon" <(E-Mail Removed)>
wrote:

>"Rob Stow" <(E-Mail Removed)> wrote in message
>news:3WV8e.1068088$6l.371871@pd7tw2no...
>>
>> FWIW, I have tried the AlCu versions of the Zalman CNPS7000A and
>> the CNPS7700 in Opty dualies using the 246, 250 and 252, and in
>> single Opteron systems using the 148 and 150.
>>
>> They do the job and run *very* quietly.
>
>But they don't come with single-digit price tags, as I recall. ;-)
>
>
>> Beware of the large size and weight of the 7700. If you have any
>> doubt about your ability to squeeze one into your system, get the
>> smaller 7000 instead.
>
>The Zalmans are magnificent if you have lots of money and space. The
>new, quiet HSFs are for folks who are limited in at least one of
>those. ;-)
>
>But only if the CPU power isn't too high.

But how often are you going to run it full tilt and will a little noise
bother you there? With Cool 'n' Quiet and a mbrd which controls fan
speeds, the idling, or even moderate use, temp stays low - my A64 3500+
90nm runs at 1004MHz (<40C... about 6C higher than system temp) most of the
time with the fan at 2200rpm. Even with the boxed HSF, the "system fan",
which funnily enough, doesn't seem to be as controllable speed-wise, makes
more noise than the CPU fan.

My Antec Sonata case is also temp controlled for the P/S and 120mm case fan
so the whole thing is pretty quiet. The worst thing is the POS fan on the
video card which occasionally makes noises like a bad bearing... and with
everything else so quiet, it's really aggravating.:-)

--
Rgds, George Macdonald

Felger Carbon wrote:

> Does anybody know how much power the 90nm K8 dissipates?

According to the Tech Report, approximately 30 W less than
the equivalent 130 nm part.

http://techreport.com/onearticle.x/7417
AMD 90nm power consumption measured

On Mon, 18 Apr 2005 22:36:00 +0000, Felger Carbon wrote:

> "Rob Stow" <(E-Mail Removed)> wrote in message
> news:3WV8e.1068088$6l.371871@pd7tw2no...
>>
>> FWIW, I have tried the AlCu versions of the Zalman CNPS7000A and
>> the CNPS7700 in Opty dualies using the 246, 250 and 252, and in
>> single Opteron systems using the 148 and 150.
>>
>> They do the job and run *very* quietly.
>
> But they don't come with single-digit price tags, as I recall. ;-)

Your 90nm Sempron has a 128K L2, if I read things right. Isn't that a
little anemic? I went with the Opteron vs. Athlon 64 to get the 1MB L2.

>> Beware of the large size and weight of the 7700. If you have any doubt
>> about your ability to squeeze one into your system, get the smaller
>> 7000 instead.
>
> The Zalmans are magnificent if you have lots of money and space. The
> new, quiet HSFs are for folks who are limited in at least one of those.
> ;-)

I may look into a Zalman, once the CFO gets over all the other charges on
the card. (*very* expensive two months)

> But only if the CPU power isn't too high.

I know high. You aren't high! ;-)

--
Keith

"keith" <(E-Mail Removed)> wrote in message
news(E-Mail Removed)...
> On Mon, 18 Apr 2005 22:36:00 +0000, Felger Carbon wrote:
>
> > But only if the CPU power isn't too high.
>
> I know high. You aren't high! ;-)

I _know_ I'm not high, Keith. Don't smoke any veggie matter. And my
diet doen't allow hootch, dammit. Oh, you're talking about the
Sempron 90nm power level? How did the Sempron become _you_? ;-)

Seriously, the .13u AMD equivalent comes in (according to Sandpile) at
66W max. I'm hoping the 90nm that I got shipping notification on an
hour ago dissipates in the 40s someplace.

Grumble wrote:

> Felger Carbon wrote:
>
>> Does anybody know how much power the 90nm K8 dissipates?
>
> According to the Tech Report, approximately 30 W less than
> the equivalent 130 nm part.
>
> http://techreport.com/onearticle.x/7417
> AMD 90nm power consumption measured

Here is another article. (Unfortunately, it is in French.)

http://www.x86-secret.com/articles/d...so/conso-4.htm