AMD’s 890GX integrated graphics chipset

AMD has been on a bit of a hot streak lately. No, I’m not talking about the dominating performance of its peerless Radeon HD 5870 GPU. I’m not referring to quad-core bargains like the nearly-$100 Athlon II X4 630, either. I speak of something far more exciting: integrated graphics chipsets.

Ok, so maybe exciting is a bit of a stretch. But AMD has definitely been on a run in the integrated graphics world, and it all started with a 780G chipset launched two years ago nearly to the day. With a DirectX 10-class graphics core, Blu-ray decode acceleration logic, and gen-two PCI Express, the 780G quickly became our integrated graphics chipset of choice—and our recommended platform for budget desktops and home-theater PCs.

This past summer, AMD replaced its mainstream integrated graphics chipset with the 785G. A refreshed graphics core with tweaked shader units and an updated video decode block punctuated this release, propelling Gigabyte’s implementation into TR Editor’s Choice territory.

AMD hasn’t been content to confine its integrated graphics chipsets to budget microATX motherboards, though. Some six months after introducing the original 780G, a hopped up version of the chipset dubbed the 790GX arrived astride mid-range ATX boards targeted at PC enthusiasts, gamers, and overclockers. The 790GX also brought with it a new SB750 south bridge chip with AMD’s first chipset-level RAID 5 implementation and an Advanced Clock Calibration capability that typically gave overclockers an extra few hundred MHz to play with.

Six months have passed since AMD lifted the curtain on the 785G, and right on schedule, an amped-up version is set to debut as the 890GX. Like the 790GX that came before it, the 890GX boasts higher GPU clock speeds and a penchant for full-sized ATX motherboards. It also sports new SB850 south bridge silicon with a 6Gbps Serial ATA controller of AMD’s own design, which is very exciting indeed. Naturally, we had to take a closer look.

The core-logic Swiss Army knife
As someone who has long chastised marketing departments for escalating model numbers without merit, I would be remiss not to take issue with the 890GX’s primary digit. AMD showed admirable restraint when it updated the 780G with the appropriately named 785G. The 790GX made perfect sense as a tuned-up version of the 780G, too. You’d think, then, that a hopped-up 785G would carry a 795GX model designation. But no, AMD apparently couldn’t resist and has dubbed its latest north bridge component the 890GX.

Look past the model number, and you’ll find that the 890GX shares the very same north bridge silicon as the 785G. The chip features just over 200 million transistors and is fabricated on a 55-nm process by TSMC. AMD sorts the chips it gets from the Taiwanese semiconductor firm, reserving only the best for the 890GX, while the rest live on as 785Gs.

The 890GX needs the cream of the crop because its Radeon HD 4290 integrated graphics core runs at 700MHz—200MHz faster than the Radeon HD 4200 in the 785G. Apart from the difference in clock speeds, though, the graphics cores are identical. Both share the same RV620 architecture, which serves up 40 DirectX 10.1-compliant stream processors.

Like most integrated graphics components, the Radeon HD 4290 is capable of carving out a slice of system memory for its own use. With such an arrangement, the IGP is forced to share memory bandwidth with the rest of the system. Fortunately, motherboard makers also have the option of pairing the 890GX’s integrated Radeon with “sideport” memory. Also referred to by AMD as a DDR3 performance cache, this sideport RAM is typically a single, 128MB DDR3-1333 memory chip. One such chip can be seen sitting next to the 890GX north bridge component in the picture above.

Since graphics chips are responsible for more than 3D pixel pushing these days, I should also note that the Radeon HD 4290’s Universal Video Decoder (UVD) block is fully up to date. The UVD supports dual-stream decode acceleration for high-definition MPEG2, VC-1, and H.264 video, which neatly covers all the formats used by Blu-ray movies. Video output can be piped over HDMI with an accompanying audio stream, but there are a few limitations on that front. The 890GX can’t pass TrueHD, DTS-HD, or uncompressed multi-channel LPCM audio over HDMI, putting it a step behind some integrated graphics platforms.

In addition to its graphics core, the 890GX north bridge features second-generation PCI Express logic. 16 lanes of connectivity are reserved for discrete graphics cards, and unlike the 785G, the 890GX can split those lanes evenly between a pair of x8 links for CrossFire. The 890GX has an additional six PCIe lanes reserved for expansion slots and peripherals, too.

