Gigabyte and MSI join the Core i7 motherboard crowd

Intel’s Core i7 processors have started trickling onto the market, and the collection of X58 Express-based motherboards ready to accept them has started to grow. We’ve already looked at X58 boards from Asus and Intel, predictably finding the former to be a much better fit for enthusiasts than the latter. Intel, it seems, still hasn’t figured out what makes a good enthusiast board. Gigabyte and MSI have been building them for years, though, and today their respective EX58-UD5 and X58 Eclipse motherboards take center stage.

Like most X58 boards, the UD5 and Eclipse aren’t cheap. But one could argue that that makes them more interesting. These are unabashedly high-end motherboards, and as such, they’re tricked out with excessive arrays of expansion slots, ports, and auxiliary peripherals in addition to novel features you won’t find on budget or even mid-range products. The question, of course, is which offers the best blend of indulgent excess, solid performance, and overclocking headroom. Read on for the answer.

Gigabyte’s EX58-UD5
Mo’ copper

Gigabyte has several X58 motherboards planned, including models that will sell for significantly less than the EX58-UD5’s $300 asking price. However, those cheaper derivatives aren’t expected to be available until early next year, allowing Gigabyte to capitalize on early Core i7 adopters willing to get in on the latest and greatest desktop processor at any cost. Besides, the UD5 is actually one of the more affordable X58-based motherboards currently on the market.

Long known for the virtual rainbow of expansion slots and ports that tend to populate its boards, Gigabyte has toned down the color palette a little for the UD5. The board is still draped in the company’s trademark shade of turquoisey blue, but the rainbow has been replaced by a subdued collection of white and light blue ports, with a little extra color here and there. Gigabyte says this new palette was a conscious attempt to make the board look more elegant than its predecessors, although unless you’re running a case window or one of Antec’s open-air Skeleton enclosures, you probably won’t notice the difference.

At first glance, it’s almost impossible to miss the giant Ultra Durable 3 billboard that Gigabyte has slapped onto the board’s gargantuan south bridge cooler. Ultra Durable refers to the use of solid-state Japanese capacitors, low RDS(on) MOSFETs, and ferrite core chokes—higher quality electrical components that Gigabyte says prolong the life of the board and improve its overclocking potential. This latest Ultra Durable iteration also doubles the thickness of the copper layers built into the board—a move that apparently lowers both impedance and board temperatures. Besides, nothing says bling like an extra ounce of copper in your motherboard.

It’s probably a good thing that Gigabyte so loudly advertises the attention it pays to the UD5’s electrical components, because they’d otherwise be lost in the sea of expansion slots, ports, and auxiliary peripheral chips that dominate the rest of the board’s real estate. To say that the UD5 is packed would be an understatement, and yet the board is largely free of clearance issues and niggling little layout quirks. Those with upside-down cases will want to note that Gigabyte puts the auxiliary 12V power connector along the top edge of the board, though. This location is perfect for traditional cases that put the PSU above the mobo, but with more obscure layouts, you might need a 12V extension cable to reach.

Gigabyte rings the UD5’s CPU socket with the usual combination of heatpipe-linked passive coolers. The heatsinks are relatively short, so they shouldn’t interfere with larger aftermarket coolers that fan out from the socket. Also note that one of the VRM heatsinks actually extends into the port cluster, providing a measure of convection-fueled exhaust out the rear of a system.

The EX58-UD5 features 12 power phases for the processor, two for the memory, and another two for the north bridge chip. To conserve power, the board is also capable of scaling back on the number of phases used through Gigabyte’s Dynamic Energy Saver software. There are even little LEDs on the board that light up according to how many power phases are active at a given time. Unfortunately, though, versions of Gigabyte’s DES software available for download on the company’s website and bundled with the board throw up BIOS incompatibility errors, rendering the board’s Dynamic Energy Saver functionality all but useless. If only Gigabyte built DES functionality directly into the BIOS—without the need for cumbersome Windows software—we wouldn’t have to wait for a software update to fix the issue.

