Intel's P45 Shows Promise, As Does ASUS' P5Q Deluxe

[Rob Williams]

From our front-page news:

It almost feels like Intel just launched their P35 chipset, but it's been close to a year. Since that release, we saw both the X38 and X48 chipsets come to fruition, so the next step is of course the P45 (and G45). We are in the process of working on a review for the ASUS P5Q Deluxe, but I thought I'd post about it quick here, since it's quite an offering.

P45 in itself is not too exciting, specs-wise. It bumps up the official FSB from 1333MHz to 1600MHz and includes support for PCI-E 2.0, a la X38. Total yawn, but what's exciting is the fact that the overclocking ability is far improved, and the options available in the BIOS are on par, and beyond, with X38/X48 options. When I first took a peek into the P5Q's BIOS, I was surprised to see brand-new options available, such as new memory enhancers and voltage options.

I won't delve much into that here since I haven't toyed with it, but one thing's for sure... P45 will be for dedicated overclockers. But that aside, the P5Q is one packed board, and putting it so simply makes it an understatement.

​

In addition to their EPU technology, ASUS boasts having the first 16-phase power solution available, which is true. Gigabyte holds a close second with their 12-phase power solutions, although others might exist as well. Another new feature is "Drive Xpert", which is essentially RAID but in layman's terms. With two special S-ATA ports on the motherboard, you can either duplicate an entire drive (even without entering Windows), or set up "Super Speed", which is essentially RAID 1.

On top of it all, the Splashtop embedded Linux environment makes a comeback, but is a much more updated version than what we first saw on the P5E3 Deluxe. In addition to being able to use Skype and surf the web (where YouTube now works), you can look at photos on the PC and talk via IM with a customized Pidgin application. I'll tackle this and a lot more in our review which will be posted within the week. Stay tuned...
​

 

[Kougar]

There have been quite a number of very interesting P45 boards set to launch, Gigabyte with their usual overboard excesses, ASUS with their now TRIPLE phase RAM, triple phase northbridge and 16 choke/phase CPU power system that is in reality not even close to 12 (Just like Gigabyte's 12phase set up is more akin to an actual 6phase design). And then there is Foxconn... they moved the NB between the socket and RAM for the shortest trace lengths yet. They also included silksreened measuring points for a multimeter user to measure vdimm and MCH voltage power.

It kind of sucks none of these will support Nehalem... it figures the best features show up on boards that'll only be useful for <6 months.

 

[Rob Williams]

How is it that the 16-phase power solution is not close to 12? I'll admit I'm not that keen on power solutions, so any insight would be great.

I still love the P5Q though. It's a feature-packed board, but the overclocking -sucks-. I am unsure if it's the P45 chipset itself, or the board, but I can't even get above 465MHz on the E8400, and trust me, I spent a good amount of time overclocking. Whereas 500MHz was easy on the P5E3 Premium, getting 465MHz was more effort here.

 

[Rob Williams]

Another thing I should add is that I've found the P5Q Deluxe to use less power overall, when compared to the P5E3 Premium, which is a little odd since I thought the more phase units on board, the more power it would suck. Is the P45 optimized that much more than X38, or is the EPU really kicking some serious ass?

Here's the results I've seen:

P5Q Deluxe
Idle: 168
Load: 273

P5E3 Premium:
Idle: 183
Load: 306

Rampage Formula
Idle: 181
Load: 306

Gigabyte X48T-DQ6
Idle: 182
Load: 310

Intel X48BT2
Idle: 186
Load: 306

The P5Q's power efficiency is undeniable. But is it P45? EPU? The 16-Phase power? I'm having a hard time figuring this one out, heh.

 

[THUMPer]

isnt P45 45nm chipset?

 

[Rob Williams]

No, as far as I know it's still based on 65nm... it just has a few features removed. Intel is keeping mum on things though, so I'll have to go by what's leaked.

They'll be going into more detail at a conference at Computex in a few weeks.

 

[Kougar]

P35/X38/X48 are all 90nm. P45 and G35 are 65nm.

Gigabyte tends to have board power efficiency down pat in comparison to ASUS. Usually I see Gigabyte's boards winning by ~5 watts, including the 12 choke models. If you compare DES verses EPU, Gigabyte widens the gap considerably. Such as http://www.tweaktown.com/articles/1348/4/page_4_test_system_setup_and_power_usage/index.html Their results perfectly mirror my own testing with an EP35-DS4, DES took a huge chunk out of the power consumption, partially by offering a dynamic undervolting the CPU.

To quote Anandtech's preview of the P5Q3. I honestly do not fully understand this myself, but previous Anandtech comments regarding the "true" circuitry designs of Gigabyte boards has proven true in the past so I believe it.

The P5Q3 may report the use of a 16-phase PWM but we know better. Although ASUS design engineers have added a lot of extra chokes and MOSFETS, the overall capacity of the power delivery circuit remains comparable to their competitors' more modest 8-phase designs. One thing's for certain, we can all agree that more phases do not necessarily make for a more stable power subsystem. Although we did not experience any problems with this particular solution we would rather see ASUS concentrate their efforts on designing an entirely new circuit based on a true 6-phase PWM.

Edit: I may as well add it to my post. While P45 is the 65nm version of P35, the actual die size has not decreased. This is because all P45 silicon is actually... G45 silicon. Intel apparently found it more cost effective to make a single design, and fuse off the GPU core. This has been the case with all P45 boards I've seen to date. Two sites have photo proof of this.

 

[Rob Williams]

I'm still confused as to how they know it's not a true 16-phase power solution. I wish I was more keen on the electrical side of things, then I could decide for myself.

If it's not a real 16-phase, then what the heck are all those MOSFETs and chokes doing?

I'll be receiving one of Gigabyte's P45 boards early next week, so I'll see how it compares to this ASUS one. From what I can see right now though, P45 offers a significant drop in overall power draw.

 

[Kougar]

What they are doing is adding some flashy marketing FUB. Gigabyte is bad enough about doing that so I can't single out ASUS there.

Gigabyte confessed their DQ6 uses a "Quad triple phase" design, not a fully 12 mosfet/choke design. Much of it is just extra redundancy more than actual extra capacity to deliver more power, they added two chokes per single mosfet. I am no electrician, but I suspect this is akin to doubling a four wheel car to eight wheels... better traction but it's not really needed. I may be wrong here though, something I need to do a good bit of reading on, and Anandtech didn't care to elaborate any on the "why" either...

Power savings would be due to the die shrink... die may be the same size, but they did fuse off the GPU on the P45 silicon so much of the silicon that is now there is simply inactive.

 

[Rob Williams]

Thanks for the info... I'd still love to know the reason it's "fluff", but I have no way to prove otherwise, so whatever. It's foolish regardless. First the megabyte race, megapixel race, gigahertz race and now the power phase race?

 

[Kougar]

ASUS is only doing it because Gigabyte was gaining market traction with that.

On the other hand while it is definitely silly, I sort of like this race. In a single chipset generation we went from ASUS & Gigabyte using only single phase power designs to them using triple phase designs to power the Northbridge & Memory.

GT 280 is has a 2+1 or a 3 phase power system on the NVIDIA reference model. I guess Gigabyte will be using quad-phase designs for their future GPUs... should be good.