The rest of the chipset’s connectivity is consolidated in its new SB850 south bridge component, which is connected to the 890GX via an Alink Express III interconnect that offers 4GB/s of bidirectional bandwidth—twice the bandwidth of Intel’s DMI interconnect. (The 2GB/s in the block diagram above refers to one-way speed). Alink Express looks a whole lot like PCIe, and I’d wager the interconnect is little more than four lanes of PCI Express 2.0.

The south bridge has two more PCIe lanes for talking to peripherals, giving the chipset 24 lanes in total. Unlike Intel’s P55, H55, and H57 Express Platform Controller Hubs, whose second-gen PCIe lanes signal at the 2.5GT/s rate typical of gen-one implementations, the SB850’s PCI Express lanes each boast a full 5GT/s signaling rate.

By far the most interesting element of the SB850 is its Serial ATA “3.0” controller, which supports transfer rates up to 6Gbps—roughly 600MB/s, with overhead taken into account—and all the usual RAID array configs. This is the first 6Gbps SATA controller we’ve seen make its way into a core-logic chipset, and it’s only the second implementation of the new standard currently in the wild. AMD designed the new SATA controller itself, too, which is a departure from previous south bridge chips that used third-party storage controller logic.

AMD’s older south bridge chips have a history of Serial ATA performance and compatibility issues, particularly in AHCI mode, which is necessary for features like Native Command Queuing. Developing the SB850’s SATA controller itself should give AMD more control this time around, and in a moment, we’ll see whether that paid off.

The dearth of storage solutions—including even high-end SSDs—capable of exceeding the bandwidth available with old-school 3Gbps SATA makes the SB850’s 6Gbps SATA support feel more like forward-looking insurance than a must-have feature. AMD didn’t look to the future when crafting the SB850’s USB controller, though. The controller design has changed from older SB700-series implementations and now features 14 ports instead of 12. But they’re all USB 2.0 rather than SuperSpeed USB 3.0. Given the speed of today’s external storage devices, that strikes me as a little short-sighted.

At least AMD has squeezed a Gigabit Ethernet controller into the SB850 alongside the usual HD audio interface and, surprisingly, an old-school ATA channel. The whole thing is fabricated by TSMC at 65 nm, resulting in a chip that measures about 50 x 70 mm. According to AMD, the SB850 draws just 0.85W at idle, which is a quarter-watt less than the old SB750.

Asus’ M4A89GTD PRO/USB3 motherboard
Next-gen connectivity with a side of throwback

It might seem odd for an integrated graphics chipset to anchor a mid-range motherboard targeted squarely at PC enthusiasts, but that’s what you get with the 890GX. For this market, the chipset’s embedded Radeon is best thought of as a backup display adapter or a source of additional monitor outputs rather than as a primary GPU. Even if the Radeon HD 4290 is the fastest IGP around, anyone who wants to enjoy recent games with all their eye candy turned up at reasonable resolutions will be plugging in a discrete graphics card.

Asus’ M4A89GTD PRO/USB3, then, is really more of a traditional mid-range motherboard than its video outputs might otherwise suggest. That’s a good thing, because over the years, Asus has become pretty proficient at building mid-range enthusiast boards.

The M4A89GTD certainly looks the part of something you might find in an overclocked gaming rig. Asus offsets the dark brown board with a peppering of blue and white slots and ports, and I quite like the understated styling.

Of course, aesthetics won’t make or break a motherboard. The layout can, and Asus has done a good job on that front. All of the onboard slots and ports are intelligently organized to avoid clearance conflicts. Users with upside-down enclosures that put the PSU below the motherboard may need extra-long power cables to reach the board’s auxiliary 12V power connector, but that’s only because the plug is situated next to the top edge of the board, which is our preferred location for traditional enclosures.

Like other mid-range boards, Asus’ M4A89GTD covers its north bridge and voltage regulation circuitry with ornate—but not outlandish—heatsinks. A single heatpipe links the two coolers, which are short enough to avoid clearance conflicts with most aftermarket cooler designs.

Asus uses an 8+2 power-phase design to feed the AM3 CPU socket. The board can reputedly handle processors with TDP ratings up to 140W, which should allow it to support the fastest Phenom II chips, including perhaps the upcoming Phenom II X6. There’s also a core-unlocking switch next to the DIMMs slots that will allow Athlon and Phenom X3 owners to try their luck at enabling the fourth core on their CPUs.

Those with keen eyes will note that Asus’ spin on the 890GX features solid-state capacitors throughout. Asus squeezes two-ounce copper layers into the four-layer board, as well.

