Gigabyte Z68XP-UD5 LGA 1155 Motherboard Review

[Eldonko]

Test Setup and Testing Methodology

Test Setup and Testing Methodology

Our test setup consists of an Intel i7 2600k Sandy Bridge CPU, Gigabyte Z68XP-UD5 motherboard, and a kit of G.Skill memory. Here are a few shots of the setup and hardware:

​

First we have the whole test bench up and running. As you can see we used watercooling (Apogee XT) for the CPU along with two heatercores. A second GTX 470 was also added for SLI testing.

Above you can have a closer look at the CPU, memory and PSU used in the setup. The 2600k used for testing was from batch L040B165.


Overclocking Methodology

At Hardware Canucks, we understand we have a diverse reader base and to represent a variety of user types, so we put the Z68XP-UD5 through three types of tests.

• Beginner Overclocker - To represent a beginner overclocker or a mainstream user that wants to have a quick and easy way to get some extra performance we used the QuickBoost Level 3 setting found in the ET6 utility.
• Experienced Overclocker - To represent an experienced overclocker that is looking for the optimal 24/7 overclock to maximize system performance while keeping voltage and temperatures in check we overclocked the Z68XP-UD5 manually and stopped when we started to get concerned with voltage levels.
• Advanced Bencher - To represent the bencher that is looking for short benchmark runs at absolute maximum CPU and memory clocks we tested with sub-zero cooling and higher than recommended voltages. More on this is found in the sub-zero testing section.

We did stability testing a little differently for the Sandy Bridge platform than usual. The main stability test used was Linpack (LinX version 0.6.4) with memory usage set to 2,560MB and 25 loops run. In the enthusiast world, Linpack is a benchmark designed to measure performance on Intel CPUs in GFlops. However, it's also a very useful tool for checking the stability of a CPU and memory. LinX picks up very quickly and if you are able to complete a 25 loop test with the specifications above your system is likely stable or very close to it. Typically we would run LinX much longer than 25 loops and add in Prime95 and OCCT; however there have been reports of degrading Sandy Bridge CPU overclockability with running these types of torture tests for long periods.

To avoid risking damage to the processor, after LinX stability was achieved, 2 runs of 3DMark Vantage and 2 runs of 3DMark 11 were run to test 3D stability. Once an overclock passed these tests, this is the point deemed as “stable” for the purposes of this review.

Windows 7 Service Pack 1 was installed to take advantage of the Advanced Vector Extensions (AVX) with Sandy Bridge processors. Intel AVX is a 256-bit SIMD floating point vector extension of Intel architecture. The BIOS used for overclocking and benchmarking was version F4, dated 09/01/2011.


Benchmarking Methodology

Benchmarks in the System Benchmarks section will be a comparison of the i7 2600k at stock speed, at auto overclock speed as set by QuickBoost, and at maximum 24/7 overclock to give an idea of how much performance a user can gain when overclocking the Z68XP-UD5.

For stock testing, optimized defaults were loaded putting the CPU at 3,501Mhz (35 x 100.3) and memory at 669Mhz and 9-9-9-24 1T timings. Optimized defaults enable Turbo by default hence the 3,501Mhz instead of 3400Mhz which is the stock speed of the processor. The auto overclocked speed on the i7 2600k for ET6 (QuickBoost) was 4,212Mhz (42 x 100.3) with memory at 669Mhz and 9-9-9-24 1T timings. The overclocked speed on the i7 2600k for 24/7 stability was 5,002Mhz (50 x 100.03) with memory at 1,067Mhz and 9-9-9-28 1T timings. Windows 7 Ultimate 64 bit was used with SP1 installed and SSD caching in enhanced mode was used.

 

[Eldonko]

Overclocking Results

Overclocking Results

As we typically do for reviews, we put the Z68XP-UD5 through countless hours of overclocking and testing; from auto overclocking through QuickBoost to manual overclocking where tweaking is king. On top of that we froze our 2600k and ran some 2D and 3D benchmarks to see how the board can handle the coveted edge of stability and high voltages runs.

Auto Overclocking

Gigabyte’s QuickBoost is a tool that allows for quick and easy CPU overclocking by the simple press of a button. There are three different levels of CPU performance, each giving a little extra speed. Basically all you have to do is open ET6, click QuickBoost level 3, and that’s it. The result was a boost from our 2600K’s stock speed (Turbo enabled) of 3500Mhz to an instant 4200Mhz with no effort whatsoever.

