X58A-OC Review- Physical Review
(Layout, Packaging, VRM analysis, ICs Analysis, Initial Air Cooling Results)
(After staring at this board for a few hours, as I do every board, I try to analyze what type of design this board is trying to model. I have a feeling this board is supposed to have heatsinks that resemble a car, and a fast one at that, but I felt as though this ICH (Southbridge) heatsink looks like a flag waving in the air.)
Today I have my first close encounter with this masterpiece of a motherboard. Like no other motherboard its sole purpose is to fulfill the needs of overclockers, and overclockers only. Yes it is true, every motherboard is overclockable, yet not all motherboards are overclocking friendly. This isn’t the first motherboard to be tuned to overclock, and it’s not the first to hold world records, but it is the first motherboard to be targeted solely at overclockers, and extreme overclockers at that. Not everyone may agree with that statement, but it is true. While many motherboards feature overclocking as their main feature, they still are marketed towards gamers as well. As we saw earlier the G1 series is targeted towards gamers, and hardcore gamers at that, this board is on the exact opposite side of the spectrum in terms of features. While the G1 Assassin might have a creative 20K2 audio processor and a Bigfoot NIC, the X58A-OC has a revamped voltage regulator module and direct hardware controlled multiplier, BLCK, and even switching frequency adjustments. This board is similar to a few other boards in terms of its features, but this board is without any features that overclockers do not need. With this board you don’t pay a premium for ICs(chips) you will never use (extra USB, dual NIC, eSATA, IEEE, or even 6 channel audio), instead GIGABYTE took the money that was saved and invested it in custom order parts for the voltage regulator module, and some nifty overclocking features.
By far the most interesting part of this board is GIGABYTE’s take at miniaturizing their 24phase voltage regulator module (VRM), following that are the OC buttons, and then small things (4GHZ easy button, PWM switches, BIOS switch, 4 way CF/3 way SLI, triple slot spacing, etc….). Today I will begin by going over the parts of the board, and then I will explain how the new VRM compares to the old one. Following the explanation of how everything works with the VRM, I will go on to explain every chip, one by one. For you to get a real sense of what this board has to offer you will need to see what makes it tick, and then how it performs. Of course that is a huge venture so just as usual I split my reviews into two parts, with this one being all about the physical stuff.
Table of Contents:
Layout and Packaging
Voltage Regulator Module Analysis
Important Overclocking ICs
All other ICs (Including Marvell SE9182 SATA6G and EtronTech USB 3.0)
Heatsink Analysis
Board Installation and initial OC results
Conclusion
Please press the power button to begin
Packaging and Accessories:
This board is standard ATX size. It will fit any case that can do normal ATX. There is no problem with it fitting into a case, even with triple slot PCI-E spacing. The box is fitting as well, plain and simple; it’s all in the name.
Overclockers don’t necessarily want any fancy boxes, or any fancy accessories like stickers, we just want a solid board that performs well.
I took the board out of the box to show you that we aren’t paying for anything we don’t need (they could have lost the CD, and probably a pack of the SATA cables, and even the back panel).
All the accessories included. We have a 3-way SLI bridge as well as a CrossFireX and a SLI bridge. The SATA cables are all black, and for those of you who think SATA3gb/s cables are different than SATA6gb/s cables, they say SATA6GB/S on them.
There are two cool parts of the accessories, the first are the voltage read cables. There are 7 of them, which means you can read every voltage available with a read point at the same time, you just need 7 multimeters. As you can see they are made to fit a wide variety of probes, this is one style that I personally like. I have used the ones that are like molex connectors, but this type will hold the tip of your probe in place, so you don’t have to worry about it falling out. I like this type of detail to attention; you don’t always see this with every brand.
Here we have the backpanel; it has slits in it for air flow I am guessing. I was told a very early version of this board has a built in fan in the heatsink. Hicookie said that it just wasn’t cutting it. Fans that are very small lack the ability to push more than a few CFM, and if they do they are loud. Most overclockers use their own large fans, so there was really no need for it I am told. You can see how cool this board is; look how it only will fill up 2/5 of your back panel!
Finally the masterpiece everyone has been waiting for, the X58A-OC, complete with its nice anti-static bag to ensure its safety into your hands. Next let’s move on to the layout of the board, to get a feel for it.
Layout and Design:
In this section I will go over everything from the buttons (including function), to connectors, and even to PCI-E spacing.
As you can see it is very gorgeous, very nice orange and black theme. It reminds me of Halloween, so maybe this board is something to be feared. I was told that they had many color schemes for this board. They had it made in many different colors, and then everyone at GB HQ at the time had a vote, and this one was picked. I think it looks great, might even match a few case setups nicely.
Here is a quick look at what we will be covering in this section:
This board has more fan headers than anyone should need in a case. It also has more buttons than you can hit at one time even if you use both hands. We will go over them one by one; I promise you won’t miss a thing.
The socket area is well done, very clean compared to many boards. This was done so that it is easier to insulate. The lack of can-type capacitors on the entire board makes insulation much easier. As you can see the main heatsink isn’t attached to the ICH (Southbridge) heatsink. This heatsink alone is quite heavy. We will take a deep look at the heatsinks in the cooling section. We will also take a look at how well some aftermarket heatsinks fit on this board, and how their fans work with these large heatsinks.
