Motherboard power, 8-phase, 16-phase power?

[GundamF91]

I'm not sure how the phase power affects the motherboard performance. Does it really matter? What "phase" power is a Abit IP35?

 

[Heidfirst]

IP35 is 4 phase analogue.
& no, it doesn't really matter on it's own - overall quality of design (however many phases) & components do.

 

[beray]

Originally posted by: GundamF91
I'm not sure how the phase power affects the motherboard performance. Does it really matter? What "phase" power is a Abit IP35?

The greater number of phases is the direct indicator of the greater amount of power available. All else equal, 8-phase had 1/2 the power of 16-phase.

"Multi-rail power supplies are more reliable and more dependable than Single-rail power supplies, even when used as Single-rail power supplies."

The reliability and dependability increased with the number of channels, in this case instead of 160A being forced through all the part components of a Single-rail, each of the 16 channel only had to withstand 1/16th or 10A per channel.

In the days prior of Intel P3 (30W class CPU), multiple phases were not an absolute neccessity hence fewer mobos in those days had multiple phases. Most mobos made do with just one phase, they were the cheapest to make.

The days of single phase mobos had come and gone, they're no longer viable. You'll only find them in custom low grade, low power demand jobs.

 

[kwantam]

Originally posted by: beray
The greater number of phases is the direct indicator of the greater amount of power available. All else equal, 8-phase had 1/2 the power of 16-phase.

Okay, maybe "all else equal" this is a true statement, but in the real world any sane person designing a 2N-phase supply will be using components each rated at half of those that would be used in an N-phase supply satisfying the same specifications.

When you designing a power supply, the important figures of merit include output power, output ripple, line regulation, load regulation, efficiency, cost, size, step response, device stress, et cetera. Using a multi-phase converter allows you to reduce output ripple without increasing the output capacitor (more cost, slower step response) or the switching frequency (lower efficiency, higher device stress, i.e., more cost).

With a single phase buck converter (for example), you have a pair of switching devices (transistor and diode or two transistors), an output inductor, and an output cap (you have many other components as well; I'm talking about the most fundamental components in the circuit, so we will ignore input filters, gate drivers, et cetera). For a given output current, the voltage on the output cap will droop by a given amount each cycle. We can calculate this as follows:

Qload = C * Vdroop
Iload = Qload / T = Qload * Fsw
Thus, Vdroop = Qload / C = Iload / (C * Fsw)

So we can reduce the cycle-to-cycle droop (i.e., ripple) by increasing Fsw or C. Now, increasing the output capacitor increases the cost of the design, and slows the transient response (important if you want to be voltage-agile to save power at low load but still respond quickly to step increases in demand). Increasing the switching frequency increases the switching and gating losses, resulting in lower efficiency and more power dissipated on your switching devices, driving up the heat production and requiring bigger devices, heat sinking, or the like.

What if instead of having a single switcher we had two running at the same frequency but half a cycle apart? Now while one is dumping charge to the output, the other one is grabbing more from the input, and vice-versa. Of course, now we have to use two pairs of switching devices as well as two inductors, but the output capacitor is shared. Now Vdroop is reduced because the output capacitor is charged twice each cycle of Fsw. Moreover, since we now have two switchers operating in parallel, we can reduce the size of the devices, since each one only has to provide half the power. Now we've reduced the output ripple and device stress while keeping the output cap the same and potentially improving the transient response. Of course, the premium we invariably pay is in board area, but perhaps that's a trade-off we're willing to make.

Now we can generalize this to more phases: an 8-phase supply has 8 switchers running 45 degrees out of phase from one another, meaning we get 8 pulses per switching cycle, and each device handles 1/8 of the total output power.

At the end of the day, it absolutely does not matter how MANY phases you use. What matters is whether you've satisfied your figures of merit, i.e., ripple, regulation, et cetera. There are a huge number of possible designs that will all satisfy the ATX power supply requirements, and certain manufacturers will make particular tradeoffs in light of other decisions they've made (e.g., they prefer to buy components from company X, and they get a better deal on the smaller devices, so they go with a 16-phase supply instead of an 8-phase supply). It's possible to build a 16-phase supply that absolutely sucks, and it's possible to build a 4-phase supply that is incredibly solid and has great performance.

Worrying about how many phases a power converter utilizes is like worrying about what brand of fuel pump your car has. Worry about the CAR's performance, don't fret about Goodwrench versus MOPAR.

 

[beray]

Originally posted by: kwantam

Okay, maybe "all else equal" this is a true statement, but in the real world any sane person designing a 2N-phase supply will be using components each rated at half of those that would be used in an N-phase supply satisfying the same specifications.

Nothing maybe about it, "All else equal, 8-phase had 1/2 the power of 16-phase". And the "sane person" didn't do 2N to save on cost. It doesn't work for saving cost, the 2N cost goes up not down "in the real world". The 2N part cost wouldn't drop by half to just break even.

Originally posted by: kwantam
When you designing a power supply, the important figures of merit include output power, output ripple, line regulation, load regulation, efficiency, cost, size, step response, device stress, et cetera. Using a multi-phase converter allows you to reduce output ripple without increasing the output capacitor (more cost, slower step response) or the switching frequency (lower efficiency, higher device stress, i.e., more cost).