Moving down the board, we find what may be the world’s largest south bridge heatsink. Seriously, this thing has more land mass than, er, Liechtenstein. And there’s more than just an ICH10R hiding beneath it. Under this mass of metal you’ll also find a “GSATA” storage controller (a JMicron JMB363 with Gigabyte’s name silk-screened on the chip) flanked by a couple of JMB322 hardware RAID chips. Each of the JMB322s is responsible for two of the white SATA ports and offers driver-free RAID 0 and 1 support. However, rather than riding the PCI Express bus, each of these chips hooks into one of the GSATA controller’s two Serial ATA ports. The hardware RAID stack in the JMB322s makes arrays connected to them appear as standard hard drives.

With ten SATA ports onboard, one might expect at least a couple to interfere with longer double-wide graphics cards. That’s not the case, though; all the ports are mounted along the edge of the board to keep cabling out of the way. I quite like this configuration, although it can cause problems for smaller enclosures that put hard drive cages or other bits of internal scaffolding right up next to the edge of the motherboard tray.

Over to the left of the SATA port cluster we can see a handy two-digit POST code display that makes troubleshooting much easier. The UD5 also has onboard power and reset buttons located up near the DIMM slots in the top-right corner of the board.

There’s certainly no shortage of expansion options on the UD5, which features a total of five PCI Express slots: three x16s, one x4, and one x1. Three-way CrossFire and SLI configurations are supported, and depending on your enclosure, you might even be able to squeeze in a double-wide threesome—the hottest of all trailer park exploits. You can even add a fourth graphics card, since the x4 slot is notched to accept longer cards. However, the position of the north bridge cooler precludes the use of longer x1 cards.

If you still have PCI peripherals kicking around, the UD5 obliges with a couple of old-school PCI slots, one of which will be blocked by a double-wide primary graphics card. Gigabyte even throws in a floppy port, although in this day and age, I think we can do without.

Before we move on, note the dual BIOS chips located in the bottom left-hand corner of the board (bottom right in the picture above). One of these chips is a backup that will come in handy if a flash attempt goes awry.

As one might expect from a board bursting with connectivity options, the UD5’s port cluster is peppered with goodies, including a clear CMOS button that overclockers should find quite useful when testing the limits of their systems. You also get plenty of USB ports, a couple of Ethernet jacks, Firewire, and a slew of audio ports, including two flavors of digital S/PDIF output. Gigabyte uses Realtek’s ALC889A audio codec, which for some might seem like an odd choice to praise. However, the ALC889A is capable of encoding Dolby Digital Live bitstreams on the fly—an old SoundStorm trick—allowing gamers to pass pristine multi-channel audio to compatible receivers and speakers.

The only thing missing from the UD5’s port cluster is external Serial ATA connectivity. Rather than putting eSATA ports on the board and arbitrarily tying them to a specific storage controller, Gigabyte provides a PCI back plate with two eSATA ports that can be connected to any of the board’s internal SATA ports. This additional flexibility is appreciated, particularly since I like the idea of running one half of a mirrored RAID 1 array in an external enclosure for easy off-site backups.

MSI’s X58 Eclipse
X-Fi comes to X58

We didn’t expect to see X58-based motherboards dip below the $300 mark, but Newegg is already selling MSI’s X58 Platinum for just $221. That’s quite reasonable for a high-end motherboard, particularly since the next-cheapest Core i7 board runs about $280. Unfortunately, we don’t have an X58 Platinum in-house. Instead, MSI sent us its flagship X58 Eclipse, which sells for closer to $350—nearly 60% more than the entry-level model.

With such a commanding price premium, one might expect a lot from the Eclipse. This is MSI’s halo Core i7 board, after all, and one whose box boldly suggests that you “Behold the future of computing.” The board itself doesn’t look all that imposing, although you could chalk that up to the subdued black and blue color scheme. MSI used to dress its boards in brilliant, fire engine red, which at least made them easy to pick out of a crowd. This latest color scheme looks like it could have just as easily come from Asus or Intel.

Of course, the aesthetic appeal of a motherboard should really only matter to the case window crowd. That segment of the enthusiast community seems to be shrinking, too, as cases designed to show off a system’s internals give way to ones focused on lowering noise levels. A motherboard’s layout is ultimately much more important than its looks, and MSI has managed to keep the Eclipse’s generous array of slots, ports, and other onboard hardware away from potential clearance conflicts.