Rather than employing an auxiliary storage controller to feed additional internal SATA ports, Asus makes do with the six 6Gbps Serial ATA ports offered by the SB850. However, Asus has elected to connect the board’s sole IDE port to a JMicron JMB361 storage controller rather than the south bridge. The JMicron chip is also linked to an eSATA port in the rear port cluster.

As you can see, the low-profile south bridge cooler won’t interfere with longer graphics cards. The SATA ports are neatly positioned out of the way of double-wide graphics coolers, as well, which is something you don’t always see even on high-end motherboards. The mix of edge- and surface-mounted SATA ports should ensure that users with extremely tight enclosures that snug the hard drive bay right up next to the mobo will still be able to connect a collection of drives with ease, too.

The M4A89GTD is stacked with half a dozen expansion slots, including dual PCIe x16 slots, one x4, one x1, and a couple of retro PCI slots. The x4 slot’s a nice touch, and I quite like the fact that double-wide CrossFire configs will still leave users with access to it and a standard PCI slot.

The board will automatically split 16 PCIe lanes evenly between its x16 slots when two graphics cards are installed. However, if you want a full 16 lanes of bandwidth running to the primary slot, you have to install an included switch card in the secondary slot. The card takes all of a few seconds to slide into place, which is hardly a hassle. An automatic or BIOS-level switch would have been slicker, though.

The M4A89GTD’s port cluster has all the bases covered: DVI, HDMI, eSATA, FireWire, S/PDIF, and USB 3.0. The blue USB ports offer SuperSpeed connectivity via an NEC controller. A VIA chip is tasked with FireWire, while Realtek’s new ALC892 codec chip handles audio. Interestingly, Asus employs a Realtek Gigabit Ethernet chip rather than taking advantage of the GigE MAC built into the SB850. We’ve seen mobo makers snub the built-in GigE offered by Intel’s chipsets for years, and it appears AMD is getting the same treatment.

As one might expect from an Asus board, the M4A89GTD’s BIOS is packed to the gills with overclocking and tweaking options. Multipliers, clock speeds, memory timings, and voltages can all be adjusted with ease and pushed well beyond reason. All of the voltages and most clock speeds can be keyed in directly rather than selected from a list, which makes trial-and-error tweaking a lot quicker for folks who know what they’re doing. For those who don’t, the BIOS also sports an auto-overclocking feature that does all the dirty work. Auto-overclocking schemes are relatively new in the motherboard world, and I like that Asus has implemented this one in the BIOS rather than tying it to auxiliary Windows software.

For me, though, the real star of the BIOS is the fan control section. I’ve long complained that rudimentary fan speed controls were inadequate for enthusiast-oriented motherboards, and Asus has finally taken notice. With the M4A89GTD, the user can set minimum and maximum fan speeds for the CPU and system fans. One can also control the temperatures at which those fans kick into high gear. The low-temperature limit is greyed out, but Asus tells me these fan controls are still a work in progress, so we could see it unlocked eventually. Props to Asus for putting some effort into a section of the BIOS that’s been largely ignored by mobo makers for far too long.

Gigabyte’s GA-890GPA-UD3H motherboard
Undercutting the competition

We’ve noticed an interesting trend develop in the motherboard world over the last little while. In an aggressive bid to increase its share of the North American retail motherboard market, Gigabyte has been selling its mobos for less than equivalent models from Asus. This pricing strategy is evident with the GA-890GPA-UD3H, whose suggested retail price is $15 cheaper than the Asus board despite the fact that both offer similar feature sets. Indeed, even Asus’ M4A89GTD PRO, which lacks USB 3.0 connectivity, is slated to sell for $5 more than Gigabyte’s SuperSpeed-equipped UD3H.

When all else is equal—including reputation and expected reliability—we’ll recommend a cheaper board that offers better value ten times out of ten. The question, of course, is whether all else is equal with this first batch of 890GX boards.

On the surface, that looks to be the case. The UD3H is another full-sized ATX board aimed at gamers and overclockers. Like Asus, Gigabyte has figured out that enthusiasts’ tastes have outgrown the days when a clashing, neon rainbow of colors was considered an acceptable palette. Bravo.

For the most part, the UD3H’s layout is uncluttered and free of potential problems. Again, though, users with upside-down cases will want to make sure that their PSUs have long auxiliary 12V cables.

Those keeping score in the power-phase pissing match between mobo makers will want to note that the Gigabyte board uses a 4×1 phase arrangement—half the number of phases available on Asus’ 890GX offerings. But the UD3H still supports 140W CPUs, and as we’ve observed with numerous other motherboards in the past, more CPU power phases isn’t necessarily better.

