Hardware Asus P6T Deluxe

[unclebulgaria]

"First of all, the CPU area; not only does it have the usual over abundance of PWMs - 16 phases to be exact - it's paired off with an extra two for the Uncore area which handles the L3 cache, memory controller and QPI. For the most part we still feel Asus is playing the numbers game, where "more is better", but it neglects to point out it uses smaller ferrite core chokes, which therefore hold less power. Plenty of MOSFETs will help with accurate switching but then why not go for Digital PWMs instead?"

Sorry, this is almost total tosh. It uses smaller ferrite chokes because they are cheaper per total power load than fewer big chokes. It will call for 16 phases to handle the (frankly insane) power requirements of the new chips. It is *much* cheaper to use more phases than add fewer big phases. Also, the average discrepancies between devices will tend to average out, meaning one of fewer phases has less opportunity to throw the delivered power significantly out of whack. Think about it - if you have 3 phases at 0.5, 0.5 and 5% accuracy, and the 5% is the outlier, you need more phases at 5% out to cause a big change in regulated output, the chances of which drop exponentially the more phases you use.

And Digital PWM? I'm sorry, you really will need to explain this one! For once Wikipedia is actually fairly accurate ...

[Splynncryth]

There really is very little difference between the dual socket server boards and the single socket x58 boards in terms of the core chipset. This really isn't a board for those looking for cheap perfromance. The price is a bit steep when compaired to some othe boards, but with the extra storage controllers, and some of the other bits and pieces, the proce isn not bad when lined up against similar products.

For a more mainstream priced but still enthusiast focused product, I think you will have the wait for Lynnfield next year.

The SAS controller is not totally worthless, you can use it to add SATA drives as the contoller handles those too (thanks for mentioning it in the article). There are a few things I can think of that it would be good for. If you use sleep rather than hibernate or powering the system totally off, will cut down on boot times. But not all boards are equal when it comes to that.

Is it too late to ask for comments on sleep mode, like how well it works, and power draw in sleep (perhaps along with hibernate, and 'off' states)?

[Sublym3]

All the X58 boards are disappointing and full of crap, not to mention they cost way to much for what they are.

Asus Rampage 2 is about ~800 aussie dollars.

Gigabyte GA-EX58-DS4 goes from about 380 to 440 aussies dollars which is the cheapest X58 mobo available.

[Bindibadgi]

Originally Posted by Splynncryth

Is it too late to ask for comments on sleep mode, like how well it works, and power draw in sleep (perhaps along with hibernate, and 'off' states)?

It's now packed away so I can't tell you power, but it goes in and out of AI Nap just fine.

[Bindibadgi]

Originally Posted by unclebulgaria

Sorry, this is almost total tosh. It uses smaller ferrite chokes because they are cheaper per total power load than fewer big chokes.

We're equating the same point. Smaller holds less power, but the total power delivery will be equivalent to that of other brands. Asus is playing a numbers game - 16 is better than 12, yet our 16 are smaller than others 12.

It will call for 16 phases to handle the (frankly insane) power requirements of the new chips.

This is BS. For starters - the Uncore area has 2 phases to itself - that's QPI, memory controller and L3 cache sorted. The Cores, four of which, are actually less transistors than Core 2 because of the smaller L2 cache. In addition, it has a per-core power control which gifts a great finesse in power control and delivery for the four cores. Core i7 is 130W TDP, but that's northbridge + CPU, unlike Core 2 which was without the Uncore area. If anything, those 16 phases have to handle LESS power than they did in Core 2, especially Kentsfield CPUs, but what they do offer is greater available power delivery on demand. You say it will call for 16 phases - then how can Gigabyte do it in 6 (a virtual 12 since there are only 6 driver ICs) and MSI does it in "6" too? The ECS I have here has 6 as well for *everything*.

It is *much* cheaper to use more phases than add fewer big phases.

Wrong, I'm afraid. while big chokes cost more than little ones - chokes to NOT equal power phases. A power phase consists of a driver IC, an up and down MOSFET (basically) and an inductor coil (choke) - some companies vary the amount of hardware, but extra total phases cost respectively more because they require the additional extra hardware, whereas the cost difference between large and small chokes will be cents at most when these guys are buying millions at a time. Lets also not forget economies of scale - Asus is a manufacturing BEHEMOTH, so it can buy its hardware cheap. No one pays the same price, it's all negotiated - the TWese, like the Chinese, are notorious for being cheap ***** (in the nicest sense of the word) and everything is cost down if possible to make the extra buck. That is, UNLESS it has a positive marketable effect that can drive further sales.

Also, the average discrepancies between devices will tend to average out, meaning one of fewer phases has less opportunity to throw the delivered power significantly out of whack. Think about it - if you have 3 phases at 0.5, 0.5 and 5% accuracy, and the 5% is the outlier, you need more phases at 5% out to cause a big change in regulated output, the chances of which drop exponentially the more phases you use.

This I agree with entirely: should 1 phase fail, then there are more to cover it. However, on the other side, more components mean more chance of failure. We've had one X58 go pop already because a singular MOSFET decided it wanted to burn up in blue smoke, despite it being a "multi-phase design".