The system was rock stable while using QuickBoost feature and we were able to run stress tests without any issue. It would have been nice to see a modest memory overclock like we have seen with ASUS boards but Gigabyte likely did not include this for simplicity’s sake.


Manual Overclocking

For manual overclocking we know our chip is capable of 5Ghz since we have achieved this milestone on a number of boards so we went right for that speed. We started with a CPU overclock without touching the memory we went right for 50x multi and 100.0 BCLK. From experience, we knew that this 2600k needs 1.43-1.46v (load) to maintain 5Ghz so we set LLC to level 7 and tried 1.43v. We know LLC level 7 is ideal from experience with other Gigabyte boards and higher than level 7 gives a fairly large overvolt. 1.43v was not quite enough for LinX so we gradually increased vCore to the point of stability which was 1.455v BIOS and 1.465v load measured with a digital multimeter. This is a bit more than we are used to but it didn’t serve as a cause for any concern.

​

After finding that point of CPU stability, the next step is always to maximize the memory overclock. Our G.Skill Trident memory is rated for 1067 @ 9-9-9 so that was a natural first step. With 1:8 (21.33) memory frequency divider, BCLK at 100, and vDIMM at stock voltage of 1.65v the memory was stable without even a bump in VCCIO. With a bit more voltage and VCCIO we could even tighten to 7-9-7 and increase the speed as you will see in the sub-zero section. This was a nice change since we had some issues with this kit on both the Sniper 2 and EVGA’s FTW board.

We ran the UD5 for a few weeks using the 5Ghz overclock and we are happy to report that we did not see any issues with stability.

 

[Eldonko]

Sub-Zero Overclocking Results

Sub-Zero Overclocking Results

By now we all know that the release of Sandy Bridge changed the sub-zero benchmarking game. Overclocking these processors doesn’t carry the simple equation of higher volts + cooler temperatures = higher CPU speed and BCLK is hardly overclockable at all, generally capping out at 107 or 108 at the most. Now overclocking is all done based on multiplier changes and CPUs have a Mhz cap and no matter how much voltage you use and how cold your temps, it cannot be surpassed. Even with a Mhz cap, sub-zero is still useful when you find a decent chip because you can keep the chip cool enough to do prolonged benchmark runs like those required for 3D benching.

​

The phase change cooler used is these tests a large single stage unit with a 10,500 btu rotary compressor, a mix of r410a and r22a gasses, and a 5 foot flex line. The cooler is tuned for a 300W heat load at -30C so it can handle Sandy Bridge with ease. Temps were low -50s idle and around -42C loaded at 1.60v. The 2600K we tested had no cold bug and it reacted normally to the sub-zero temperatures.

​

As you can see above left the UD5's temperature readings don't not go below -10C regardless of temperature so a digital thermometer is handy.

The Z68XP-UD5 worked great during sub-zero operation and exhibited no cold bugs, boot issues or anything else out of the ordinary. All-in-all, the board was very easy to work with.

In the end we were able to match our test CPU’s max of 5756Mhz; a speed achieved with a few other motherboards in the past. Also note that our memory is running at 2233Mhz at timings of 7-9-7 so that alone is a nice accomplishment for any board. We were lucky enough to get a 2600k with a Mhz cap that approaches the 5800Mhz mark and the UD5 seems to have the capability to reach the limits of our CPU with minimal effort.

Click to enlarge
​

Above you will find a few of the 2D benchmark runs we were able to achieve. With 2D benching, memory clocking really helps and since we could tighten timings to 7-9-7 and still use the 1:8 divider plus push BCLK we have to conclude that some impressive runs are possible with the UD5.

Click to enlarge​

For 3D benching we ran dual GTX 470s in SLI with stock cooling on the GPUs so improvements are possible with better cooling. In the end we were able to bench 3D and CPU intensive benchmarks at around 5700Mhz and shorter benchmarks such as PIFAST and 1M up to about 5750Mhz.

 

[Eldonko]

System Benchmarks

SuperPI Benchmark

SuperPi calculates the number of digits of PI in a pure 2D benchmark. For the purposes of this review, calculation to 32 million places will be used. RAM speed, RAM timings, CPU speed, L2 cache, and Operating System tweaks all effect the speed of the calculation, and this has been one of the most popular benchmarks among enthusiasts for several years.

SuperPi was originally written by Yasumasa Kanada in 1995 and was updated later by snq to support millisecond timing, cheat protection and checksum. The version used in these benchmarks, 1.5 is the official version supported by hwbot.