Like Asus, Gigabyte uses two-ounce copper layers that purportedly offer lower impedance than typical one-ounce layers. There are solid-state capacitors across the board, as well, and nerdy puns at no extra charge.

The stumpy heatsinks for the north bridge and voltage circuitry do a good job of staying out of the way, which should allow users to run larger CPU coolers without issue. The low-profile south bridge heatsink shouldn’t interfere with expansion cards, either.

Massive graphics cards like those in AMD’s Radeon HD 5800 series can stretch all the way across an ATX motherboard, creating all sorts of clearance problems for SATA cabling. Gigabyte neatly avoids the issue by lining up all eight of the board’s SATA ports along the board’s edge, where they’ll tuck just under longer cards and coolers. This arrangement isn’t without potential for peril, though. A hard drive cage mounted right next to the motherboard tray may not leave enough room to plug into edge-mounted SATA ports.

Gigabyte squeezes an extra PCIe x1 slot into its stack, but the close proximity of the north bridge cooler may complicate compatibility with longer expansion cards. Of course, there are still two x16 slots, two more x1 slots, and a couple of PCI slots from which to choose. Unlike the Asus, this board doesn’t require a switch card to juggle lanes between the PCIe x16 slots, either. When the secondary slot is empty, all 16 lanes are automatically routed to the primary slot. Install a graphics card into the secondary slot, and the board will split the lanes in a dual-x8 config.

Look familiar? The Gigabyte board’s port cluster nearly mirrors what we saw from Asus. The only difference that really matters is the UD3H’s lack of eSATA connectivity. One could argue the presence of those blue USB 3.0 ports makes external Serial ATA ports unnecessary for this board. However, for the overwhelming majority of users, I suspect an eSATA port would’ve been more useful than the seventh and eighth internal Serial ATA connectors, especially if it was one of those fancy new hybrid eSATA/USB ports.

Although the Asus and Gigabyte boards both use the same Realtek ALC892 codec chip, only the latter appears to have implemented support for real-time Dolby Digital Live encoding, which allows multi-channel game audio to be passed to a compatible digital receiver or speakers over a single S/PDIF cable. Without real-time encoding, only source material with pre-encoded audio tracks, such as movies, can take advantage of multi-channel digital audio output.

For the most part, the first 890GX boards from Asus and Gigabyte offer identical features. Their interfaces differ, but the two boards’ BIOSes serve up similar tweaking and overclocking functionality. Both include embedded BIOS flashing utilities and support for multiple configuration profiles, too.

So where do they differ? In the automatic overclocking department, for one. You won’t find a BIOS-based auto-overclocking utility on the UD3H. Gigabyte doesn’t even ship the board with Windows software that’ll turn up your CPU’s clock speed automatically, although it has done so in the past with other boards.

By far the biggest between the two BIOSes comes when we look at fan speed controls. Those on the UD3H look positively prehistoric. The user has the option of turning automatic fan speed control on or off for the CPU and system fan headers, and one can toggle whether the CPU fan is a three- or four-pin model. That’s it. We’ve been asking Gigabyte for control over temperature thresholds and actual fan speeds or voltages for years now, and nothing has changed. Apparently, the ability to tweak obscure system voltages by hundredths of a volt is more important than meaningful fan speed controls.

The devil’s in the details
This wouldn’t be TR motherboard coverage without a painstakingly detailed assessment of each board’s BIOS options and specifications. These details don’t exactly lend themselves to eloquent prose, but you should be able to find what you need in the tables below.

Gigabyte prefers that you adjust clock speeds via explicit multipliers, while Asus gives users control over actual clock speeds. You say tomato, I say, uh, tomato. To Asus’ credit, the M4A89GTD does have a couple of extra voltage knobs to twirl. Asus’ voltage controls also offer more granularity than Gigabyte’s, although that doesn’t strike me as a difference that has much practical import.

The multiplier options listed above are what’s presented with an Athlon II X4 635 processor. Expect support for higher multipliers if you’re using a Black Edition CPU with an unlocked upper multiplier.

Lots of similarities here. Asus and Gigabyte differ on a few auxiliary peripheral chips but little else.