And Digital PWM? I'm sorry, you really will need to explain this one! For once Wikipedia is actually fairly accurate ...

Digital PWM control it notorious for allowing a greater degree of voltage accuracy and greater power per phase, because the power switching is more tightly controlled. However, it takes a strong degree of design and knowledge to get it right: most companies don't employ people who know how to do it or have enough time to research it (which costs money = less profit = basic cost:benefit ratio).

In the early versions there was a large vDrop (or Droop I cant remember) and they were seriously hot running, however the latest versions are much, much better but the accuracy still needs to be appropriately calibrated. It's also expensive which is why DFI and EVGA have gone for it (enthusiast: DFI knows already, EVGA just employed a load from Foxconn, Abit, EpoX) but given the super tight time constraints of product development no one else has.

What I was wondering was why Asus don't consider it, considering it's such a diverse comapny employing many talented people, including some ex-Abit staff that must have some knowledge of digital PWMs... perhaps. The cost:benefit should be better for Asus that designs a lot of different computer hardware.

[Kúsař]

So what's the biggest difference between digital PWM and "analogue" PWM that makes digital PWM much better? "...power switching is more tightly controlled...". I thought that "analogue" PWM responds to changes in CPU voltage immediately as it's using part of CPU voltage as feedback comparing it to set reference voltage(and adjusting mentioned MOSFETs accordingly).

[unclebulgaria]

Mr badgi,

PWM is inherently digital. The switching device cannot control the amplitude of the output directly, so ferrites / inductors / chokes are used with capacitors to lower the deviation of the output signal and limit the rate of change. You can factor their effect into the design pretty easily using Laplace transforms and considering the effect of switching frequency (put everything in the s-domain and it all falls through pretty easily). PWM stands for Pulse Width Modulation - increase the on-to-off ratio (also known as duty cycle) and the output will veer towards a value set by the supply voltage and switching losses in the circuit, which in turn are frequency dependent. Therefore, by controlling the duty cycle you control the nominal amplitude of the output voltage. So PWM is inherently digital - as it cannot be done without a digital component to the control loop - and I repeat: what does Digital PWM mean?

In the grand scheme of worldwide electronics manufacture, I'm afraid Asus are a drop in the ocean compared to the likes of BCM in Malaysia or Nokia. Even if it were on these scales, ferrites are the most economic way to add induction to the output of an SMPSU. Also look at the price against rated switch voltage and quiescent current for an IC and you will see again: more expensive. Then look at Rds(on) characteristics of MOSFETs against price. Guess what? You can research this at FEC or RS. I'm sorry, more phases = cheaper.

[trmbne2000]

Will you be reviewing EVGA's 132-BL-E758-A1? Its selling for the same price as the P6T Deluxe on Newegg, and it comes with 3-way SLI. I'm definitely thinking about one of those as I love EVGA products.

[Splynncryth]

Originally Posted by unclebulgaria

Mr badgi,

PWM is inherently digital. The switching device cannot control the amplitude of the output directly, so ferrites / inductors / chokes are used with capacitors to lower the deviation of the output signal and limit the rate of change. You can factor their effect into the design pretty easily using Laplace transforms and considering the effect of switching frequency (put everything in the s-domain and it all falls through pretty easily). PWM stands for Pulse Width Modulation - increase the on-to-off ratio (also known as duty cycle) and the output will veer towards a value set by the supply voltage and switching losses in the circuit, which in turn are frequency dependent. Therefore, by controlling the duty cycle you control the nominal amplitude of the output voltage. So PWM is inherently digital - as it cannot be done without a digital component to the control loop - and I repeat: what does Digital PWM mean?

In the grand scheme of worldwide electronics manufacture, I'm afraid Asus are a drop in the ocean compared to the likes of BCM in Malaysia or Nokia. Even if it were on these scales, ferrites are the most economic way to add induction to the output of an SMPSU. Also look at the price against rated switch voltage and quiescent current for an IC and you will see again: more expensive. Then look at Rds(on) characteristics of MOSFETs against price. Guess what? You can research this at FEC or RS. I'm sorry, more phases = cheaper.

The PWM is what runs the DC-DC converter, it should be a buck converter in this case. The very basic principle of the system is to 'squirt' a bit of power into some output caps as needed to keep the voltage at a constant level. IIRC, the inductors provide the current, but I was never all that good at the theory
One problem with all this is that there is still high frequency noise, that of the base frequency that gets modulated. More phases is to try and do a sort of noise canceling and provide cleaner power as I understand it. More phases means less power between the phases, and that should mean less stress to the components because there is a LOT of current at those voltages and wattages.

I belive the differences between analog and digital are in what is controlling the duty cycle. In an analog system, an analog timer generate the duty cycle. In a digital system, it would be a microcontroller or logic circuit.

I don't know which set up is better. With analog, I would think component value drift would be a problem. With digital, I would guess it would be the granularity based on the microcontroller's clock, or sample rate of the ADC it would probably need.

[Bayaz]

I have a P6T Deluxe V2 and find that the BIOS fan control doesn't work and a lot of people have complained on the Asus forums about problems with waking from sleep mode.