Results: A 9% increase in SuperPI 32M is noted going from stock speed of 3501Mhz (Turbo is on by default) to the QuickBoost speed of 4212Mhz on the i7 2600k and Z68XP-UD5. Jumping up to the manual overclock speed of 5002Mhz a 26% gain is noted.


CINEBENCH R11.5

CINEBENCH is a real-world cross platform test suite that evaluates your computer's performance capabilities. CINEBENCH is based on MAXON's award-winning animation software CINEMA 4D, which is used extensively by studios and production houses worldwide for 3D content creation.

In this system benchmark section we will use the x64 Main Processor Performance (CPU) test scenario. The Main Processor Performance (CPU) test scenario uses all of the system's processing power to render a photorealistic 3D scene (from the viral "No Keyframes" animation by AixSponza). This scene makes use of various algorithms to stress all available processor cores. The test scene contains approximately 2,000 objects which in turn contain more than 300,000 polygons in total, and uses sharp and blurred reflections, area lights, shadows, procedural shaders, antialiasing, and much more. The result is displayed in points (pts). The higher the number, the faster your processor.

Results: The CINEBENCH R11.5 results also show an impressive increase in performance in rendering moving from a stock system to two levels of an overclocked system. For CPU rendering, a 12% to 44% improvement (in points) was achieved when moving to QuickBoost and manual OC speeds.


Sandra Processor Arithmetic and Processor Multi-Media Benchmarks

SiSoftware Sandra (the System ANalyser, Diagnostic and Reporting Assistant) is an information & diagnostic utility. The software suite provides most of the information (including undocumented) users like to know about hardware, software, and other devices whether hardware or software. The name “Sandra” is a (girl) name of Greek origin that means "defender", "helper of mankind".

The software version used for these tests is SiSoftware Sandra Professional Home XII.SP2c and the two benchmarks used are the Processor Multi-Media and Processor Arithmetic benchmarks. These three benchmarks were chosen as they provide a good indication of three varying types of system performance. The multi-media test shows how the processor handles multi-media instructions and data and the arithmetic test shows how the processor handles arithmetic and floating point instructions. These two tests illustrate two important areas of a computer’s speed and provide a wide scope of results.

Results: Sandra processor arithmetic and multi-media show very impressive improvements on an overclocked system, with 43% gains in performance across the board in arithmetic and multi-media for the manual overclock and 17-20% gains for the QuickBoost OC.


MaxxMem Benchmark

Created by MaxxPI², the MaxxMem benchmark tests your computer’s raw memory performance, combining copy, read, write and latency tests into one global score. This memory benchmark is a classic way to measure bandwidth of a memory subsystem.

MaxxMem uses continuous memoryblocks, sized in power of 2 from 16MB up to 512MB, starting either writing to or reading from it. To enable high-precision memory performance measurement, they both internally work with multiple passes and averages calculations per run.

Further, the main goal was to minimize (CPU) cache pollution on memory reads and to eliminate it (almost completely) on memory writes. Additionally, MaxxMem operates with an aggressive data prefetching algorithm. This all will deliver an excellent judge of bandwidth while reading and writing.

Results: Moving from stock speed to the QuickBoost on the UD5, we see modest gains of 3-8% in memory read, write, and copy. When we move memory speed up to 1066Mhz (1T) and CPU speed to 5002Mhz we see more substantial gains (up to 50%) in memory read, write, and copy!

The results are similar when looking at latency, only a 3% gain is noted at QuickBoost settings but when a manual overclock is implemented along with a memory overclock the gains jump up to 31%. These findings show that CPU power can do a bit for memory latency and bandwidth but optimizing your memory will help a lot more.

 

[Eldonko]

Load-Line Calibration Testing

Load-Line Calibration Testing

Load-line Calibration is defined by Intel’s VRM spec and affects CPU voltage. In short, the CPU’s working voltage will decrease proportionally to load so higher load-line calibration nets higher voltage, increased stability and better overclocking performance. However, in some cases LLC has been known to apply too much voltage in order to compensate for these decreases while other boards have been known to apply too little voltage and thus, stability remains a bridge too far.

The Z68XP-UD5 has 10 levels load-line calibration, which is quite a bit more than most other boards.

In this section we will test each level of LLC at three different voltages (1.3v, 1.4v, 1.5v) so we can give users an idea of what to expect from each level. Both actual readings (measured via a multimeter) and BIOS reported voltage is displayed below. Optimally, we would like to see both the BIOS and DMM report the same voltage to ensure user inputs result in accurate outputs.

