ASRock Z68 Extreme7 Gen3 Review
A Very Close Look at the Extremities of the Extreme7
In this segment we will take a upclose look at two of the most important physical aspects of this motherboard; power delivery and the PCI-E 3.0+NF200 implementation. We shall begin this section with a bare shot of the board, without the giant heatsinks.
With the heatsinks removed it is easy to see a straight 12 phase voltage regulator that provides power to the CPU. It's divided into an 8+2+2 phase design, using three different PWMs. First lets cover the question we asked in to the intro, how does this board achieve PCI-E 3.0 with an NF200. First let me show you a nice shot of the PCI-E system of the motheboard:
Let’s start with the parts, and maybe you will figure it out on your own. First in this picture above the NF200 we have eight 2-lane bi-directional PCI-E 3.0 NXP L04083B PCI-E switches, which can switch a total of 16 lanes in two different directions. The green chip underneath is the NF200. Then we see the PLX Bridge chip PLX PEX8608, but the PLX chip really has little to do with the NF200, instead ASRock says in their advertising that it helps better manage the PCI-E 2.0 lanes of the PCH. Since there are so many third party controllers (USB 3.0, SATA6G, NICs, PCI, 1394a) that use the PCI-E 2.0 lanes of the PCH, the PLX bridge can better manage the lane allotment. One 16x slot is a 4x slot and that 4x is directly connected to the PCH. Now there are 4 other PCI-E 2.0 switches, the ASMedia ASMT1440 which switch the output from the NF200 for SLI.
Now here is how it gets PCI-E 3.0 with the NF200: those 8 switches are able to switch 16x lanes from the CPU to the second PCI-E 16x slot, or to the NF200, totally bypassing the NF200 if only one card is in slot 2. As long as those switches are PCI-E 3.0 capable then you can have PCI-E 3.0. Of course if you run any type of SLI, or put any card in any of the other 16x slots (except the 4x slot) the second PCI-E 16x slot with automatically disable.
Here is ASRock's "caution" directly from their manual:
Now we made a chart to show you the PCI-E/NF200 configurations of different LGA1155 motherboards.
To begin, the NF200 isn't the strongest chip in the clan, yes it provides 32x lanes whether it has an 8x or 16x input, but it does cause additional latency. The Z68X-UD7 has an all lane attachment to the NF200, meaning none of its lanes bypass it, GIGABYTE wants users who don't need 3-way SLI to just buy the Z68X-UD5 as it’s basically the same board without an NF200 at a better price point. The ASUS Maximus IV Extreme-Z probably has the best configuration; you can have a single slot at 16x without NF200, as well as 8x/8x SLI/CF without NF200, which is the best way to do it on LGA1155 platform. On top of that since the Maximus IV Extreme-Z has 8 lanes directly connected to the first slot, and the NF200 still outputs 32 lanes with the other 8 lanes, it can run 3-Way SLI/CF at 16x/16x/8x. EVGA's Z68 FTW is almost the same, except it can’t run 2-Way SLI/CF at 8x/8x native, but it can run 3-way at 16x/16x/8x. The ASRock can run one slot at 16x native with PCI-E 3.0, or it can run SLI/CF like the Z68X-UD7. That pretty much sums up LGA1155 NF200 configurations in all their simplicity and complexity.
Now that we have covered that we can move on to more interesting stuff, the voltage regulators. The question lies, does the voltage regulator on this board stand up to other top of the line boards? Well let's take a look!
We can see that the board uses some very nice inductors, looks like the same type as on the GIGABYTE X58A-OC, GIGABYTE called them Max Power Ferrite Choke, and they were capable of up to 50A per inductor. These are probably similar, but they look a bit different and have a different rating. Now you only see so many output capacitors, but don't be fooled these caps have a very high capacity, 820uF a pop. Now we move on to the MOSFETs, the Low RDS(ON) MOSFETs in this case are the same brand that ASUS uses on their mainstream boards, NXP, but these are NXP PH9030AL for both high-side and low-side, they can provide about 40-50A per phase in good temperature conditions, but their specs show a great decline in output with increased in temperature. Of course their output at high temperatures is still very great.
There are two odd things about the VRM on this board, first of all the VRM uses three PWM control ICs and second ASRock uses ST driver instead of the CHiL ones.
The CHiL 8328 is a powerful 8-phase VRM, in this case all 8 phases are used and not doubled, they go through ST6743 drivers, 8 of them. Apart from the ASUS boards ASRock is the only company from which I have seen the use of this specific PWM. The funny thing is that ASRock decided to use a 3+1 phase Richtek RT8859M and only use 2 of its phases for the iGPU, and uses two RT9618 drivers instead of the ST drivers. A separate unknown Richtek controller is used to power the VCCIO/SA two phases, the unknown controller has integrated drivers. The two phases for the VCCIO/SA use different inductors.
The memory VRM is shown below, it has two phases.
Each phase uses the same high current MOSFETs as well as the unknown PWM from Richtek which has integrated drivers.
On the back of the board you can see the drivers in a straight row, as well as 3 peripheral ICs. Guess there just wasn't enough room on top of the board.
Here is something we thought should be pointed out. The PCI-E Slot locks are very cheap, and they aren't automatically engaged when a card is in the slot. Users have to manually lock the card in, and unlocking is not so easy.