4 Phase Power

[Operandi]

I?ve noticed a lot P4 based motherboards featuring four phase power designs for the CPU, while not many/any for the Athlon. The power consumption between the P4 and the Athlon always seemed pretty comparable. Is there any reason why there are more P4 based systems with 4 phase power then the Athlon? Are P4 motherboards built to higher standards or is it not as big of a necessity for Athlons?

 

[lexxmac]

Yes, most likely in anticipation of prescott and its enormous power needs. AMD has so far not put out a chip, as far as I know, that exceeds 100 watts of heat dissapation, whereas prescott CPUs put out that much or more, depending on the clock speed, voltage, and so on.

 

[AsiLuc]

Check out these links:

AMD K7
AMD k8
Intel P4

Athlon 64: 89 Watt Max.
Pentium 4 3.4E: 103 Watt Max.

Remember Cool'n'Quiet? Most of the time the AMD are cooler.
That's a significant difference.

 

[Cogman]

LOL!!! Man I remember just a couple of MONTHS back that people would not go to AMD because they ran hotter then Intel and where harder to cool. This is just funny, for me, to see the tables turn.

 

[AbsolutDealage]

Originally posted by: Cogman
LOL!!! Man I remember just a couple of MONTHS back that people would not go to AMD because they ran hotter then Intel and where harder to cool. This is just funny, for me, to see the tables turn.

For me it was never that the AMDs were harder to cool, it was the fact that if your cooling failed you were SOL.

 

When you say 4 phase power, you can't be talking about AC power (as in 3 phase), what exactly is 4 phase power in relation to DC that the uP requires?

 

[Operandi]

Originally posted by: AsiLuc
Check out these links:

AMD K7
AMD k8
Intel P4

Athlon 64: 89 Watt Max.
Pentium 4 3.4E: 103 Watt Max.

Remember Cool'n'Quiet? Most of the time the AMD are cooler.
That's a significant difference.

Nice link, but that?s still only 14 watts difference between Intel and AMD's fastest. Different topic, are there any AMD boards with 4 phase designs?

 

[Bassyhead]

Originally posted by: Operandi

Originally posted by: AsiLuc
Check out these links:

AMD K7
AMD k8
Intel P4

Athlon 64: 89 Watt Max.
Pentium 4 3.4E: 103 Watt Max.

Remember Cool'n'Quiet? Most of the time the AMD are cooler.
That's a significant difference.

Nice link, but that?s still only 14 watts difference between Intel and AMD's fastest. Different topic, are there any AMD boards with 4 phase designs?

i'm pretty sure i've seen one or two before, but cant remember what the boards were off the top of my head

 

[IvanAndreevich]

Marketing IMO. ASUS A7N8X gets away just fine with 2-phase power. However, more is not less. Can't argue against it really.

 

[HappyCracker]

Gigabyte's "Dual Power System" kicks it up to four phase if memory serves. Could be wrong though.

 

[Operandi]

Originally posted by: HappyCracker
Gigabyte's "Dual Power System" kicks it up to four phase if memory serves. Could be wrong though.

No, your right it dose I forgot about that board.

I also completely forgot I have a friend who just graduated with electrical engineering degree. I'll see what he says.

 

[AsiLuc]

Some links:
AMD newspost Oct 2002
Overclockers.com
God bless you!!!
my querie

 

[Spencer278]

What does Phase even mean when refereing to power and is it better to have a high phase number.

 

[PrincessGuard]

For DC-DC conversion, it means the number of regulators connected in parallel to the output. A controller switches between each regulator to balance the current load so one single converter doesn't overheat. They use some sort of feedback mechanism to stabilize the output and reduce noise generated by the load and by the regulator itself.

The greater the number of phases, the less strain you put on each output driver but the larger the number of components you need to have. Having more phases also inherently increases the ripple frequency, which is actually nice since higher frequencies are more easily filtered by the output inductors.

 

[imported_rustybx]

Originally posted by: PrincessGuard
Having more phases also inherently increases the ripple frequency, which is actually nice since higher frequencies are more easily filtered by the output inductors.

Incorrect. Interleaving converters does not change the inductor frequency. The inductors see the same frequency regardless of number of phases. However, ripple current in the output capacitors does have a higher effective frequency with interleaving. This leads to lower ripple current in the output capacitor. For a given ripple voltage this allows fewer and smaller capacitors.

 

[Mark R]

There is another advantage due to multi-phase conversion - the out-of-phase inductor ripple currents actually cancel in the output capacitors. This means that the capacitors can be even smaller than required for the apparent operating frequency.

Additionally, some multi-phase controllers offer a rapid-response feature. Under a sudden increase in load, they can increase the power delivery through the next phase to trigger, rather than waiting for the entire cycle to complete. This can improve transient response and can approximate the performance of a single phase converter with a higher operating frequency.

 

[imported_rustybx]

Originally posted by: Mark R
There is another advantage due to multi-phase conversion - the out-of-phase inductor ripple currents actually cancel in the output capacitors. This means that the capacitors can be even smaller than required for the apparent operating frequency.

To be precise, the currents completely cancel only at specfic duty cycles (Vo/Vin) depending on number of phases. Further, cancellation depends on current sharing between the phases.

With spiffy controllers the sharing can be quite good...but generally not better than 5-10%. Without balancing circuits the phases can be off by 3:1 or more.

At low duty cycle there's essentially no cancelation. At 50% duty you have total cancellation for 2-channel. Cancellation at 33% and 66% duty for 3-phase and 25/25/75% for 4phase. With a 4-channel the maximum ripple current is ~20% of the single channel value for duty cycles higher than 20%.