Since it’s a pet peeve of mine, I should note that MSI puts the auxiliary 12V connector along the top edge of the motherboard, where associating cabling won’t interfere with airflow between the CPU socket and rear chassis exhaust. This position works well for traditional enclosures, although it can require an extension cable when paired with upside-down enclosures like Antec’s P180, depending on the length of your PSU’s 12V line.

Speaking of power cabling, MSI includes a GreenPower Genie pass-through connector for the primary 24-pin power connector. This pass-through also plugs directly into the motherboard through an included two-pin connector, and it works in conjunction with GreenPower Genie software to monitor and reduce system power consumption. Unlike Gigabyte’s DES app, GreenPower Genie actually works with the X58 Eclipse, and it’s capable of monitoring 3.3, 5, and 12V power consumption in addition to dynamically scaling the number of power phases used to feed the processor, memory, and north bridge chip. When configured in auto power saving mode, GreenPower Genie also lowers system voltages very slightly (usually by only a few millivolts). We’ll look at its impact on system power consumption in a moment.

While the X58 Eclipse is capable of dynamically scaling power phases, it has fewer of them to work with than the Gigabyte board. The Eclipse dedicates two power phases to its north bridge chip and another two to memory, but there are only six phases available to the processor—half as many as you get on X58 boards from Asus and Gigabyte. We’ve yet to see six-phase power solutions limit system performance or stability, even with overclocked configs, so I wouldn’t worry too much about it.

MSI may not have a fancy Ultra Durable brand that denotes the use of higher quality electrical components, but it’s worth noting that the Eclipse is littered with solid-state capacitors and ferrite core chokes. We’d expect nothing less on a $350 motherboard.

What we didn’t expect, however, was the relatively sedate heatsinks MSI uses on the board. Gone is my favorite obnoxious motherboard heatsink of all time—the infamous double loop—and in its place a collection of short VRM and chipset heatsinks that aren’t even linked with an elaborate heatpipe network. Isolating voltage circuitry cooling from the chipset isn’t a bad idea, although MSI does link its north and south bridge heatsinks with a couple of heatpipes. Either way, there’s plenty of room around the Eclipse’s socket for larger aftermarket coolers.

Turning our attention southward, note that all six of the Eclipse’s ICH10R-backed SATA ports are edge-mounted to ensure plenty of clearance for longer graphics cards. The board also comes with an additional four Serial ATA ports tied up in a similar combination of JMicron chips as on the EX58-UD5. There’s certainly an elegance to the use of driver-free hardware RAID chips, but it’s not like configuring arrays on the ICH10R is all that difficult, nor is installing RAID drivers when doing a Vista setup.

A small jumper block sits right next to the SATA ports and plays host to MSI’s D-LED 2 display module. This little OLED screen displays descriptive post progress messages like “CPU Ini” rather than cryptic hex codes, making troubleshooting easier than with a typical two-digit post code display. D-LED displays other information, too, like the voltage of the processor and north bridge chip, and a system temperature fed by a separate probe with a 20″ reach.

This apparent second coming of the D-LED certainly has potential, but I can’t help but think it’s been squandered here. The device would be much more useful if it were capable of cycling through a more extensive suite of hardware monitoring variables, such as processor and chipset temperatures, fan speeds, and other voltages. I’d also like to see the D-LED linked to the motherboard by a longer cable to allow the display to be run outside of a case.

There are no surprises in the Eclipse’s slot stack, which features three PCI Express x16 slots, two x1s, and a couple of standard PCI slots. CrossFire and SLI are both supported, and the slots are spaced such that you can run three double-wide cards without issue.

Below the slot stack you’ll find onboard buttons for power, reset, and the D-LED display. MSI even provides a little love for old-school overclockers with a set of “Easy OC” dip switches that control the board’s base clock. Of course, the switches only provide options for 133, 166, and 200MHz; you’re much better off overclocking through the motherboard’s BIOS, which offers base clock steps in 1MHz increments.