Our testing methods
The 890GX is really the only mid-range integrated graphics chipset on the market. Direct competition simply doesn’t exist, but we can cobble together a competent rival using Intel’s latest Clarkdale platform. Intel puts an integrated graphics processor—the Graphics Media Accelerator HD—right next to the processor core on its Core i3 and i5 CPUs. Slap one of those into an H55 or H57 Express-based motherboard with video output ports, and you’ve got yourself an integrated graphics platform.

With four 2.9GHz cores and a price tag around $120, the Athlon II X4 635 is one of AMD’s most attractive CPUs and a perfect match for the 890GX. On the Intel side, the most appropriate competition is probably the Core i3-530, which is the same price as the X4 635 and features two 2.93GHz cores that can process four threads in parallel thanks to Hyper-Threading. We’ve paired the i3-530 with an H55 Express motherboard. The H55 chipset should offer equivalent performance to the H57, since the only major difference between the two appears to be support for multi-drive RAID arrays, which we won’t be using today.

I conducted our application, gaming, and video playback tests with only the Gigabyte 890GX board, but all the other tests were run on both the Asus and Gigabyte 890GX mobos. We used the integrated graphics processor for each platform during testing. Windows 7’s power plan was set to “balanced” for all but a subset of our power consumption tests.

With few exceptions, all tests were run at least three times, and we reported the median of the scores produced. For IOMeter, we’ve reported average rather than median scores. Also, our power consumption tests were only run once.

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:

• Stream 5.8 64-bit
• CPU-Z 1.41
• The Panorama Factory 5.3 x64
• x264 HD benchmark 3.03 with AviSynth 2.58
• 7-Zip 4.65
• TrueCrypt 6.3a
• Cinebench 11.5
• Call of Duty: Modern Warfare 2
• DiRT 2
• Borderlands
• Left 4 Dead
• CyberLink PowerDVD 9 9.0.2528.51
• TCD Labs HD Tach 3.01
• Intel IOMeter v2006.07.27
• NTttcp
• RightMark Audio Analyzer 6.2.3

The test systems’ Windows desktop was set at 1280×1024 in 32-bit color at a 60Hz 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
Memory bandwidth doesn’t always dictate real-world performance, but it’s a good place to start when testing systems whose integrated graphics processors consume a chunk of main memory. Both of our 890GX boards feature 128MB of dedicated sideport video memory, however, which should reduce bandwidth sharing.

Despite its lack of dedicated video memory, the H55 Express platform delivers higher memory bandwidth than either 890GX board. Score one for the Core i3-530. The Asus and Gigabyte 890GX boards offer nearly equivalent memory bandwidth, with the Gigabyte board having a slight edge overall.

Shift your attention to memory latency, and the Asus 890GX board edges out the Gigabyte by a nanosecond. The real story there is how much the H55 Express trails—its memory access latency is a full 30 nanoseconds slower than the 890GX boards.

The following graphs are a little indulgent, but they paint the latency picture in three dimensions, across multiple block and step sizes. I’ve arranged the graphs in order of highest latency to lowest. Yellow represents L1 cache, light orange is L2, red is L3, and dark orange is main memory.

Here we’re focused on how different processor platforms compare, so I’ve omitted scores for the 890GX board. The H55 Express and its Core i3-530 may have much higher memory access latencies than the 890GX and Athlon II X4 635 combo, but thanks to the Core i3’s L3 cache, that platform doesn’t have to hit main memory until the block size exceeds 4MB. The AMD system starts dipping into main memory once we pass 512KB, giving it higher access latencies than the Intel rig at block sizes between 512KB and 4MB.

Application performance
Core-logic chipsets don’t have a huge impact on application performance, but we’ve whipped up a diverse suite of application tests to give you a sense of how our test platforms stack up when their respective CPUs become the bottleneck. Most of these tests are effectively multi-threaded, so keep in mind that we’re testing a proper quad-core Athlon II X4 against a dual-core, four-thread Core i3.

Cinebench’s rendering test is handled by the CPU, and the 890GX platform has a sizable lead over the H55. The results of the OpenGL modeling test are even more striking. In that test, the Radeon HD 4290 easily outruns the Intel GMA HD.

The Core i3-530 is hardly a slouch when it comes to HD video encoding, but the Athlon II X4 635 is faster—by a healthy margin, too.

In another highly multi-threaded test, the 890GX system prevails once more.

TrueCrypt isn’t even close. Core i3 and i5 processors do have new instructions designed to improve encryption performance, but TrueCrypt doesn’t yet support them.

The Panorama Factory’s stitch operation is nicely multithreaded, and the Athlon II X4’s four cores are faster than the i3-530’s two.

