Re: GIGABYTE Z77X-UP5 TH Review with Ultra Durable 5 Analysis
The IR3550 can do 60A at 10W power dissipation, and that is very impressive considering typical LF-PAK MOSFET the 7030AL which is the high-side MOSFET(limit for power stage current output) on many high-end boards can output 68A at 62.5W power dissipation at 25C, those conditions will never happen. That would require you to cool each individual 7030AL high-side MOSFET with as much cooling as you use for your CPU!!! If you wanted to you could use six IR3550 at 60A and have the same power dissipation (which is possible because of expanded area of six IR3550) but also 6x the current output when compared to the 7030AL! However we can use the graphs in the datasheet provided with the 7930AL to calculate how much current at 10W it can output, and answer is about 27A in the same conditions as the IR3550.
So in the same conditions the IR3550 will output twice as much current, thus that is why it is so damn attractive. The IR3550 also is at 90% efficiency at 50A, compared to a typical 50A DrMOS(Vishay SiC780) which is cutting edge for DrMOS can do 35A at 90%. Thus the 50A DrMOS isn’t even comparable! The only MOSFETs that come close that have ever been used on a motherboard are the DirectFETs also from IR, and that is no surprise as the IR3550 uses technology from the DirectFETs, such as full body copper design, as well as gold internal connectors, and a lot of output surface area for current. The full body copper clip means that the current flows through the low-side MOSFET’s body instead of the PCB, which means less resistance for the output current.
The IR3550 are one of the best solutions, and that is why we are going to see them on many future high-end boards. For instance the EVGA Z77 FTW uses the IR3550, 7 of them; however IR has no marketing for these new power stages other than some press releases, thus GIGABYTE made its own marketing which I will show below. The parts of the marketing below are somewhat educational and correct as well:
First up from GIGABYTE’s marketing, what is an AIO Powerstage?
(From GIGABYTE Marketing:)

(A standard powerstage however just requires 1-high side MOSFET and a diode, however you will see the diode replaced by a low-side MOSFET in modern systems, and in good systems like on the current Z77 GIGABYTE and ASUS boards you will see 2 low-side MOSFETs and 1 high-side.
Next up, what is inside the IR3550 and why is it special?
(From GIGABYTE marketing:)

Now I have taken some of their marketing, and added some tidbits which show first the comparision to the DirectFET technologies:
(Images from GIGABYTE Marketing, Text from Me)

Connecting all of the low-side MOSFET to the high-side MOSFET is something IR said only the IR3550 series does.
Next up we have a comparison I made between the IR3550 and the DrMOS I just talked about:
(Image on the left from GIGABYTE Marketing, rest from ME)

In the above image, none of those is the 50A DrMOS, it is just the 60A IR power stage in two different views to show you the inside. There are pads underneath the power stage which you cannot see in the image, they are under the MOSFETs.
I just want to say this, the IR3550 is NOT a DrMOS, DrMOS is a specification made by Intel for an AIO power stage, a DrMOS is a type of power stage, however not all power stages are DrMOS. Texas Instruments, Volterra, and IR have power stages which are AIO but are not DrMOS. DrMOS has its own protocols and designs, however these power stages are an improvement on that design, they also have different pins and different internal circuits, the IR IR3550 is made to work with an IR PWM (CHiL PWMs included).
Now that we have covered the MOSFET, we have to cover the inductor, which I personally believe hasn’t been talked about enough, as it has a HUGE part in making 60A per phase possible.

They are also rated 60A and they had to be custom made for GIGABYTE. The correct DCR as well as inductance and current rating where combined to bring about an extremely strong choke.
The IR3563 is also a first for GIGABYTE, it is the first time they have used a true 8-Phase PWM to power their 8-Phase VRM. Now you are wondering, what is so great about 8 phases and don’t many boards have true 8-Phases? The answer is no, and let me explain.
First of all you need multiple outputs from a PWM to drive different voltage outputs, the reason we have multiple output PWMs is to save money!!!!!:

That is one reason the UP5 has more digital PWMs than the UD5H, because the CPU VCore PWM is a 8+0 phase PWM, and all 8 are being used. Many boards like the Z77 Extreme4 use 4 phases and double them to 8. The M5G uses the same doubling.
First of all there is a reason when you look up a PWM it says 6+2(IR3567 found on GIGABYTE boards) or 7+1(CHL3828 found on ASUS boards), 6+1(ISL6367 found on ASRock boards). That is because all phases need the same output, so you cannot per say take the IR3567 6+2 phase PWM and run 4 phases on the VCore and 4 on the iGPU, you can however do 6 on VCore and 2 on iGPU.
The PWM used on the Z77X-UP5 TH is a true 8 phase PWM:

The International Rectifier IR3563A is a true Digital PWM, incorporating cutting edge digital PWM technology. This PWM is possibly the best VRD12.5 PWM you can find ATM, it incorporated VRD12 and VRD12.5 protocol design, and many transient algorithms to correctly and efficiently supply power to the CPU. This little IC is responsible for the high overclocks and tight LLC that we see on the UP5.
Benefit of more True phases: But why does that matter you are probably asking, as more components do help lower the temperature and the amount of work that needs to be done even when you double phases virtually. The answer is that the increase in true phases has performance benefits. First of all each phase will only switch 1/8 of a cycle after the one before it instead of ¼(if compared to 4 phases doubled to 8), meaning that per cycle each phase is doing less work when there are more true phases. Also there is an interleaving affect which is caused by this per cycle addition of phases. It reduces total ripple amplitude by increasing the ripple frequency. Having 8 true phases will result in a total ripple frequency 8x greater than that of any single phase, however it will only increase the ripple frequency 4x greater in the virtual 8 phase(true 4 phase) system.
That means if you had two of the exact same VRM, one with its phases doubled and one with its phases not doubled, then the one which wasn’t doubled will have much higher ripple frequency and thus much lower ripple current, which is what we always want to reduce. Thus the output ripple and the noise from the VRM will be greatly reduced when you use the true 8 phase VRM, such as the one on the Z77X-UP5 TH. Of course this affect will reduce ripple by a much larger amount when increasing from 4 to 6 phases than it will from 6 to 8 phases, however nevertheless the affect is there and has an impact.

So let’s do a quick overview on why this VRM is one of if not the best:
- IR3563A provides true 8 phases for a true 8 phase VRM
- 8x IR3550 are able to output 60A a pop in real-life scenarios, and thus can output a total of 480A
- 8x 60A Custom made Ferrite chokes are able to sustain 60A current flow and not limit the MOSFETs.
- 2 OZ copper PCB is required to be able to transfer this type of output, if it doesn’t say 2oz then most likely it isn’t.
- 8x 820uF OSCON Sanyo brand polymer capacitors provide very low ESR to reduce current ripple, and they are high enough in capacitance rating to store power which isn’t used.
Memory VRM:

The memory and the iGPU get their phases from this IR3570, which is a 3+2 phase digital PWM, only 2+2 phases are used. Two phases for the memory are provided to a dual driver, the IR3598, which outputs to two sets of Rensas K03B7 for the high-side and one K0393 for the low-side. Each phase is capable of 30A on a good day.

A second IR3570 provides PWM support for the VTT/IMC phases are pictured below. Also support for the PCH power is also provided by the MOSFETs shown above.

Above we have the same 2 sets of Rensas MOSFETs which we saw for the memory VR; however this time we have 60A chokes to go along with them. Here we see an IR3598 which is used as a dual output driver instead of a doubler.
The PCB:

On the back of the PCB we see two sets of backside metal brackets, which usually are used to cool down back-side MOSFETs, however we have no back-side MOSFETs, and these brackets are used to stiffen the board and support the heatsinks. Thus the motherboard can support very heavy heatsinks without issue.

PCB is nice and black, right? That is true black, like a dark knight.
Circuitry Analysis:
ThunderBolt:

Here we have thunderbolt, with all the support ICs, one would think this could all go on a separate card. It is even complete with its one VRM, which is in the image below.
- DSL3510L: Intel dual output Thunderbolt controller which allows daisy chain. It has a 3.4w TDP and supports 10GBps per channel.
- 2x L04DP211 :(one is on the back) made by NXP are display port 2:1 MUX and provide DP to each thunderbolt port.
- 2x TPS22980: made by T.I. are 3.3v to 18V high voltage MUX, used to power the ICs inside the expensive thunderbolt cables. 18v is max output which is needed as DC voltage doesn’t hold over long distances.
- PI3HDMI412: made by Pericom is a 4 channel 2:1 MUX for providing HDMI bandwidth to the thunderbolt controller
- PI3C3125: made by Pericom is a high bandwidth 4-bit 2 port bus switch which is used to provide the controller with bandwidth.

Here we have a single phase VRM to power the Thunderbolt controller and ports!!! It is powered by a Richtek RT8120 single phase PWM with integrated drivers. Some good quality PowerPAk MOSFETs are used in lieu of D-PAK.
Audio, NIC, Connectivity ICs:

On the Z77X-UP5 TH GIGABYTE has provided buyers with a beefed up audio system. Basically the same as on the Z77X-UD5H, the UP5 uses a 110 dB SNR Realtek ALC898. However while most high-end motherboards use this codec, a few license Creative XF-I sudio software and rebrand it as XF-I, which is what GIGABYTE does here. However GIGABTYE doesn’t stop there, they add in two audio amplifiers one for the backpanel TOSLINK, and the other powers the internal header for a front panel device. GIGABTYE uses two T.I. DRV632 AMPs/Line Drivers.

Many users like Intel NICs, and thus the Intel WG82579V is a PHY to complinet the controller inside the PCH which is popular on high-end motherboards.

To exapnd USB 3.0 ports as well as amplify and rejuvinate the USB 3.0 signal over a long distance two VLI810 1:4 USb 3.0 hubs are used to provide a total of 4 USB 3.0 ports on the back and 3 internal headers capable fo 6 ports together.
930@4.4 1.4v
GA-X58A-UD5 rev. 2.0
XMS3 2gb x 2 1600
GT 220
Custom Water
Custom Lian-Li PC 70
Bookmarks