Like the EX58-UD5, the X58 Eclipse features a handy CMOS reset button right in the port cluster. Eight USB ports are also provided alongside Firewire and Ethernet jacks, and a couple of eSATA ports backed by a JMicron JMB362 SATA controller. There’s more eSATA love, too, thanks to a PCI back plate that allows an additional two external Serial ATA ports to be connected to any of the board’s internal SATA ports.

As you’ve no doubt noticed, the Eclipse’s port cluster is entirely devoid of audio jacks. Rather than putting an audio chip on the board, MSI has gone with an external riser card based on Creative’s X-Fi Xtreme Audio.

This riser sits in a PCIe x1 slot and provides a collection of analog audio jacks in addition to a digital S/PDIF output. Don’t get too excited, though; this isn’t a real X-Fi with hardware audio processing capabilities. The card features a CA0110-IBG audio chip and performs all of its EAX 5.0 magic in software. Drivers shipped with the card also lack much of the functionality present in drivers for full-fledged X-Fi cards, although you do have access to Creative’s 24-bit Crystalizer and speaker virtualization features. What really hurts the Xtreme Audio is its lack of support for real-time Dolby Digital Live or DTS encoding. This essentially limits multi-channel digital audio output to source material with pre-encoded audio tracks, which games lack.

BIOS options
Core i7 overclocking is a brave new world, with all sorts of extra multipliers and voltages with which to fiddle. Even this early in the game, both Gigabyte and MSI have their respective BIOSes relatively up to speed. There are a few key differences worth noting, however.

Unless you feel like dropping a grand on an Extreme Edition with an unlocked upper multiplier, Core i7 overclocking comes down to how high you can push a system’s base clock. Gigabyte and MSI serve up plenty of base clock options, with the Eclipse topping out at 400MHz and the UD5 pushing the ceiling to a ridiculous 1200MHz, both in fine-grained 1MHz increments.

While increasing the base clock will pump up the processor’s core speed, it also increases the speed of the chip’s uncore component and its QuickPath Interconnect, both of which we’d like to keep running as close to their stock speeds as possible. Only the Gigabyte board provides explicit control over the uncore multiplier, although to be fair, this control is somewhat limited. If you want to adjust the board’s memory bus multiplier, you have to use an auto uncore multiplier, otherwise the memory bus setting doesn’t take. Fortunately, the auto setting works well enough. The MSI board seems to do a good job of managing the uncore multiplier on its own, too, even if doesn’t expose this functionality to end users.

On the QuickPath front, the UD5 and Eclipse both offer QPI link speed options. However, only the former allows users to crank the link speed beyond the processor’s default value, which for the Core i7-920 is a 36X QPI multiplier that yields a 4.8GT/s link speed. Overclockers will be more interested in slowing the QuickPath link, and both BIOSes oblige with a “slow-mode” option that drops the QPI multiplier to 24X.

Memory multipliers bring us to an interesting little roadblock that Intel has built into its 920 and 940 processors. While the 965 Extreme has no problem running its memory faster than 1066MHz, provided proper multipliers are available in the BIOS, the 920 and 940 are effectively locked to a memory bus speed of 800 or 1066MHz. The UD5 doesn’t even offer higher multiplier options with a 920 installed, and while the Eclipse does, we couldn’t get them to work with our 920 processor and OCZ DDR3-1600 DIMMs at either 1333 or 1600MHz.

While neither board will allow users to circumvent Intel’s hard limit on Core i7 memory speeds, both offer more memory timing controls than even most seasoned enthusiasts will know what to do with. Gigabyte pushes the envelope further than most here, providing individual timing control for each of the UD5’s three memory channels.

Despite the fact that Intel quite explicitly states that running memory bus voltages higher than 1.65V risks serious damage to Core i7 processors, both the EX58-UDT and X58 Eclipse offer memory voltages well in excess of 2V. There are plenty of CPU voltage options available, too, alongside a long list of chipset and other voltages to tweak. Gigabyte serves up a few more voltages than MSI here, but both boards have more than enough overvolting options for even extreme overclocking, with plenty of granularity, too.