Moving to the RAM area, we can see that it is pretty close to the CPU socket. This helps cut down on parasitics, as the trace length is shorter from the CPU’s memory controller to the RAM slots. Also notice that the RAM slots don’t extend below the board’s heatsink. This is so you can switch out RAM without having to remove your extra long GPU. The buttons for OC are situated facing the right side of the board. This is a great position for benchmarking. GIGABYTE expects that most users won’t use a case; even if they do the buttons are still workable.
Here we have the lower right hemisphere. The SATA ports are all angled, and we even have a nice SATA6G port hooked up to the Marvell SE9182, Marvell’s new SATA6G controller. The ICH heatsink looks like a flag, I swear. Many overclockers won’t use the IOH/VRM heatsink, but this way they can still use the ICH heatsink, because it does get hot. You can also notice the PEG-OC (PCI-E power plugs) in our case they are actually SATA power connectors. I am told that you can hook up two different PSU’s to these plugs, one for each PSU. In the component test section you will see how well it works out. The position the SATA power connectors are in makes them perfect, in a case or outside, we can see how on a PSU’s daisy chain of SATA power connectors, the SATA drive can be hooked up right next to the SATA ports.
Here we have our trusty PCI-E layout. The 1st and 3rd PCI-E slots are full 16x slots, the 2nd and 4th are 8x slots. The way this is done means that you can fit two triple slotted GPUs and have them each run at 16x. The lack of NF200 makes 4-way SLI impossible, but keeps costs down. It is obvious that this board is made for 2D and single or dual card benching. If this board had an NF200 chipset, it would cost more, but also if you only use one or two cards you have added latency that otherwise wouldn’t be there. NF200 is really meant for 3 and 4 way SLI setups. We can see that GIGABYTE kept our trusty PCI slot, for anything from an iRAM to an extra diagnostic card.
Here we have our cute backpanel. Honestly more connectivity than needed, but the least I have ever seen. GIGABYTE is now using red USB 2.0 ports. The audio has been cut down to 3 ports, and the blue are USB 3.0 ports. I love the fact that they kept the PS2 keyboard and mouse ports.
Lets start again from socket area around, but this time with closeups of noticable features.
We have the trusty X58 Lottes socket. Some earlier X58 Foxconn sockets showed burn. It seems they have fixed it for P67, but for X58 Lottes is preferred. We also have out clear CMOS button, as well as our 3 DIP switches for VRM switching frequency. The stock value for switching frequency is 400 kHz, switch 1 = 600 kHz, switch 2 = 800 kHz, and switch 3 = 1 MHz. Each frequency is good for a different condition, or a different CPU. You have to play around and see which is best for you, and I will explain what switching frequency is and how it has an effect on overall VRM performance.
Here we have our trusty OC workspace. We have everything from voltage monitoring points, to multiplier and BLCK adjustment. You press the “Gear” button to change between 1Mhz and 0.3Mhz BLCK increments. We also have our trust LED POST display. I really like the LED POST display as it’s an easy troubleshooting tool. The Power and Reset buttons are together as well. The power, 4G, Gear, BLCK +/-, and Multiplier +/- buttons all are lit with orange LEDs inside. The 4G, power, and Gear buttons will stay lit, but the incremental buttons ONLY stay lit when being pressed. You will see how it works in the SetUP section.
Here we have a close-up for the voltage measurement module (that is what GIGABYTE calls it). You can either use their provided connectors, or free measure the pads if you only have one DMM (Digital Multi Meter). Each read point is also labeled.
Here we have the SATA ports, the SATA6G port is grey colored. The PCI-E power connectors are really a great idea. This is a first I have seen on any motherboard, and I hope it catches on because it really is ingenious.
As I said earlier somewhere, that I love DIP switches. Here we have one long overdue for the dual BIOS feature on the GIGABYTE boards. Rather than using fancy key combinations we can finally just flick a switch. This will no doubt cut down on board RMAs as well as increase user satisfaction and control. It is a small thing, but one that is important for serious overclockers. Not only can you save 16 profiles, but you can also use one BIOS to test a BETA while the other is your 24/7 OC, and you have full user control.
I just wanted to say again that there are 7 fan connectors, 6 of them at 4-pin PWM connectors. This is a feature we saw on the G1 Assassin; the user has full control over every fan port via EasyTune6 or even SpeedFan.
Now I know what a lot of you guys want to see, how does GIGABYTE do all of this? Well to begin here is the board, nude:
You can see one gorgeous Voltage Regulator Module (VRM). In this next section I am going to tell you and show you where all the saved money from the extra ICs go. Follow me to the most important and most interesting part of this review, Voltage Regulation Section!
Voltage Regulator Module:
X58A-OC Voltage Regulation.
