Review: Gigabyte EX38-DQ6
03.18.08 - 11:06am
I’ve got a treat for you guys today. This review will be one of my most complete reviews ever. On the table I’ve got Gigabyte’s flagship motherboard, the EX38-DQ6. Gigabyte is one of the major motherboard vendors and for Intel’s Core 2 processors they have a motherboard covering every price-point from $50 to $300. The EX38-DQ6 boasts many of the features one would expect to see on a top-tier board but lets see if it justifies it’s lofty price.
The Package and Board
The packaging on the EX38-DQ6 commands attention with it’s brightly decorated box and flashy emblems over the box face. Inside the box, the motherboard is securely stored in a tight fitting plastic tray while all the peripherals reside in a cardboard box underneath the board. Compared to many other boards in which the board simply rests in an anti-static bag inside the box, this board is already ahead of the competition before I even fired it up.
The first thing you notice with the EX38-DQ6 when you lift it out of the plastic tray is it’s weight. The board features a pure copper heat-pipe and heatsink system that cools the mosfets, northbridge, and southbridge. Unlike lots of the competition that use anodized aluminum, Gigabyte has spared nothing with this board. After looking around the board you’ll notice the two PCIe 2.0 connectors supporting CrossFire, the eight SATA ports, twin Gigabit Ethernet ports, and a slew of USB ports. Unless you need SAS connectivity, Gigabyte has you covered with this board.
Dynamic Energy Saver
DES is a new feature unique to Gigabyte. Since the EX38-DQ6 is equipped with 12-phase power it is capable of providing some of the cleanest power available to your processor. By utilizing so many phases, the power supplied to the processor is extremely stable, permitting even higher overclocks. At the same time though, these 12 phases all would typically be operating and each phase has it’s own associated efficiencies. Since converting 12vdc to 1.35vdc isn’t a 100% efficient process, some power is lost as heat. Dynamic Energy Saver strives to counter these power losses by dynamically adjusting the number of phases the board utilizes during various load situations. In my testing I saw via the DES application and the on-board DES LEDs that it typically utilized 4 phases while at idle or lightly loaded, 8 phases around 50% to 75% load, and 12 phases when nearing 100% load. Over the course of 12 hours DES reported a savings of 23 watt-hours while 1600 watt-hours were consumed. While the savings from DES were less than 2% of what was consumed, this concept of energy efficiency is a step in the right direction and I applaud Gigabyte for this.
BIOS
I unfortunately don’t have any shots of the BIOS as I managed to kill this board when switching from CPU to 3D tests. Somewhere in the process of mounting my insulated GPUs with their dry ice pots I must have scratched a trace or dislodged a component. I’m chalking this 100% to user error, while there are no visual signs of damage it wasn’t an issue with the build quality and the board failing from a manufacturing defect. Now if I did have shots of the BIOS you’d see it’s a rather nicely done-up BIOS, if such can be said of a BIOS. You’ve got control over CPU FSB, multiplier from 6x to stock, PCIe frequency, memory dividers and northbridge strap speeds, CPU/MCH and SB skew controls, and then a slew of voltage options. It is in these voltage settings though that I had my only issue with this board. Gigabyte uses +0.x volt settings rather than giving you outright voltages. I’ve forwarded a request along to see if anything can be done about this as I’ve heard a lot of people comment on this, desiring actual voltages rather than the +0.2v or +0.35v and such. If memory serves correctly you can boost vDDR up to 2.6v, vFSB up to 1.8v, vMCH up to 1.8v, and vCPU up to 2.35v. There is a feature labeled Loadline Calibration which is meant to reduce the effects of the ever famed vdroop and from what I can tell it droops your voltage about 0.03v and then fluctuates 0.02v to 0.05v depending on the current draw. Disabling loadline calibration resulted in larger vdroops from idle but in the end either method resulted in similar load voltages.
Overclocking
In theory if you are buying this board, you had better be doing some overclocking. This board really excels at overclocking and with sufficient cooling I could see this being my new 5+GHz platform. I performed all my sub-ambient cooling on my new mini-cascade so I haven’t got any 5.5+GHz runs but the mini-cascade was good till around 5GHz. Sadly the mini-cascade isn’t quite finished and it was running an un-tuned charge of carbon dioxide in the second stage, limiting my temps to -75C and my capacity to around 200w, thereby limiting my effective clocks. Make no mistake, the EX38-DQ6 and QX9650 were just begging for higher clocks but I simply couldn’t keep it cool enough.
Test Setup
For those familiar with my reviews, this setup should start to look very familiar.