With a single phase buck converter (for example), you have a pair of switching devices (transistor and diode or two transistors), an output inductor, and an output cap (you have many other components as well; I'm talking about the most fundamental components in the circuit, so we will ignore input filters, gate drivers, et cetera). For a given output current, the voltage on the output cap will droop by a given amount each cycle. We can calculate this as follows:

Qload = C * Vdroop
Iload = Qload / T = Qload * Fsw
Thus, Vdroop = Qload / C = Iload / (C * Fsw)

So we can reduce the cycle-to-cycle droop (i.e., ripple) by increasing Fsw or C. Now, increasing the output capacitor increases the cost of the design, and slows the transient response (important if you want to be voltage-agile to save power at low load but still respond quickly to step increases in demand). Increasing the switching frequency increases the switching and gating losses, resulting in lower efficiency and more power dissipated on your switching devices, driving up the heat production and requiring bigger devices, heat sinking, or the like.

What if instead of having a single switcher we had two running at the same frequency but half a cycle apart? Now while one is dumping charge to the output, the other one is grabbing more from the input, and vice-versa. Of course, now we have to use two pairs of switching devices as well as two inductors, but the output capacitor is shared. Now Vdroop is reduced because the output capacitor is charged twice each cycle of Fsw. Moreover, since we now have two switchers operating in parallel, we can reduce the size of the devices, since each one only has to provide half the power. Now we've reduced the output ripple and device stress while keeping the output cap the same and potentially improving the transient response. Of course, the premium we invariably pay is in board area, but perhaps that's a trade-off we're willing to make.

Now we can generalize this to more phases: an 8-phase supply has 8 switchers running 45 degrees out of phase from one another, meaning we get 8 pulses per switching cycle, and each device handles 1/8 of the total output power.

At the end of the day, it absolutely does not matter how MANY phases you use. What matters is whether you've satisfied your figures of merit, i.e., ripple, regulation, et cetera. There are a huge number of possible designs that will all satisfy the ATX power supply requirements, and certain manufacturers will make particular tradeoffs in light of other decisions they've made (e.g., they prefer to buy components from company X, and they get a better deal on the smaller devices, so they go with a 16-phase supply instead of an 8-phase supply). It's possible to build a 16-phase supply that absolutely sucks, and it's possible to build a 4-phase supply that is incredibly solid and has great performance.

Worrying about how many phases a power converter utilizes is like worrying about what brand of fuel pump your car has. Worry about the CAR's performance, don't fret about Goodwrench versus MOPAR.

"At the end of the day, it absolutely DOES matter how MANY phases you use." FEWER PHASES = LOWER COST, ONE PHASE = COST CUTTING CHAMPION as long as it is viable.

No it isn't "like worrying about what brand", I don't pick 4, 8, or 16-phase by faith, belief, trust, feeling, or any brainless methods to design power supplies. I do it the same way Asus's engineers did.

Asus's engineers definitely didn't pick 16-phase because it served no purpose whatsoever but marketing and a supreme idiotic cost cutting method which made them have even less profit margin.

Instead of making more lengthy elaborated excuses up as to why it not matter for the next 6 months, you can read why they did it here below in perfect English, there's no "reasonable and believable", fictions, nor lies in it...

ASUS´ revolutionary True 16-phase power design utilizes true hardware power regulation to guarantee genuine power efficacy. During heavy CPU loadings, the intelligent power design automatically switches to 16-phases; and conversely during low processing periods, it uses a responsive 4-phase system to power the CPU ? raising VRM efficiency. A power design that does not have auto phase switching will be unable to increase power efficiency; and lower VRM efficiency will still drop off the Output Current and result in wasted power and increased heat. The new True 16-phase design maintains an exceptional power efficiency of up to more than 96% ? resulting in less power drawn, lower temperatures and excellent delivery of performance in comparison to competing models. With the True 16-phase power design, users will enjoy reduced operating temperatures and extended lifespans of key components such as the CPU and motherboard.

 

[VirtualLarry]

I love it. A power engineer replies to you, and you reply to him quoting Asus's marketing copy. It doesn't exactly show that you have a good grasp of the domain.

 

[wolverineI]

Well the engineer is talking about the PSU while the other guy is talking about the onboard
motherboard power regulation.Two completely different things.They are both right about the respective parts they are talking about.

 

[kwantam]

Originally posted by: wolverineI
Well the engineer is talking about the PSU while the other guy is talking about the onboard
motherboard power regulation.

I'm talking about both, because fundamentally they are identical pieces of technology.

beray, you're welcome to believe whatever you'd like. From what you've written it's clear that your experience in the matter clearly outpaces my own meager understanding of the field. Strange, considering that the power converters I've designed are sold in quantities on the order of tens of millions of units a year.

*shrug*

 

[Old Hippie]

From what you've written it's clear that your experience in the matter clearly outpaces my own meager understanding of the field.

LMAO!! :laugh:

 

[beray]

Originally posted by: VirtualLarry
I love it. A power engineer replies to you, and you reply to him quoting Asus's marketing copy. It doesn't exactly show that you have a good grasp of the domain.

Amazing eh? Asus's marketing had no lies in it.

Originally posted by: Old Hippie

From what you've written it's clear that your experience in the matter clearly outpaces my own meager understanding of the field.

LMAO!! :laugh:

I'd been truly impressed by the experience stated. But I'd rather have the reasons of Asus's engineers doing 16-phase in place of the 8-phase they'd already done instead of impressive experience.