The X58A-OC is an overclocking board done GIAGBYTE style. This means that they have paid attention to the smallest details and expanded the purview of overclocking features. They did this to a large extent in the VRM, but it’s not noticeable by looking at the box, but I will show you how they did it. Any overclocker will tell you that power delivery is one of the most if not the most important part of any motherboard. When it comes to overclocking many think that more phases equals better. While more phases can equal more current output, the truth is that you won’t ever need more than 6-8 phases, that is if each phase can output 30 or more amperes. On the X58A-OC you are getting a souped up Voltage Regulator Circuit/Module (VRC/M) which here on out we will call the VRM. GIGABYTE has reduced the phase count, but increased quality of many aspects of the VRM. In this section we will look at how.
On current motherboards a buck converter which is a DC to DC step down circuit is used to power the CPU. The issue is that we need to step down a lot and carry over a great deal of that power at very high (>90%+) efficiency while tightly regulating the power, to top it off, it needs to be done fast. A buck converter is a switch mode power supply (SMPS), much like your computer’s main PSU, but this one is a DC/DC power supply not an AC/DC PSU which powers all your main components. Changing 12v at something like 30a to 1.2v at 120a is no easy task. Every component including; CPU, CPU Uncore, RAM, IOH (Northbridge), and ICH is powered by a buck converter.
I need to cover a few terms before we begin this section.
Glossary:
PWM: Pulse Width Modulator, in our case an interleaved synchronous buck converter. This device outputs pulses on each of its channels to the drivers that control the MOSFETs.
Parasitic: characteristics of the operation of a component which are undesirable (ESR, ESL, DCR (Direct Current Resistance)) for the mode of operation we want. Example: a capacitor has resistance, a MOSFET has capacitance.
Switching loss: Energy lost per operation
ESR: Equivalent Series Resistance
Switching Frequency: Is simply defined as the frequency (amount of times) the same operation occurs, in our case the amount of times per second each phase is switched on.
Duty Cycle: is thought of as the time the object is on divided by the total time the object can be on. It can be expressed as a percentage or ratio. For our purpose and explanation we can also think of it as Vo/Vi=duty cycle. Duty Cycle is used in the control scheme of many PWMs.
The PWM:
Now the PWM used on this board is the ISL6336, a proven powerful mixed signal interleaved synchronous buck converter. It is the same analog one used on all high-end and mid-end GIGABYTE X58 boards. This PWM is responsible for modulating and sending out signals to the drivers that control the MOSFETs to change 12v at X amount of amperage into 0.9-2.1v at more amperage than your PSU can put out on a good day. At the same time the PWM is responsible for regulating the voltage. How does it do this?
How does it regulate the output?
It senses an output voltage and compares it to a reference voltage that you set, depending on the LLC setting its going to also effect the duty cycle as well as other factors. In a true analog PWM this is done with an amplifier that amplifies the error signal, which is then compared against the reference signal through a comparator, the difference will then let the PWM know how much to change the duty cycle and whether to turn on or off the MOSFETs. Now that is just a simple way to put things, in fact as time progresses PWMs are one of the hardest parts of the SMPS to analyze. For that you need to take a course in SMPS buck converter design.
In the ISL6336 current is controlled in this type of feedback loop, a measurement of the drop of voltage across a current sensing resistor lets the PWM know much current is flowing through each phase so that it can adjust the current per phase accordingly. Taking it to a broader sense of things this is a negative feedback loop, which are all but the most common in the natural world (which has nothing to do with this). If you are wondering, negative feedback is a means of a response loop in where the final result is meant to deviate from the original, positive feedback is to reaffirm the same result. The PWM is always trying to correct the voltage. Quick fact, in the human body there are ONLY 2 positive feedback loops, child birth and blood clotting. Of course Digital PWMs determine how to regulate voltage a bit differently.
So what does Digital Do Different?
Digital PWMs use an algorithm called PID (to help process the error) and instead of an amplifier they use an ADC (analogue to digital converter) so that the error can be dealt with digitally. Instead of turning on the MOSFET and changing the duty cycle until optimum conditions are reached, the digital PWM can calculate exactly how much time to turn on each MOSFET. There are other small differences, but in the end the major part of the difference between analog and digital is the error feedback processing and control. Digital PWM error compensation circuitry has come a long way, and really is up to par with analog PWM’s speed (which was a problem for transient response in the past); on the other hand analog PWMs have increased their effectiveness and precision as well and are up to par with digital PWMs in that sense. To be very honest after talking to some others who know a lot more about this than I, they say that Digital and Analogue achieve the same goals through different means, saying one is better than another is pretty hard to do, and that companies love to market these technologies. Until now digital PWMs had the one good advantage; much easier end user control.
Integration of Analog PWM End-User Control
What does that exactly mean? Well one of the proponents to analog PWMs is that they do not have control firmware that can be tweaked by a motherboard company or end users to fit their needs. They have to basically order the PWM pre set. So to give the manufacturers the ability to control things in their PWM they have to add off die ICs, such as a GPIO (general purpose I/O) to control for instance LLC. In our case they used the iTE IT8275E which controls phase switching and LLC. In the end analog PWMs do have end user control it’s just that it won’t be integrated easily, it takes more work and costs more. Now it doesn’t take a GPIO to control LLC, it actually takes a resistor connected between the FB and VDIFF pins on the PWM to control LLC, but to give users control they put in a GPIO so that the end user can control this. Switching frequency is actually a simpler integration, it takes a resistor to the FS pin, and this time GIGABYTE had the option to just leave it as hardware controlled by DIP switches. GIGABYTE did it this way because they wanted to give the user real-time control over switching frequency.