The presence of ample overclocking and memory tweaking options is pretty much a given with high-end enthusiast boards. However, mobos with robust fan speed controls are still few and far between, which is particularly surprising considering how obsessed many enthusiasts have become with having their rigs running as quietly as possible. Automatic fan speed control on the EX58-UD5 is limited to a toggle switch for the processor fan that supports three- and four-pin fan configs. MSI does much better on this front, allowing users to set a target processor temperature and minimum processor fan speed. The X58 Eclipse’s BIOS also provides static fan speed controls for three system fan headers, with options to run those fans at 50, 75, or 100% of their default speed.

Specifics on specifications
We’ve consolidated all the key specifications of the EX58-UD5 and X58 Eclipse below.

Our testing methods
Today we’ll be pitting the EX58-UD5 and X58 Eclipse against not only each other, but also Asus’ P6T Deluxe and Intel’s DX58SO. Note that we’re now using a Core i7-920 for motherboard testing, which is why scores for the P6T and DX58SO on the following pages don’t match the results from our initial look at those boards.

All tests were run three times, and their results were averaged.

All of our test systems were powered by OCZ GameXStream 700W power supply units. Thanks to OCZ for providing these units for our use in testing.

Thanks to OCZ for providing the DDR3-1600 DIMMs we used for testing.

Finally, we’d like to thank Western Digital for sending Raptor WD1500ADFD hard drives for our test rigs.

We used the following versions of our test applications:

• SiSoft Sandra Standard 2009 15.42

• WorldBench 6.0 Beta 2
• HD Tach 3.03
• Call of Duty 4 1.4
• RightMark Audio Analyzer 6.06
• NTttcp
• CASE Lab Euler3d CFD benchmark multithreaded edition
• Crysis 1.21
• Enemy Territory: Quake Wars 1.4
• Half-Life 2: Episode Two
• CPU-Z 1.41

The test systems’ Windows desktop was set at 1280×1024 in 32-bit color at an 85Hz screen refresh rate. Vertical refresh sync (vsync) was disabled for all tests.

All the tests and methods we employed are publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.

Memory performance
With the Core i7’s memory controller integrated right into the processor, one might not expect much difference in memory subsystem performance between motherboards. However, as we’ve seen with Athlon 64, X2, and Phenom platforms, motherboard makers do have some freedom to tune on-die memory controllers, and that can have an impact on memory performance.

Gigabyte apparently has some secret sauce working in the EX58-UD5, which manages to score nearly 2GB/s higher than the rest of the field in Sandra’s memory bandwidth tests despite running its memory at the exact same clock speed and timings as the other boards. The UD5’s advantage is short-lived, however. In our memory access latency tests, all the boards are clumped within less than a nanosecond of each other.

STARS Euler3d computational fluid dynamics
Few folks run fluid dynamics simulations on their desktops, but we’ve found this multi-threaded test to be particularly demanding of memory subsystems, making it a good link between our memory and application performance tests.

Our Euler3d fluid dynamics benchmark is largely bound by memory performance, so it’s no surprise to see the Gigabyte board leading the field. However, it’s only a hair quicker than Asus’ P6T Deluxe. MSI’s X58 Eclipse has to settle for third place, just ahead of the DX58SO.

WorldBench
WorldBench uses scripting to step through a series of tasks in common Windows applications. It then produces an overall score. WorldBench also spits out individual results for its component application tests, allowing us to compare performance in each. We’ll look at the overall score, and then we’ll show individual application results alongside the results from some of our own application tests.

Only three points separates the fastest Core i7 board from the slowest in WorldBench. The MSI and Gigabyte boards are only one point apart, wedged between offerings from Asus and Intel.

Gaming

Don’t read too much into these results. Although Crysis, Quake Wars, and Episode Two spread the field a little, we’re running a relatively low display resolution and only modest in-game detail settings. With the resolution and eye candy cranked up to levels gamers use in the real world, all the boards offer roughly equivalent performance.

Power consumption
We measured system power consumption, sans monitor and speakers, at the wall outlet using a Watts Up Pro power meter. Power consumption was measured at idle and under a load consisting of a multi-threaded Cinebench 10 render running in parallel with the “rthdribl” high dynamic range lighting demo. Results that fall under “No power management” were obtained with Windows Vista running in high-performance mode, while those with power management enabled were taken with Vista in its balanced performance mode.