- CPU: Intel QX9650 3.0GHz 12MB L2 Cache
- Mobo: Gigabyte EX38-DQ6 with backplate removed
- RAM: Buffalo Technology 2×1GB PC2-8000 Firestix, D9GKX
- PSU: OCZ 700w GameXStream
- GPU: 1x Sapphire 512MB HD3850
- OS: Windows XP Professional SP3
- Cooling: Scythe Ninja Rev B and 2x 1/4 HP R290/R744 Cascade
Max FSB and Max Frequency
Even though my QX9650 is a 45nm processor, it still runs insanely hot with 1.5v+. I managed to hit 474FSB on air with the FSB voltage maxed out, vcore set to 1.45v, and vMCH set to 1.45v. Switching to the cascade didn’t yield a single MHz FSB but this chip has a documented hardwall at 474FSB. Max frequency on air was 4000MHz 4-core stable with 1.4125v and only limited due to temps and 4800MHz 4-core stable with 1.65v on the cascade. I was able to suicide-shot up to 5200MHz but it wasn’t bench-stable as core temps would drop from -74C to -18C before even SPi 1M would complete. In my limited memory testing I was able to hit DDR2-1220 5-5-5-15 with my FIrestix at 2.2v which typically wall around 1200MHz to 1250MHz.
Performance Tests
I performed a few CPU/Memory benchmarks to help gauge how this board performs under a variety of settings. My aircooled runs were done at 4GHz with the memory at 1GHz while the cascade cooled runs were done at 4800MHz and then the maximum bench-stable for each bench ranging from 4900MHz to 5000MHz. I ran Super Pi 1M and 32M for single-threaded clocking, wPrime 32M and 1024M for multi-threaded clocking, and 3DMark 2006 CPU Test for an application CPU score.
3DMark 2006 CPU Test
Unlike wPrime, the 3DMark 2006 CPU Test is a much less harsh benchmark and at about 1 minute of running time for both tests it is very light on the processor. I ran this test at 4GHz, 4.8GHz, and 4956MHz to show how the setup scaled with clocks. This test was definitely cooling limited as temps drooped down to -35C by the end of the test.
wPrime 32M and 1024M
wPrime is one of the harshest benchmarks on the circuit right now. Unlike Super Pi, which is single-threaded, and 3DMark 2006, which doesn’t load your core at 100% for minutes on end, wPrime will literally bring your system to it’s knees. That being said, I was only able to run this at 4GHz and 4.8GHz. I got one run in at 4852MHz but I didn’t bother recording it’s results as it was a sliver higher than the 4800MHz run. Through this entire test one should note that the voltages barely fluctuated more than 0.01v, those 12 phases were working hard on keeping the power supplied to the CPU clean and stable.
Super Pi 1M and 32M
SuperPi and overclocking go hand in hand. SuperPi 1M and 32M have been the de facto high-speed single-threaded tests for years and I don’t see that changing. These runs are un-tweaked and simply are the results from a default Windows XP SP3 install. This was my first time under 10 seconds with Super Pi, for some odd reason I had never used my own hardware to bench SuperPi, so I did a little celebration for this momentous occasion.
Conclusion
The EX38-DQ6 is everything you could expect out of a flagship motherboard. While my testing was cut abruptly short due to me accidentally killing the board, the short duration that I had it on my bench was a great time and the board ran flawlessly. The Silent-Pipe motherboard cooling system worked like a charm even when the board was overclocked, the 12-phase power brought additional stability to the system at stratospheric clocks, and the Dynamic Energy Saver put a new twist on efficient computing. The only flaw I could find in this board was the BIOS voltage selection and even that was a trivial issue that didn’t affect performance one bit. The hearty PWM setup, use of high-quality capacitors, and superb layout made benching on the EX38-DQ6 a breeze with my only limitation being processor cooling. If I have a chance to revive this board or use another one definitely expect some 5.5+GHz runs out of this setup. As my aircooling runs showed, this same platform is more than capable of running a modest aircooled overclock and with two PCIe slots, this board would be great for the hardware enthuasiast and gamer.
Pros
- Excellent Power Management
- High Build Quality
- Excellent Performance at Normal and Extreme Speeds
Cons
- BIOS Voltage Selection
Nice review Chris, if you hadn’t killed it I would have loved to see some more bench runs on it ;)
Hey now it wasn’t all my fault.. sort of. It’s replacement is doing a great job though, kicking butt and taking names under the mini cascade in 3DMark. It’s a solid board, so simple to overclock. I’m still learning X38 though, there are a few quirks with memory bandwidth and straps.