ISL6336 Features
Now multiple levels of switching frequency control is cool, both sides have it in this round. Another thing that the Intersil PWM has going for it is the fact that the phases (channels) are interleaved. The PWM also runs every phase in parallel. What does this mean for us? Well interleaving the channels means that their individual ripple frequency and ripple amplitude can be multiplied together, this means lower ripple. This leads us to two Intersil technologies, Adaptive Pulse Positioning (APP) and Active Phase Alignment (APA).
The APA scheme can during VERY high current loads align all phases (usually they work one at a time so each channel switches 1/6 of the cycle after the channel before it) with this load scheme the PWM can turn all 6 channels on at ONE time when needed. The APP scheme banks on the interleaving and ripple effects which allows the use of lower bulk output capacitance which has an effect on dampening ripple. That is why on almost all GIGABYTE X58 and P67 boards you see so little bulk output capacitors, because of their high phase count.
The MOSFETs:
(Vishay DrMOS Gen III. SiC769CD rated 35A continuous 1Mhz Switching Frequency)
Next let’s talk about the good old MOSFETs and Drivers.
The DrMOS (DriverMOSFETs) are the same DrMOS used on all other GIGABYTE boards (35A rated continuous output current). These 35A generation III DrMOS have some very good advantages. First off all they are fairly small, which means more room on the board for other components. They also are able to operate at switching frequencies at 1 MHz. They boast over 90% efficiency in multiphase buck topology (which is the case here). Another huge factor is that the driver is integrated. The reason this is important is because driving signals need to be as close as they can to the MOSFET’s terminals. Integrating both the high-side and low-side FETs into a single IC with the driver allows room for other components while helping to increase the overall efficiency of the system, there is no need for extra PCB traces and extra impedance. These MOSFETs are classified as PowerPAK MLP6x6 package. The FETs inside the PowerPAK have low RDS(on) for the high-side FET is 6mohm and even lower for the low-side at RDS(on) of 1.5mohm. They are actually tri-state DrMOS. I was told that if you use LN2 to cool the CPU it will indirectly cool the copper in the PCB, and then through conduction it will indirectly cool the DrMOS which could help push more current (as I have been told). So let’s get down to business, how can GIGABYTE say that this motherboard has a VRM equivalent to their 24 phase boards? To be honest it’s ALL about the current rating of the inductor in this case.
The inductors:
The new: The inductors used on this motherboard are 56nH rated at 50a, GIGABYTE specifically special ordered them for this board, and I was told that the data on this specific inductor is still off limits, but that it has a current rating of 50A. GIGABYTE calls these inductors MPFC (Max Power Ferrite Choke).
The old: On all GIGABYTE boards in production today we have these ferrite core chokes. GIGABYTE isn’t the only company that uses these shiny blue/gray cubes (inductors), EVGA and some others do as well depending on the board and which buck converter. Now the rated current of those 1uH (100nH) inductors (chokes) on previous GIGABYTE boards is about 20A. Inductors always have a rated saturation current. What does this mean? Well rated saturation current is where you being to store energy in the “air” instead of the inductor’s core. Once you reach and exceed saturation current the inductor is saturated with magnetic flux, it loses its inductance and stops operating like and inductor, this poses a problem as it produces electrical noise and can damage other components (mainly the MOSFETs) because the current shoots up as inductance is lost. Different MOSFETs will blow at different times depending on the current rise and the max temperature they can take. The inductor can also be damaged and it depends on the material used and its max rating.
(Example of MOSFETs burning because of current pull through the inductor is too high, above GTX 570)(Special thanks to bassplayer for the picture)
Example of low rated inductors: This is the reason why you see many newer NIVIDA GPUs burn their MOSFETs with stock VRM design. It’s simply the fact that the MOSFETs are putting out more current than the inductors can handle. This is very dangerous. On previous GIGABYTE boards you have maybe 12 or 16 even 24 phases on a board, but GIGABYTE expects you only to pull maybe ½ or ¼ of the current output possible from the DrMOS. What GIGABYTE expects isn’t unfair, it’s the truth, you see these VRMs advertised as being able to handle 1200watt CPUs, or 800watt CPUs, and in reality you can never pull wattage even close to those numbers (more like 300-4000watts max). This is evident by looking at other boards used for subzero benching. As you will see towards the end of this section, I have made a table of the top X58 motherboards; Rampage 3 Extreme, the EVGA X58 Classified E760, X58A-UD9, and the X58A-OC. This is a very common practice, and all manufacturers do it to some extent. In the case of the X58A-OC, GIGABYTE opted out.
The Capacitors:
Usually in a review like this I never cover the inductors or capacitors because they are usually the same on every board. This board uses POSCap TPE, highly conductive polymer tantalum core, ultra low ESR capacitors. Boards such as the R3E and the EVGA Classified, both use a Proadlizer capacitor and then a few normal electrolytic can-type aluminum capacitors. Almost all other boards use normal aluminum can-type capacitors. For the motherboard’s CPU power supply we have a few different types of capacitors. The ones that are important right now are the bulk output capacitors. Each one is rated at 470uF, and has ESR of 6-7 mΩ. That gives us a total bulk output capacitance of 9400uF, a very hefty number.
