Z68X-UD3H Review Exclusive*

[sin0822]

Z68X-UD3H Review
-Steven B.



Today we take a look at GIGABYTE’s Z68X-UD3H, their mainstream consumer motherboard. If we sit back and take a look at this motherboard, as we do every motherboard, we can see right off the bat that it is feature filled, yet modest within its own respects. It has USB 3.0, extra SATA6G, 1394A, 7.1 audio codec, and even PCI. It has every feature a consumer would want and even more. So you have to ask, how is this board priced so well? The answer is that it drops off overclocking features, voltage regulator phases specifically, for a more feature rich feature set. Its entire spec list is pretty long, and it seems that its features match the price pretty well. This board even has the USB 3.0 turbo mode that we have seen on the Z68X-UD7, of course to a bit lesser extent. This board has the Ultra Durable 3 label, which means that its PCB has 2oz of copper opposed to the 1oz standard, 100% solid Japanese Capacitors, and low RDS(ON) MOSFETs for peripheral voltage regulators. The extra durability factor associated with the Ultra Durable 3 label is not something many other boards in this class have to offer; in fact it is what makes the GIGABYTE board shine. It seems that with Z68 GIGABYTE felt like they needed to give their mainstream boards a VRM upgrade, but not in the phase count. Instead GIGABYTE opted to replace the Low RDS(ON) MOSFETs with Driver MOSFETs that they use for their high end boards, which can increase efficiency, power output, and thermal spec. Now the phase count on this motherboard is a bit on the lower side, in total it is a 4+2+1(CPU+Uncore+iGP) phase design. So in theory each of the DrMOS can output 35A to the CPU cores, so we can look at 140A output, which can land us about 220watts for the CPU cores, that is about a 5-5.2 GHz OC to our estimates. We notice that overcurrent protection enacted by the PWM, is limited to +10% through the BIOS settings. On the other hand, we noticed the PWM frequency setting. In the case of this board, at stock it’s ~360 kHz, and can be increased to 680 kHz, the highest we have seen on any GIGABYTE LGA1155 board thus far. We will discuss the meaning of this and other features through the review, so sit back, relax, and enjoy.
Box, Accessories, and Specs.







The box is pretty well designed; the black background matches the black PCB.
Sin’s Take: Now compared to the Z68XP-UD3 this board has the fancy black PCB, something that users are really demanding. Now the UD3 and the UD3H have a very close feature set, the UD3H costs a bit more, and the features are almost identical, so it is up to you whether or not the extra good looks are worth the price, I say go for it if you have the money.

Now these are the accessories. There are a total of (4) SATA6G cables, all black, as well as a SLI connector. We have the back panel, as well as the manuals, and some GIGABYTE sticker.


Here are all the specs:




The board is well laid out; the board looks like a standard Z68 board. The PCB is very busy



Here we have the backpanel, everything you need is right there, including Display Port.



Looking at the top left side of the board we have the CPU socket area. It is well laid out; the only fight we have to pick is that not all of the DrMOS are heatsinked. On a board with only 4+2+1 phases, those phases will be running a bit hot when the CPU is overclocked. On the other hand, the lack of heatsinks surrounding the socket means that you won’t have any trouble fitting a large heatsink in that area.
Sin’s Take: GIGABYTE could have heatsinked all the DrMOS, but if you think about it, it won’t increase the amount of current they can produce, so in the end if you are going to overstrain them, you are going to overstrain them with heatsinks on or off. I think maybe this way it’s sort of a warning, this board is not meant for high frequency benchmarking, so take it easy please.



Here we have the memory DIMMs, the Trusted Platform Module connector, as well as the memory phases.
Sin’s Take: Memory overclocking should be very good on this motherboard, as it isn’t lacking in producing memory power. The BIOS allows for 5mv steps for DDR3 voltage control as well. We see that GIGABYTE also left room between the first GPU slot, and the DIMM latches, you can replace your memory with a long GPU in place.



Now you are probably wondering, why aren’t there 8 SATA connectors? Well we have 3 black SATA3G from Intel PCH Z68 (2 angled & 1 straight), 2 white ports are SATA6G Z68, and the 2 gray are SATA6G Marvell. The last SATA3G is on the back panel as eSATA, a common design technique. This board has many, many, many internal headers. Really anything you want, this board has in terms of connectivity, it even has a COM header.
Sin’s Take: After using a few other boards, I am glad that GIGABYTe color codes their front panel connectors, as well as the fact that they use a two pin design right above the front panel connectors for the clear CMOS. What I do, which I am sure others do as well, is that I hook up my reset button on my case to the Clear CMOS jumper, so that I can OC without on-board buttons. Just make sure if you do that, don’t hit reset while the power is on.