​

In this case we measured core voltages in the areas indicated above.

To get a good idea of how the UD5 will react to different levels of voltage we tested all 10 levels of LLC at 1.3v, 1.4v, and 1.5v and measured idle and load voltages with a digital multimeter accurate to three decimal places. All tests were run with the CPU speed at 4Ghz and memory at 1066 9-9-9.

At 1.3v the idle voltage remains at 1.303v for all 10 levels of LLC and the load voltage ranges from 1.237 to 1.350v, a difference of 0.113v. The closest level of LLC to the BIOS set voltage at 1.3v is LLC level 7 which gives 1.291v under load.

Increasing voltage to 1.4v we saw load voltage ranges from 1.318 to 1.441v, a difference of 0.123v. Again this is quite good and the closest level of LLC to the BIOS set voltage at 1.4v is LLC level 7 which gives 1.399v under load.

Moving up to 1.5v, we saw load voltage ranges from 1.408 to 1.558v, a difference of 0.150v. Consistent with the previous two results, the closest level of LLC to the BIOS set voltage at 1.5v is LLC level 7 which gives 1.507v under load.

One difference we did see at 1.5v was the idle voltage increasing starting at LLC level 4. This did not occur at 1.3v or 1.4v so it appears LLC works a bit differently as you increase the vCore.

Above is a table summarizing all LLC testing results that also includes the voltage readings in ET6. As you can see, ET6 always reads voltage a bit low (up to a 0.038v difference) so keep that in mind when overclocking. It is also important to note that using LLC above level 7 starts to overvolt by a fair amount so be careful using the higher LLC levels or use a digital multimeter so you know the final load voltage.

 

[Eldonko]

Conclusion

Conclusion

While Gigabyte may be flooding the market with Z68 motherboards, they haven’t skimped on any of their offerings and the Z68XP-UD5 is proof of this. It offers an absolutely excellent all-round computing experience while allowing users to overclock to their hearts content with complete stability past the 5GHz mark. At this point we really couldn’t ask more of a motherboard.

As seen with quite a few other Gigabyte boards, they have made it easy for inexperienced overclockers to get the most out of their systems. The QuickBoost feature in ET6 gave our 2600K an instant 700Mhz overclock without any prior overclocking knowledge being needed. More advanced users will also feel right at home and can manually overclock to their heart’s content. We achieved 5Ghz with our 2600k with minimal effort and throughout testing the UD5 was absolutely rock solid while sub-zero overclocking also went very well and allowed this board to truly stretch its legs.

Naturally, the Z68 and its expanded feature set will always edge out the P67 and although we did not provide SSD caching and Virtu tests in the review both features worked without a hitch. We also have to mention that Gigabyte didn’t make any friends around these parts with their Z68 release strategy. By rebranding P67 boards (which didn’t support Virtu GPU switching) with Z68 stickers and passing them off as new products, they lost a valuable customer base and alienated many early adopters. To add insult to injury for all those who bought into the initial Gigabyte Z68 product stack, these new “XP” boards with their support of Virtu were released only a few months later. Sure, our Virtu tests and SSD caching results can be found in the G1.Sniper 2 review and in the Z68 chipset article but in our opinion, Gigabyte missed the boat from the getgo and is currently playing catch up.

Unfortunately, trying to justify the UD5’s expense may be tough for many consumers. The $60 less expensive Z68XP-UD4 may not boast its sibling’s expanded PCI-E layout, four additional USB 3.0 ports and four extra CPU power phases but it does more SATA 6Gb/s ports and a much lower price tag. ASUS has also thrown a wrench into Gigabyte’s parade with their excellent P8Z68-V Pro which not only offers similar features but also includes a full user friendly UEFI BIOS; something that Gigabyte just can’t seem to nail down.

Looking at everything we have gone over in this review, we highly recommend the UD5 regardless of Gigabyte fumbling the ball when Z68 was first introduced. While this may not be the best value around – especially with the UD4 and ASUS P8Z68-V Pro hovering around the $200 mark – it is currently one of the better Z68 motherboards available.


Pros

- Good layout, no clearance issues
- QuickBoost worked flawlessly
- Solid BIOS, relatively bug free
- 20 phase power
- Excellent board for manual overclocking or sub-zero benching
- Memory clocking was great
- Software suite is huge

Cons

- Only 2 SATA 6GB/s ports (plus eSATA)
- No UEFI BIOS interface and Touch BIOS doesnt seem useful
- Limited amount of users will see benefits over a UD4