Here is a table I made with some ratings that are important for us today.
There are many benefits that tantalum capacitors have over their aluminum electrolytic counterparts. First off they have much lower ESR than comparable electrolytic. They also store more energy per unit volume; which means at the same farad rating the tantalum capacitor will store more energy. They also have a much larger temperature range and their reliability is much higher. Almost all tantalum capacitors are rated to go down to -50C. Aluminum electrolytic capacitors show capacitance changes over time, temperature, and even frequency. Although pretty small, it can be a factor for those capacitors right next to the CPU during very cold temperatures. Their lower ESR allows tantalum capacitors to have a higher ripple current rating. This is very important for input capacitors.
(from POSCAP)
Input capacitors:
Usually always rated a few volts (16v ratings common) above the input voltage (12v), these capacitors are under the most strain, especially when it comes to transient loads. While the bulk output capacitors are very important and take the most notice, it’s the input capacitors which take the harshest load changes. The reason is while the bulk output capacitors have the inductors to help smoothen the large current changes, the input capacitors are directly feeding the MOSFETs which step down the power. A higher ripple current (Irms) is directly related to lower ESR, in which tantalum capacitors are well suited.
Output Capacitors:
The bulk output capacitors have a few jobs. A few of the more important jobs are; decoupling the output voltage, feeding current to the CPU while the inductors are recovering from a large transient event, acting as energy reservoir, and ripple suppression. The ideal parameters for the bulk output capacitors are that they have low ESR because, if it’s too high, it can take a toll on transient response. Higher ESR can cause more ripple, which would require a higher switching frequency counteract.
How is the X58A-OC suited?
The output filter:
The output filter is made up of two major components, the Inductors and Capacitors. Replacing all the bulk capacitors on any other GIGABYTE X58 board without calculating the effects of the reduced ESR could actually cause problems with performance. The rule is that if you replace either capacitors or inductors you really have to retest and will probably end up replacing both. That is most likely why they special ordered the Inductors. Together the LC (inductor and capacitor) circuit acts as a low pass filter. This means that it blocks high frequency signals. The MOSFETs output the current in pulses and the LC circuit stores the energy and then outputs it to the CPU. Effectively this method ensures that the output is a steady stream of current, you can’t pulsate power to the CPU like you can to an inductor. The PWM has some cool features, such as APP, which can turn all the phases on at once, I also believe that some other PWM companies (Chil) have this tech too.
The Transient Response:
Transient response is dependent on a lot of different factors, more factors than we can cover here today, but I will cover a few big ones including; efficiency, switching losses, and ESR. Speaking from a pure cause and effect relationship, the switching frequency is one of the key determinates of transient response. Higher switching frequency equates higher transient response. High switching frequency also has many benefits such as much better ripple control and less over and undershoot. The issue with high switching frequency is parasitics of the different components that are affected by switching frequency. Every time a MOSFET is switched (determined by switching frequency) on to output current to the inductors, we have something called a switching loss. Switching loss as defined earlier is a certain amount of power that is lost every operation. You can actually observe this by looking to see which part of the VRM gets hot, most of the time the most noticeable part is the MOSFETs and this is because of a loss of power to heat. If the MOSFETs are getting hot, or the efficiency is too low for design standards, or it’s just too expensive to cool then the engineers have to figure out how to lower the switching frequency or change out the MOSFETs for ones with better characteristics. Right now the parasitics of these MOSFETs are very low, but as you see the VRM heatsinks still can get very hot. The higher the ESR in your output bulk capacitors then the higher the ripple. That is why switching frequency is increased to a certain point where it can help negate the effects of the ESR. One other very important factor that affects transient response is the speed of the error feedback loop. But that is a discussion for another day.
Current output and some other factors:
Below is a table of some of the best top tier motherboards from different companies for the LGA1366 platform. You can see MOSFET current output rating VS inductor current ratings. You can see how almost all the VRMs are limited in one way by their inductor current rating, except for the X58A-OC.
(Above are the top best X58 OCing boards. These are also the best VRMs each company has to offer for X58. Current output is not very important at this level, I just wanted to show how the inductors limit current on every model, except the X58A-OC. 400watts is enough for any max OC with a LGA1366 CPU, 350watts is pushing it. ) (One other NOTE because many ASUS fans are going to challenge this. There have been no reports of the Rampage 3 series MOSFETs or inductors being damaged, this is for two reasons. First reason is that it is a properly designed VRM, which has very high quality parts. The current output isn't low its actually higher than ~97% of the motherboards out there today, with the ones up top being exceptions. Current output isn't the most important factor when it comes to VRMs, as long as your VRM can output 250amps you are golden for any OC. The other thing is that the MOSFETs used on the Rampage 3 series are top quality, very high quality and very good build means that they can take a harsh beating and survive, that plus saturation current is never really going to be reached is why even though current output is lower than the others, its still one of the best)
Final words on the X58A-OC board’s CPU VRM?