Here we can see the triple slot GPU spacing. This board support SLI technology, and as we will discuss later, GIGABYTE has given it PCI-E 3.0 support. Now this board has no PCI-E 3.0 PCI-E switches, but we will discuss later how we can still have some PCI-E 3.0 capability, something many people have been wondering about.
Sin’s Take: What people don’t realize is that on a board with 16x PCI-E lanes, like this one, only half the CPU’s PCI-E lanes are connected to PCI-E switches, so they can go to either the first slot for 16x or second for 8x/8x SLI/CF. The rest of the 8X lanes, the first 8X to be exact, are hard wired to the first slot. That is how GIGABYTE can say these boards are PCI-E 3.0 compatible.



Now this board has 4 LEDs, they are for the 4 phases. They will almost always be all lit like a Christmas Tree.


Deep Look at ICs, VRMs, and PCI-Es.



PCI-E 3.0 Capability, Fact or Fiction? Or in-between?



Please take a look at that picture above. You can see I have outlined what I want you to look at. You can see those in green are the PCI-E lanes from the CPU. You will see that half of them go to the area of the PCI-E switches, and the rest go to the first 8x of pins of the first 16x PCI-E slot. This means that when a PCI-E 3.0 capable CPU, Ivy bridge CPU to be exact, is placed into the board you can use PCI-E 3.0. The first 8x of the first 16x slot will have PCI-E 3.0 capability, allowing half of the first slot to have double the bandwidth, which PCI-E 3.0 provides. It might not be like the Sniper 2 which has full PCI-E 3.0 support, for all 16x lanes, but it’s better than nothing. GIGABYTE announced BIOS with PCI-E 3.0 support, this board was on the list. A board like the UD7 which uses an NF200 cannot have this PCI-E 3.0 capability as all the lanes go through the NF200.





The rest of the PCI-E/PCI layout. Notice you can see how the USB 3.0 turbo mode works. When USB 3.0 turbo mode is enabled, it might reduce the number of PCI-E lanes for the GPUs. The Turbo mode will provide you a good performance boost if you have a lot of devices hooked up to the board, but it can hurt GPU performance.







Now I want to talk more about this small voltage regulator. Now we see a ton of capacitors, more than let’s say the Sniper 2, which has 10x560uF, here we have 12. The increased capacitor count will help with ripple suppression. Now because of the reduced phase count, the phase interleaving affect, which is used to reduce the ripple current, is now reduced. This means that an increased capacitor count is required to keep ripple at bay, as well as a higher switching frequency to help transient response and also help keep the ripple at bay. We usually see high switching frequency with lower phase count systems, and it makes sense since the MOSFETs will run hotter, and thus are easier to cool down in fewer numbers. On this board I am glad to see lower OCP setting in BIOS, on this board we have up to 10%, on the Sniper 2 it’s up to 60%. That is 50% more, but then again the Sniper 2 has double the phase count. Why am I glad that this board doesn’t allow higher over current protection? The reason is because I would rather be told my limit than burn out a DrMOS, and that OCP is a way of doing so. I think if you surpass the limit that the CPU might throttle, or that the PWM will limit the current thus limiting the performance. The switching frequency on almost all GIGABYTE LGA1155 boards is 264 kHz, on this board it’s ~360KHz at auto setting, and can go up to 680KHz for best ripple control. Now please be careful, do NOT just crank up the switching frequency, you will have a much higher chance of overstraining the DrMOS, but also lower switching frequency results in higher current output from the DrMOS. So leave it at auto, or take it to +42-68% for good balance. The DrMOS have highest current output at around 300khz.