Gaming
The 785G is certainly a competent integrated graphics processor, but it’s hardly a recommended solution for gamers looking to play the latest titles with the details turned up at reasonable resolutions. The 890GX’s Radeon HD 4290’s GPU is clocked 40% higher, so it should fare better. But is it really good enough to manage playable frame rates with the latest games? To find out, we tested a handful of recent releases at a charitably low resolution of 1024×768. We used the in-game timedemo features built into Left 4 Dead 2, DiRT 2, and Borderlands, and relied on FRAPS to log gameplay sessions in Modern Warfare 2. Median scores were taken from five runs of DiRT 2 and Modern Warfare 2, while three runs were used with the other games, whose scores were more consistent.

Lesson 1: even the latest incarnation of Intel’s Graphics Media Accelerator has serious issues. Our H55 Express system crashed repeatedly in Left 4 Dead 2 regardless of whether we were running a timedemo or just trying to play the game. DiRT 2 didn’t work, either, presenting us a blank screen instead of the game’s usual menu. Both are recent, popular titles that really should work.

With the GMA HD pulling up lame in two games, it’s easy for the Radeon HD 4290 to look good. But there is some bad news. I had to run DiRT 2 at the lowest in-game detail levels to manage playable frame rates at 1024×768. Fortunately, Left 4 Dead 2 was much more accommodating. The Source-engine title yielded decent frame rates even with all details turned up, albeit with extras like antialiasing and anisotropic filtering disabled.

Modern Warfare 2 ran acceptably on the 890GX, too, but only after I disabled antialiasing and a couple of in-game effects. Amusingly, the GMA HD’s average frame rate is the same as the Radeon HD 4290’s low point.

The only game of the bunch that I wouldn’t deem playable at 1024×768 on either platform is Borderlands. For whatever reason, this game doesn’t seem to scale well down to middling graphics hardware. Even with the lowest in-game detail levels, the Radeon HD 4290 only managed an average of 20 frames per second. This is why real gamers run discrete graphics cards—even sub-$100 models are a heck of a lot faster than either of these IGPs.

Blu-ray playback
We conducted our Blu-ray playback tests across three high-bitrate movies covering the major formats available on the market. 28 Days Later was used to represent the H.264 camp, Nature’s Journey for VC-1, and Click (which it pains me to even admit that we purchased on Blu-ray) for MPEG2. The latest version of PowerDVD, which supports the decode acceleration built into both the 890GX and GMA HD, was used for testing. Playback was run full-screen over HDMI at 1080p resolution.

Both systems played back our three movie samples flawlessly. Based on the scores above, the H55’s decode logic looks to be much more efficient. However, the Core i3-530 doesn’t lower its clock speed below 2.93GHz when playing the movies. The Athlon II X4 635 drops its CPU multiplier from 14.5X to 4X, taking the CPU clock from 2.9GHz down to just 800MHz. 20% of 800MHz is less than 11% of 2.93GHz.

Serial ATA performance — IOMeter
We’ll begin our storage tests with IOMeter, which subjects our systems to increasing multi-user loads. We used IOMeter’s workstation and database test patterns, since those are more relevant to desktop systems than the file or web server test patterns. This particular test makes good use of the Native Command Queuing capability built into the AHCI specification.

Drives capable of taking advantage of the SB850’s 6Gbps SATA controller are few and far between. The best candidate is currently Crucial’s new RealSSD C300, which is the first 6Gbps solid-state drive to hit The Benchmarking Sweatshop. Naturally, we couldn’t resist testing with it. But plenty of folks also use mechanical drives, and likely will for some time, so we’ll kick things off with a look at SATA performance using a Western Digital VelociRaptor.

AMD has a history of poor storage controller drivers, so in addition to testing the Gigabyte 890GX board with AMD’s own drivers, we tested it with the Microsoft AHCI drivers included with Windows 7.

When they’re both using AMD’s AHCI drivers, the Asus and Gigabyte 890GX boards offer nearly identical transaction rates. Performance levels off after we hit 32 concurrent I/O requests, which just happens to be the queue depth for Native Command Queuing. Interestingly, there’s no performance plateau when we combine the 890GX with Microsoft’s own AHCI drivers. The H55 Express’ transaction rates don’t trail off after 32 I/Os, either.

Regardless of which drivers are used, the Gigabyte 890GX board uses relatively more CPU time than the Asus. The H55’s CPU utilization is lower than both boards, although we’re still looking at less than 4% CPU utilization overall.