The X58A-OC has a really great VRM. To incorporate a higher switching frequency, and to increase overall transient response and ripple suppression, GIGABYTE revamped the output filter topology significantly. The issue with changing a current GIGABYTE design of the Inductor/Capacitor circuit is that if you drastically change an aspect of either the inductors or capacitors you have to change the aspect of the other. This VRM is built to work over a large range of switching frequencies. This is just speculation but I am guessing Hicookie wanted increased transient response like on the UD9, wanted to cut down on overall size(easier to insulate), wanted full user control over switching frequency, wanted to keep current output extremely high, and wanted to improve ripple control. Many of these wants are hand in hand. For instance when you increase switching frequency you are also able to decrease inductance required as well. Less phases for GIGABYTE means that their 6 phase PWM’s bandwidth isn’t stretched thin among 24 phases and the VRM will have better transient performance as well. To help really increase the effectiveness of ripple suppression GIGABYTE wanted to do something, and luckily for them they were able to change out the whole output filter (inductor/capacitor) and increase switching frequency.
In the end we have one of the most capable VRMs to be built on a motherboard. Not only in current output, but also in ripple suppression and transient response. Of course there are other options out there as well, and you can see how they add up. Up to now the EVGA Classified has had a crown on transient response, but to test transient response is very tough. There was an article in the OCMag which compared the transient response of the EX58-UD5, the Rampage 2 Extreme, the EX58-UD7, and the Classified among other boards. The Classified was the transient response winner, but now that we have this new GIGABYTE VRM, it might just come close or even beat the Classified. Of course this is mostly speculation, and only real world tests will show. I just wanted to point out that the change in inductors and capacitors are very great, and GIGABYTE isn’t playing when they say this design can be compared to their 24 phase design. It can easily be compared, and in some cases win.
The Uncore VRM:
Above is the Uncore’s VRM. The Uncore is the part of the processor that houses the internal memory controller, and that is what its best known for. It’s actually everything in the processor that is not the cores, such as L2 cache. The Uncore doesn’t use that much power, and is almost well equipped on every motherboard. Above you can see the different parts, same Inductors and Capacitors.
The MOSFETs are different and so is the PWM. For this GIGABYTE used the ISL6312, and it’s not the only ISL6312 on this board, we have two more which power the RAM and the IOH. The MOSFETs are made by Renesas (uPA2724UT), and they can output a nice 25A per phase. The ISL6312 is a native 3 phase PWM with integrated drivers (no need for external drivers), but has the ability to be a 4 phase PWM if there is an extra external driver. The Uncore is well equipped.
The RAM and IOH(X58) Voltage Regulator Modules:
As you can tell by now these VRMs are setup just as the Uncore VRM. Without regard to accuracy I outlined which components go to which output. The BLUE=IOH VRM and the RED=RAM VRM.
In orange are two ISL6312s, one per each VRM. The same Renesas MOSFETs are used, one pair (high-side and low-side) per phase at 25A per phase. Same high quality inductors and capacitors for each VRM. If you count we are missing 3 MOSFETs, those are can be found on the underside of the board, the only MOSFETs on the underside.
(3 missing MOSFETs, and two of these ISL6312s)
You are probably wondering how the other parts of the board are powered. The thing is that those other parts like the SATA and the ICH (ICH10R) don’t really require a variable output, and if they do then the need for regulation isn’t so much. The ICH has something like a few watts of heat output so it barley draws any power. Just like on all other GIGABYTE X58 and even P67 boards, the ICH is powered by a small ISL PWM hidden under the battery holder, and uses Nikos MOSFETs. As you can see above that IC says sexy on it, even though the IC really isn’t too sexy, I thought it was cool. Actually that chip is an operational amplifier which powers some onboard components that don’t need variable output. There are 4 or 5 of these SEXY chips on the board. LDOs (LowDropOut) regulators, which are linear regulators, are used to power most everything else.
The DDR termination voltage is actually powered by its own little IC, the Richtek RT9173D is a high power linear regulator used to specifically power the DDR’s termination voltage. Max 3A output. To end off this section we have below a tantalum POSCap replacing some ceramics.
Overclocking and Feature ICs.
These next ICs are the ones responsible for the temperature monitoring, fan control, and even the overclocking features.
This IC, the NuvoTon NCT2032D is obviously the IC that controls the overclocking buttons. It is not listed anywhere on NuvoTon’s website. NuvoTon is known for providing the same type of features on other boards, but not to this extent (multiple BLCK levels AND multiplier clocking). With the ability to preprogram a very nice 4 GHz overclock on touch, this IC provides us with true hardware overclocking. Hardware overclocking is nothing new, I remember that my old Abit boards had the uGURU standalone OC panel, and other boards have that too now. GIGABYTE unlike a lot of other motherboard companies still makes their own motherboards. Not having a contract in place, they would have to make a very large amount of special PCB design to have it for their boards. This is why they opted to keep the buttons on the board. Honestly the buttons work flawlessly, I was able to walk my 990x to 4.2 GHz without even increasing the Vcore, and of course it was only CPU-Z stable, which isn’t that practical. I can see the precision and variety of overclocking buttons really help those reach insane CPU-Z clocks.