[sin0822]

It is not by chance that we see so many less output capacitors on the Z68X-UD7 compared to the Sniper 2. In a multi-phase system we can see that if we have for example 4 phases, that the 4 phase will have combined ripple frequency 4 times greater than any one phase alone. Higher ripple frequency and lower ripple amplitude allows for less output capacitors as the ripple is being suppressed by the interleaving of the phases instead of solely by the bulk output capacitors. That is a theory and in practice is does apply to certain extents. In a 4 phase system this interleaving affect isn’t as great as if it was a 12/16/20/24 phase system, so we see an increased amount of output capacitors. That is one reason why a 24 phase system has a switching frequency of 260 kHz, while this 4 phase is at 360 to 680 kHz. Now motherboard voltage regulators are very complex, and very hard to understand. But in the coming months I will write an article comparing and outlining them, hopefully making them easier to dissect and compare. I will also do testing on this board specifically to see if switching frequency has that much effect on ripple. For now, know this, it’s a give and take game. I will compare on this board 160khz vs 640khz. Then i will compare this board directly to the Sniper 2 at the same Switching frequency to see if increased phase count helps. Then we will compare against the UD7. Comparing with all these GB boards helps, because we are using the same DrMOS and PWM(2 different models (one 4 and one 6 phase) same company(Intersil) and same mechanisms) we will be able to see the affect of a single change in VRM compost hardware's affect on the whole while keeping other factors constant as best we can. WE will also use some other VRMs of other boards int eh comparison.


Here are the peripheral Voltage regulators, we see that they use low RDS (ON) MOSFETs, no crappy Nikos MOSFETs present for dc-dc buck converters on this motherboard.

The Heatsinks:

They only cover half the VRM, 1 CPu phase and 1 iGPU phase is left uncovered.

The BIOS:






PWM Frequency & LLC Examination:


Overclocking + Benchmarks
Overclocking was pretty straight forward. Everything acted as it is supposed to, the only issue I have is with the LLC controls. The best this board can do with this current LLC (which is partly set by BIOS) is a 70mv drop under load. No LLC setting allows for increase of voltage under load, but that might be partly because of the low phase count. In this case I had to set a very high idle voltage to get 1.5v load, about 1.58v, something I would never recommend one do for a 24/7 overclock. After tweaking this board, I have to say that a 4.5 GHz overclock is simple with this board; you can probably pull of a 5ghz overclock if your CPU likes lower voltages. Otherwise buy something like the UD4/5/7 if you want to have a higher 24/7 overclock.
I did like the fact that I have 5mv control on almost every voltage, including the DRAM voltage.













Now we see that the board isn’t really tweaked that well for all these benchmarks, but its performance is very good for what it is, a $160 LGA1155 motherboard. It doesn’t win every benchmark, but as you see in some cases such as SuperPI 32M it is very quick. If you look close at the 3DMark benchmarks, it has good GPU performance.
Now we will look at the A3850+ A75-UD4H VS 2600K + Z68X-UD3H, but this time only using the internal graphics processor of each system, no discrete GPU.






If we compare against the APU A3850, the 2600K wins in every CPU benchmark, of course there is a slight frequency difference with both at stock. For IGP gaming, there is no question that the APU is stronger, but in video transcoding, it seems that the Intel has won.

Conclusion:

When I review a motherboard, I sit back and take a look at it for a while. I stare for probably a good 30 minutes, while I take time and analyze things. I have to say I spent a good amount of time testing out this board, and I was surprised to find very little BIOS issues. Usually I might find a cold boot, boot loop, or BIOS glitch with memory, but this time I found none of those things. Instead the debugging process was pretty straight forward. The only part in which I think this board falls a bit short is the CPU voltage regulator. I think that GIGABYTE should have given it a few more phases, but for $160 you can’t ask for too much more. I did realize something though, in the past on P67, the UD3 series and below used Low RDS (ON) MOSFETs for the CPU core voltage regulator, in which each phase could produce 22-25A. When I look at the specs of these Driver MOSFETs (DrMOS) we see that they are rated 35A. So if you think about it, this board would be as good as a P67 variant that has 6-8 phases. It is an example of how the number of phases isn’t a good way to judge a board, instead judge by the quality of each phase. Of course 4 phases for the CPU cores is the minimum, but the fact that they are DrMOS makes it a bit better. On a side note, if this were LGA1366 would you ever dream of a 4.5ghz overclock on a 4+2+1 phase motherboard? As we all known Sandy Bridge CPUs are very power conscious, they are extremely efficient, and they have lower TDPs than past platforms.