We ran the X58 Eclipse with and without its GreenPower energy saving scheme enabled, and found that it saves about eight watts overall. Even without GreenPower, the Eclipse still consumes less power than the other boards when under load. The EX58-UD5, on the other hand, has the highest power consumption of the lot. Gigabyte’s Dynamic Energy Saver software should be able to lower the UD5’s power consumption a little (we’ve seen up to an 8W drop in power consumption on other Gigabyte boards), but since DES software isn’t yet compatible with the UD5, we were unable to measure its impact.

Overclocking
Enthusiasts hungry for Core i7 processors are likely to select the 920, which is by far the most affordable of the bunch. Since this chip lacks an unlocked upper multiplier and is hampered by an arbitrary memory speed ceiling (at least until motherboard makers find a way around that particular limitation), we’ve focused our overclocking attention on pushing the base clock of each board. To take the processor out of the equation, we dropped the CPU multiplier to 12X before cranking on the base clock, testing for stability with an eight-way Prime95 load along the way.

Although we initially set a static 6X memory multiplier to drop the Eclipse’s memory bus speed, we found that leaving this setting at “auto” actually produced better results. In fact, with the board set to manage the memory bus speed on its own, the Eclipse effortlessly sailed up to a 180MHz base clock. 190MHz wouldn’t post until we set the QPI link to “slow-mode,” but it was perfectly stable once we did. However, even with slow-mode enabled, the Eclipse couldn’t avoid BSODs under load with a 200MHz base clock. Playing with chipset, QPI, and other voltages didn’t help, either. Since a 190MHz base clock will take a Core i7-920 up to 3.8GHz, we’re not inclined to complain too loudly.

In order to get the UD5 to run a 6X memory multiplier, we were forced to leave the uncore multiplier set to “auto.” This apparently wasn’t a problem for the board, which like the MSI, cruised up to 180MHz without making a fuss. 190MHz was a little more problematic, with the UD5 posting but crashing under load. Fortunately, switching the QPI link to “slow-mode” quickly remedied the problem, allowing us to hit a stable base clock of 200MHz. The UD5 posted at 210MHz, too, but quickly crashed under load, no matter what we did with system voltages.

Motherboard peripheral performance
Some of the biggest performance differences you’ll find between motherboards come on the peripheral front, where it’s easy to spot where mobo makers have skimped on auxiliary peripheral chips.

Poor Gigabit Ethernet performance seems to finally be a thing of the past, at least on high-end Core i7 boards. The Eclipse and UD5 are evenly matched here, with the former consuming a few more CPU cycles than the latter. Both use similar Realtek networking chips, but the Gigabyte board features newer RTL8111D chips to the MSI’s RTL8111Cs.

While we’ve managed to avoid shoddy networking implementations this time around, some Firewire solutions are clearly superior to others. The MSI board’s Via Firewire chip is faster across the board than the Texas Instruments chip used by the EX58-UD5.

USB performance is pretty much a wash, which is expected given that all four boards tap the same ICH10R south bridge chip.

The ICH10R is also responsible for close scores through much of our Serial ATA testing. However, there are differences in performance between the auxiliary SATA options offered by the MSI and Gigabyte boards. Despite the fact that both are essentially using a similar combination of chips, the Eclipse boasts higher read and write speeds. I’m not quite sure what to make of the incredibly high burst speed of the Eclipse’s JMB322-backed SATA ports, though. These results were consistent across multiple test runs, and while the high burst rate suggests the use of cache memory for the auxiliary SATA controller, there are no memory chips on the board.

The EX58-UD5 does well in our 24-bit/192kHz RightMark Audio Analyzer loopback tests. However, despite its fancy X-Fi riser card, the X58 Eclipse scores quite poorly. RMAA identifies inter-channel leakage as a major problem for the X-Fi in loopback tests that route stereo output into the card’s line input, and while that’s something we’ve been able to correct with full X-Fi implementations, the lack of a robust audio control panel on the Xtreme Audio prevented us from doing so here. Fortunately, inter-channel leakage doesn’t seem to affect standard audio playback. We listened to a few high-quality MP3s on both the Gigabyte and MSI boards, and it was difficult to discern obvious differences in playback quality between the two.