Switching over to a high-end SSD exposes a weakness with Microsoft’s AHCI drivers, at least at lower loads. The H55 Express still manages slightly higher transaction rates than any of our 890GX configurations—and that’s with a 3Gbps SATA controller.

Again, the Gigabyte 890GX board exhibits higher CPU utilization than the Asus. Both consume relatively more CPU cycles than our H55 system, but the differences don’t amount to much as the load scales upward. I didn’t expect CPU utilization to peak with the fewest outstanding I/O requests, but that’s what happened with all four configurations.

Serial ATA performance — HD Tach
We used HD Tach 3.01’s 8MB zone test to measure basic SATA throughput and latency.

The RealSSD has some of the fastest read burst speeds we’ve ever measured, and it’s clearly quicker on the 890GX than it is on the H55 Express. Even the VelociRaptor has higher burst speeds on the 890GX, but only with the Gigabyte motherboard. The Asus board’s VelociRaptor burst speeds are nearly 20MB/s slower than those of the Gigabyte.

Here’s where things start to get a little weird. With the VelociRaptor, average read speeds don’t vary all that much from one platform to the next. However, the RealSSD’s read speeds are all over the map, with the H55 Express wedged between the Asus and Gigabyte 890GX boards. The read speeds on each platform were quite consistent from one run to the next, making the differences between platforms all the more puzzling.

We haven’t had much time to work with the RealSSD, but before testing it with each configuration, I ran a secure erase and a full-disk HD Tach write speed test on the drive to establish an even, used-state playing field. This should prevent things like the block-rewrite penalty from affecting our results.

The intrigue continues when we look at average write speeds, which again find the Asus and Gigabyte boards separated by quite a margin. Asus is fastest again, but neither 890GX board is quick enough to catch the H55 Express.

More troubling than the wide gap in SSD write speeds between the Asus and Gigabyte 890GX boards is the fact that both pull up lame with the VelociRaptor. The H55 Express’s average write speeds are 30MB/s faster than the 890GX’s when the mechanical drive is installed.

HD Tach doesn’t quote a margin of error for its random access time test, but I can’t help but wonder if it’s more than a tenth of a millisecond. Things are pretty even here, with the obvious exception that the RealSSD’s access times are more than an order of magnitude shorter than the VelociRaptor’s.

There is a +/- 2% margin of error for HD Tach’s CPU utilization tests, but the differences in CPU utilization are much greater than that. With both the VelociRaptor and the RealSSD, the Gigabyte 890GX scores much higher than the Asus in HD Tach’s CPU utilization test.

USB performance
Our USB transfer speed tests were conducted with a USB 2.0/FireWire external hard drive enclosure connected to a 7,200-RPM Seagate Barracuda 7200.7 hard drive. We tested with HD Tach 3.01’s 8MB zone setting.

The SB850 may have a rearchitected USB controller, but it’s not as fast as the one inside the H55 Express. The 890GX boards pull up short in the burst and average read speed tests, and the Gigabyte trails behind the leaders with writes, too.

Matters get worse for the Gigabyte board when we look at CPU utilization, which is notably higher than the Asus. The 890GX boards are both running the “Balanced” Windows power plan with Cool’n’Quiet enabled, so there shouldn’t be this much of a difference between them.

When asked about the SB850’s slower USB transfer rates, AMD suggested that real-world transfers shouldn’t be affected. The company also indicated that Cool’n’Quiet can react oddly to CPU utilization tests, although that wouldn’t explain the difference in CPU utilization between the Asus and Gigabyte boards.

PCI Express performance
We used NTttcp to test PCI Express Ethernet throughput using a Marvell 88E8052-based PCI Express x1 Gigabit Ethernet card.

PCI performance
To test PCI performance, we used the same NTttcp test methods and a PCI Intel GigE NIC.

A Gigabit Ethernet controller may not be the most bandwidth-intensive peripheral to throw at an expansion interface, but it’s certainly the most common. All of our system configurations do well in the throughput tests, but the 890GX rigs have higher CPU utilization than the H55 Express. I suspect we’re seeing the Athlon II’s clock throttling in action again, but that doesn’t explain why the Asus board has lower CPU utilization than the Gigabyte in the PCI test.

Power consumption
That covers the chipset-specific portion of today’s festivities. Now it’s time to switch gears to exploring variables more dependent on motherboard attributes than core-logic components. First up, we have power consumption tests. We measured system power consumption, sans monitor and speakers, at the wall outlet using a Watts Up Pro power meter. Readings were taken at idle and under a load consisting of a Cinebench 11.5 render alongside the rthdribl HDR lighting demo. We tested with Windows 7’s High Performance and Balanced power plans.