This allows manufacturers to reduce the amount of phases. Of course we have to realize that this board is not meant for benchers, it’s meant for the mainstream user and occasional overclocker. It is meant for that person who need a 24/7 computer to last them years to come without failing. GIGABYTE delivers in that respect, and even put a few features in there that are just excellent for the price. The fact that this board rocks the Ultra Durable 3 label means that isone of GIGABYTE’s best built motherboards, and that is a great sign. GIGABYTE is known for making some of the best built motherboards, and having the best quality control, because they actually have their own factories. In the case of the Z68X-UD3H, they took a mainstream motherboard would you find in any desktop PC and slightly tweaked it to match the lower end of the overclocking spectrum. This board will allow you to OC your processor to your 24/7 OC, it just won’t be able to do any 5.2 GHz + benchmarking. It is packed with every connection you would ever want out of a motherboard; it even has legacy PCI and COM ports. For $160 you really can’t beat the deal that GIAGBYTE is offering here. Even compared to Asrock boards in a higher price range, this board has the same features. The bottom line is that if you want an affordable motherboard that will sustain a mild overclock for many years, is feature packed, and stylish at the same time, then this motherboard is for you.


What Sin Likes: I really like the fact that GIGABYTE kept the black PCB, and filled the board up to its neck with features. The fact that the board has Display Port means that those wanting a resolution above 1920x1080 can use the video out from the board. I like how we have good PWM controls through the BIOS, and I like that GIGABYTE is giving finer voltage increments for the DDR3 Voltage (5mv now). I like how GIGABYTE was able to implement USB 3.0 turbo mode on a board this affordable, as that is a feature you find on boards like the Z68X-UD7. The board’s performance was pretty good, in terms of benchmarking it wasn’t as ground breaking in the CPU segment, but it did a fine job in the GPU segment. Oh and the BIOS I was using to test, the latest Beta BIOS, is very solid. No boot loops, no cold boots, no odd bugs.

What Sin Thinks Can Be Better: I think that GIGABYTE should have put heatsinks on all the DrMOS, and maybe even added a few more phases. I think that Load Line Calibration needs to be further tuned as all 10 levels droop under load. Could have some more fan control.

Bottom line: In terms of connectivity this board has everything ones heart can desire, and then a bit more. To many that is all that matters, that the board have more USB 3.0 and as much SATA6GB/s as possible, maybe a COM port, decent audio, and PCI for connectivity. GIGABTYE does a good job supporting USb 3.0, extra power for mobile devices, and the so. Their boards use very good quality controllers for USB 3.0 and SATA6G. It seems that they dropped the crappy JMB controllers for SATA, and upgraded the audio on their low end boards to ALC889 (better speced than 892) .

Price: $160 USD


Score: 9.2/10 (Very Good)
Strong Point: Price versus Performance
Strong Place on/in: 24/7 machine, home theatre, media center.

[Chike]

GIGABYTE is releasing new revisions (1.3), at least for the higher end boards that seem, at elast according to their overview page, to support PCIe 3.0, the UD3H included.

[Chike]

It's a pitty the video transcoding does not include the UD7.
It's important IMO for users who need their PC for this purpose.

[sin0822]

well the UD7 doesn't have iGPU out, so of course its going to be slower. its general knowledge that the iGP in the SB CPu is strong than a discrete GPU. I wanted to compare IGP to IGP.

An yeah i heard about the rev 1.3 boards, they should be interesting.

[Chike]

well the UD7 doesn't have iGPU out, so of course its going to be slower. its general knowledge that the iGP in the SB CPu is strong than a discrete GPU. I wanted to compare IGP to IGP.

When someone that needs a system for video transcoding he might think "I'll buy the biggest and meannest" board and actually shoot himself in the foot.
It's maybe obvious to you that UD7 will be slower but not to all users.
With all the benchmarks I have seen till now I still don't know if i7-980X is faster than i7-2600k or i5-2500k with Quick Synch Video.
This would not be an issue if GB made all boards with the iGP enabled like other manufacturers

[sin0822]

They do, their top model for that purpose is the Z68XP-UD5. I didn't have a chance to run the transcoding on the UD7, but i can find some time and do it if you like.

[Chike]

It's not for me. I am probably skipping SB and IB.
I just think it's fair if performance comparison includes video transcoding.
Comparing it to the 3850 that was inferior in all other tests is not very impressing.

[sin0822]

well i was just trying to compare iGP to IGP, to see if video out on the Intel system is better, basically targeting integrated GPu solutions instead of discrete.