Above we have the iTE IT8720F and Winbond W83L786, which both are hardware monitoring Super I/Os. The iTE monitors a temperature, provides the PS2 ports in the back, monitors voltages, and even provides CPU fan control. The Windbond gives us the rest of the fan control and temperature sensors. Both of these ICs are found on the G1 series motherboards, where they provide the same features. Together along with the NuvoTon IC they make one feature RICH overclocking motherboard.
NEW IC’s Marvel SE9182, EtronTech EJ168A
Apart from the new NuvoTon IC pictured above, we have the new Marvel SE9182 and the new EtronTech USB 3.0 controller. These make up the new ICs found on this board.
EtronTech is a small company that is known for memory ICs, and now they have come out with their first System IC, the EJ168A, which provides us with 2 USB 3.0 ports. I have seen this IC on a few other boards to date, and some to be released, so we will see this new USB 3.0 host controller in other boards as well. It is nice to have some USB3.0 competition with Renesas/NEC who has owned the field until now.
The Marvell SE9182 is another beast all together. Right now the self proclaimed KING of SATA6G off die solutions, this IC out performs its old SE9128 counterpart without hesitation. It even gives Intel’s SATA6G a more than fair fight, even though I’d use an Intel controller for RAID0 with SATA6G SSDs, this controller should handle one just as well as Intel SATA6G.
The IOH/ICH and Other ICs
Here we have our gorgeous Tylersburg Intel X58 IOH. With a native 36 lanes of PCI-E, with only 32 PCI-E lanes useable, this IOH gives us the power to run 2-way and 3-way SLI as well as up to 4-WAY CrossFireX.
The X58 IOH has the PCI-E lanes, and these ASMT ASM 1440 PCI-E switches distribute them. There are 8 of these switches, and they configure the PCI-E lanes accordingly.
Here we have our great ICH10R ICH (Southbridge), this IC provides us with all the connectivity we could want, and then more. Of course most of that connectivity was left out on purpose.
Next we have out trusty clock generator. The ICS9LPRS914HKLF so you can put it into SetFSB, OH WAIT! You have hardware buttons that can do the same thing!!! Here we have both of our 16MBit BIOS chips, each one can be selected by a switch, which is a great feature for GIGABYTE to finally add in.
Finally we have our ALC889 and then our RTL8111E which provide us with audio and our LAN.
That covers all the ICs on this board except for some OAMPs and LDOs.
Now let’s take a look at the cooling on this board.
The Heatsinks! And Cooling
As you can see these heatinks are of a brand new design. They are much different than any other X58 board and even more different than any other GIGABYTE board. Each heatsink has LEDs in them. I like them in the heatsinks rather than on the board. I think GIGABYTE understands that if an LED doesn’t have function then it’s just decoration. Here the heatsinks are part decoration in themselves, and I personally enjoy how this board glows in the dark.
If we take a closer look we can see that these heatsinks have all sort of different cuts and angles. The heatsink for the IOH and the VRM is pretty damn heavy. This aluminum heatink actually does have a heatpipe in it as well which connects the VRM block to the IOH block. We can also see that the heatink covers both the main and uncore VRMs. The heatink also covers the inductors/chokes, but doesn’t provide them with any cooling, which of course is unneeded.
This picture shows us the heatpipe as well as the thermal interface material used. The IOH is given TIM as normal and 100% of the DrMOS and the Low RDS(on) MOSFETs for the CPU power rails use thermal pads. The USB 3.0 EtronTech IC also has its own thermal pad, probably just to even the hold down balance. These heatsinks honestly do a good job, and you have to use them for air cooling. When you use Dry Ice or Liquid N2, you don’t necessarily have to use these heatinks, because the copper in the PCB will cool down these components through conduction. The problem with not using these heatinks is that you still are required to cool down the ICH (Southbridge) and the IOH (Northbridge), many users use ghetto solutions like a small Northbridge heatsink from a Pentium 4 era motherboard. The good thing GIGABYTE did is that they disconnected the heatink into two parts.
The introduction picture is of the ICH’s heatsink. This heatsink has a very long extension that covers a lot of PCB. This might actually help extend its surface area to help cool it better.
You guys are probably interested in what lights the heatinks up, and the answer is LEDs. The heatsinks have LEDs, the main heatsink has many and the ICH heatsink has 3. I think they give a really nice ambient glow.
Here is a close up image. The LEDs are surface mount, so replacing them isn’t too hard, but finding replacements are. Of course you need to be handy with a soldering iron to do this, and they are just fine the way they are.
Now above are probably two images that don’t seem like they come from the same board. Metal screws are how GIGABYTE has been attaching their heatsinks on to their P67 and later revision X58 boards. This time we see the use of metal screws for the ICH and IOH, 8 all together, 4 per block. Surprisingly they used plastic push through fasteners for the VRM part of the heatsink, 3 of them. I honestly have no idea why, but as I was removing the heatsinks and putting them back on I realized something. It’s a real pain to have to unscrew these heatinks so often and then keep track of them, using half pushpins and half metal screws makes some sense. Use the metal for where it counts, and use plastic for where you have variable pressure requirements. Many would want all metal, but after removing so many heatsinks so many times, push pins aren’t too bad of a concept.
The Board in Action and Some Initial Results
To start this section off I will start with how the board looks once you have it setup, and how everything works. I did some quick air results on two different CPUs as well.