Motherboard makers usually ship their boards with energy-saving software that’s supposed to lower power consumption without impeding performance. We’ve tested each board with and without this software installed. Gigabyte’s H55 Express board uses Dynamic Energy Saver software, while the company’s 890GX offering uses a new app called EasySaver. The Asus board uses an EPU app that must be configured in “auto” mode to avoid performance-sapping clock throttling.

Even with fewer power phases than its Asus counterpart, the Gigabyte 890GX board draws notably more power. Running each company’s power-saving software is good for a watt or two, but that’s about it.

As one might expect, our H55 system has the lowest idle power draw of the lot. The Core i3-530 is more power-efficient than the Athlon II X4 635, and the Intel CPU is also running on a smaller microATX motherboard.

Under load, the Intel system has even more of a power-efficiency advantage. It’s not even close.

Between the 890GX boards, the Asus draws less power under load by about 10W. Power-saving software has more of an effect here than it did at idle, particularly on the Asus board.

Overclocking
I had wanted to dip into IGP overclocking with this review, but there simply wasn’t time. And then I got to thinking and figured there wasn’t a point, either. Sure, the Asus and Gigabyte 890GX boards both give users the ability to tweak Radeon HD 4290 clock speeds, but if you’re that desperate for graphics performance, you’re better off saving your pennies, shoveling a few driveways, and buying a real graphics card, even if it’s a generation or two old.

Fortunately, I did have time to run a few quick overclocking tests on the motherboards themselves. First, I experimented with the auto-overclocking utility built into the Asus BIOS.

There wasn’t much to it: select the option in the BIOS, wait for the reboot, and see what you get. My system settled on a 233MHz base clock, which combined with a 14.5X multiplier, yielded a 3.4GHz CPU clock speed—not bad for auto-tuning.

Since the Gigabyte board lacks an in-BIOS auto-overclocking utility and doesn’t come with Windows software that accomplishes the same task, I kicked it old school with some base clock overclocking on the Asus. First, I lowered the CPU and memory multiplier to take those components out of the equation. Next, I turned up the base clock speed, checking for stability along the way using a four-core Prime95 load.

The Asus board cruised up to a 300MHz base clock speed with ease, but it would go no further. 310MHz wouldn’t post, even with extra voltage applied to the CPU and chipset. Still, it’s hard to complain about a 50% boost for the base clock.

Next up: the GA-890GPA-UD3H.

Much like the Asus board, the Gigabyte didn’t put up a fuss as I turned up the base clock—but only up to 280MHz. Try as I might, I couldn’t get the system to post at 290MHz. A 280MHz base clock speed is still capable of taking an Athlon II X4 635 up to an even 4GHz, which ain’t half bad.

Motherboard peripheral performance
Core logic chipsets integrate a wealth of peripherals, but they don’t handle everything. FireWire, Ethernet, USB 3.0, and audio are farmed out to auxiliary chips, for example. To provide a closer look at the peripheral performance you can expect from the motherboards we’ve tested today, we’ve compiled Ethernet, Serial ATA, USB 3.0, FireWire, and audio performance results below.

The 890GX boards have similar FireWire performance, which is interesting considering that they use completely different controller chips. CPU utilization is a little higher on the Gigabyte, but the scores are within that test’s +/- 2% margin of error.

USB 3.0 may be the new hotness, but it just isn’t stable on the Gigabyte 890GX board. Scores were wildly inconsistent from one run to the next, and the system even locked up a couple of times during testing. Using the exact same SuperSpeed USB hard drive, I saw much more consistent performance on the other two boards.

All three boards use the very same NEC USB 3.0 controller, yet the H55 has much higher average read and write speeds than the Asus 890GX. Clearly, some implementations are superior to others.

We’ve already covered the interesting SATA results, but these fill out some missing scores for the auxiliary storage controllers on each board. The scores above were all obtained with the VelociRaptor serving as the test drive.

The Asus and Gigabyte 890GX boards use Realtek’s RTL8111E and 8111D Gigabit Ethernet controllers, respectively. We don’t see much difference in throughput between the two, but again, the Gigabyte turns in a higher CPU utilization score.

The 890GX boards score identically in our 24-bit, 192kHz RMAA loopback test. That said, the Gigabyte H55 board scores one point higher in three of the eight component tests—and overall.