First the only thing lit up when the board is off is the power button. It’s orange and glows pretty nicely. You can find your board even in pitch black!
The board looks fantastic when it’s lit up. They did a really nice job with the lighting. It would look pretty sick on a bench. I am sure some will build whole systems around this orange glow.
The OC button area is lit only when you press the “4GHz” or “Gear” buttons. They stay lit when active.
The +/- buttons only light up when you press them. Here I have pressed all of them.
Here we have the board setup with a nice air cooler. I have also plugged in the SATA power cables. The SATA drive is right next to the board and the daisy chain of the SATA cables makes it easy to hook up your SATA drive right next to the SATA ports. In a case these plugs are well positioned as well.
Now let’s take a look at the spacing:
As you can see the spacing with a very large 120mm fan like this is pretty right. The fan actually ends up just fitting correctly; it sits on the heatsink in this case, but doesn’t make it hard to fit, just a bit different than usual.
The other side of the heatsink as you can see is a tight fit. Your RAM should be in the orange slots on X58 GIGABYTE boards, and then the black ones once you have filled the orange ones up. Notice the difference in thickness of the two fans.
This is a cool shot of the back-panel in use. A little air flow might not be bad.
Here we have all 7 of the included connectors hooked up to the motherboard. It’s a very standard but nice design. Something I would like to see on all GIGABYTE boards.
Now for some easy overclocking results done on air!
All I did for this was press the 4G button. I had two different dual channel kits in this board, one a 1600 MHz kit(8,8,8,24) and one a 2133mhz(9,10,9,27) kit. Both are of the same density (2GB each). Took my RAM to 1600 MHz at 7,9,8,24 at 2T, which is very impressive. The CPU was taken to 4 GHz, but with a 20x multiplier 200 MHz BLCK, very easy OC. The voltage on the CPU was 1.36v and the voltage on the IMC (QPI/VTT) was 1.36v. It is actually a very impressive OC for any auto-OC button.
MAX AIR Overclocks!! Two processors, X58A-OC, and some easy aircooling fun.
Your overclocks WILL vary from mine.
MAX OC for i7 930
This is very nice, I was never able to get to 4.6ghz with my i7 930, it took 220blck, and the board did it with ease. This is a personal best for this CPU on air.
MAX OC for i7 990X
Here we have my 990x, which does 5.2 GHz easily, it also did 5.2 GHz on my G1.Assassin, and the difference is the Vcore used. I had to use 0.03V less!!! That is pretty impressive. There was more room in this chip, but that is for the LN2 and DICE. No need to degrade this beauty.
ONCE AGAIN, your OC results WILL NOT be the same!!!!!!!!!!!!!!
The conclusion and preview:
Above is a picture of my pot on the board. It looks like it’s going to be really fun. More results will come about performance very soon. I will compare this board against my other X58 board, the G1 Assassin. This board’s VRM really packs a hefty punch; the switching frequency DIP switch is just a little extra. The new POSCaps and 50A inductors really make this board worth it. This board is not only simple to use, but it’s also a really great overclocking board. For instance in the BIOS once you save and reset it zips through POST like it’s no one’s business. On top of that BIOS quick boot, C1E and EIST are DISABLED at stock. I was able to reach either a higher clock (i7 930) or use lower vcore (0.03v less). I like the buttons, they make overclocking so easy, CPU-Z WRs will be snatched without a problem, because you have total control over increasing CPU clock through hardware, and even in minute levels.
This board really is worth the money, original estimates put it around the price of the UD5, but of course that is always how it goes. I expect it to be a bit higher. This board is for overclockers just like the G1 Assassin is for gamers. It still can do everything any other X58 board can do; it just has much less connectivity. This board really packs a lot of great overclocking features. The way GIGABYTE has focused in on different market segments is impressive to say the least. This model is one I hope to see included in many future GIGABYTE lines. Honestly voltage read points and the BIOS switch are long overdue in GIGABYTE boards. This board is a bit late as well, considering X58 is at the end of its run, but it’s better to be late and win than never enter the game. It glows pretty nice in the dark too! This board is the Ultimate OC board in my opinion, and my favorite motherboard ever.
Pros:
High quality VRM and with switching frequency control.
Whole board has POSCaps only.
Triple slot PCI-E spacing on ATX size board
BIOS is stable and steady, tuned for overclocking
OC Buttons are excellent
Voltage read points and BIOS switch (finally)
Extra PCI power form SATA power cables.
Cons:
Heatsink is just going to cut it, it’s a bit big.
Plastic push-pins for half of the hold down
VTT is outputted higher than what is set, by a fair amount ~0.02v to 0.03v
A little late in the X58 game.
Now we are done with part 1(physical review), part 2(performance review) is coming very soon. Stay tuned!
I would like to thank everyone at GIGABYTE who made this review possible!
PLEASE feel free to PM me if you want me to change something in this review because it might be wrong. I will need proof that it is wrong as well. Everything in this review has been checked over and over.
I also know MANY users are interested in how the VRM works, I have a lot of good links to current reading material if you can understand basic electronics lingo. Just ask.
Reply With Quote
Originally Posted by Brian@ASUS
Bookmarks