The ICFX3200 is an event not just because it’s the first—and thus far only—motherboard based on the RD600, but also because it’s DFI’s only high-end LANParty board for Core 2 processors. DFI has quickly established itself as a favorite among enthusiasts, and many have been waiting with great anticipation to see what motherboard whiz Oskar Wu can cook up for a Core 2-compatible LANParty.
Has DFI’s combination of seemingly contradictory ingredients—an AMD chipset with an Intel socket—produced a motherboard with mouth-watering potential? Keep reading to find out.
AMD’s RD600 is easily the most interesting attribute of the ICFX3200, but the chipset itself is a little short on bells and whistles. The RD600 doesn’t even have enough PCI Express lanes to supply CrossFire configurations with a full 16 lanes of bandwidth to each card. Instead, it must resort to a dual-x8 configuration—a trick Intel’s 975X chipset has been pulling off for some time now.
As one might expect from a Core 2-compatible chipset, the RD600 supports quad-pumped front-side bus speeds up to 1066MHz and dual-channel DDR2 memory up to 800MHz. AMD is, of course, well known for the memory controllers it’s integrated into Athlon 64 processors, but the RD600’s memory controller is an ATI design.
|CPU support||LGA775-based Celeron, Pentium 4/D, Core 2 processors|
|North bridge||AMD RD600|
|South bridge||AMD SB600|
|Interconnect||PCI Express x4 (2GB/s)|
|Expansion slots||3 PCI Express x16
3 32-bit/33MHz PCI
|Memory||4 240-pin DIMM sockets
Maximum of 8GB of DDR2-533/667/800 SDRAM
|Storage I/O||Floppy disk
2 channels ATA/133
4 channels Serial ATA with RAID 0, 1, 0+1 support
4 channels Serial ATA with RAID 0, 1, 0+1, 5 support via Promise PDC40719
|Audio||8-channel HD audio via SB600 and Realtek ALC885 codec|
|Ports||1 PS/2 keyboard
1 PS/2 mouse
6 USB 2.0 with headers for 4 more
1 1394a Firewire via VIA VT6307 with headers for 1 more
1 RJ45 10/100/1000 via Marvell 88E8052
1 RJ45 10/100/1000 via Marvell 88E8053 1 analog front out
1 analog bass/center out
1 analog rear out
1 analog surround out
1 analog line in
1 analog mic in
1 coaxial digital S/PDIF output
1 coaxial digital S/PDIF input
|Bus speeds||FSB: 0-511MHz in 1MHz increments
DRAM: 133-658MHz in 3MHz increments
NB PCIe: 100-200MHz in 1MHz increments
|Bus multipliers||CPU: 6x-10x (Core 2 Duo E6600)|
|Voltages||CPU: 0.44375-1.6V in 0.00625V increments
CPU special: 100-121.25% in 0.05-0.1% increments
CPU VTT: 1.21-1.59V in 0.01-0.03V increments
DRAM: 1.5-3.01V in 0.01-0.02V increments
NB Core: 1.3-2.18V in 0.06V increments
NB PLL 1.8: 1.83-2.51V in 0.08-0.1V increments
NB PLL 1.2: 1.21-1.46V in 0.08-0.09V increments
NB PCIe: 1.22-1.6V in 0.11-0.13V increments
NB Convert: 1.81-2.39V in 0.08V increments
SB Core: 1.23-1.59V in 0.11-0.14V increments
Clockgen: 2.92, 3.4V
|Monitoring||Voltage, fan status, and temperature monitoring|
|Fan speed control||CPU, system, north bridge|
Follow the four-lane PCI Express link from the RD600 down to the chipset’s south bridge component and you’ll find the SB600, a chip we’ve seen in everything from high-end CrossFire Express 3200 chipsets for Socket AM2 to AMD’s latest 690G integrated graphics chipset for budget Micro ATX boards. The SB600 is a little light in the features department, sporting only four Serial ATA ports and lacking RAID 5. DFI makes up the difference by including an auxiliary storage controller from Promise that brings an additional four SATA ports to the table along with support for parity-powered RAID arrays. However, the Promise chip is tied to the PCI bus, so it will have to share limited bandwidth with other devices.
Fortunately, the Promise chip won’t have to share bandwidth with the board’s dual Gigabit Ethernet controllers. DFI calls upon a pair of Marvell chips to supply GigE connectivity, and both use PCI Express interfaces. In fact, the GigE chips hook directly into the RD600 north bridge, so network traffic won’t even consume chipset interconnect bandwidth.
The rest of the LANParty’s spec sheet fills out as one might expect, with VIA providing a Firewire chip and Realtek handling audio codec duties. You won’t find a more ubiquitous tag team of onboard peripherals; if only VIA had a mascot, it could do a buddy movie with the Realtek crab.
LANParty designs are known for their flash, and the ICFX3200 is no exception. The board is peppered with fluorescent yellow and orange, and the ports and slots even glow under UV light, so it makes quite a statement in a properly-lit and windowed enclosure.
Most of us will be more concerned with the board’s layout, but it doesn’t disappoint there, either. DFI gets off to a good start at the top of the board, where we find both sets of power connectors. Each connector is located on or near an edge where cabling won’t interfere with airflow around the CPU socket or rear chassis exhaust. The auxiliary 12V power connector has eight pins, although the board is still compatible with four-pin power supplies.
Taking a closer look at the CPU socket reveals few capacitors—a testament to the board’s digital power delivery circuitry. Digital PWMs can deliver cleaner, more consistent power to the CPU, and they used to only be found on high-end server and workstation products. Lately, though, we’ve seen digital PWMs pop up on a number of high-end enthusiast boards, including DFI’s own LANParty designs for Socket AM2.
The ICFX3200’s low-profile power delivery circuitry leaves loads of room for gargantuan aftermarket coolers, but unfortunately, the board’s chunky north bridge heatsink does not. The north bridge cooler is happily a passive design, yet it may interfere with wider heatsinks that fan out from the processor socket.
Moving down the board, we also see a passive cooler attached to the SB600 south bridge. This one’s short enough to avoid clearance problems, leaving plenty of room for longer double-wide graphics cards. Even the board’s IDE and south bridge SATA ports are neatly arranged along the edge of the board to prevent them from being obscured by massive cards like the GeForce 8800 GTX or whatever land mass next-gen Radeons will arrive on.
Over to the left, you’ll notice that the LANParty also comes with a set of handy onboard power and reset buttons. This is a great feature to have when troubleshooting or testing the board on an open test bench, but that’s about where the utility ends.
Another little extra that helps with troubleshooting is the ICFX3200’s two-digit POST code display, which will save you from having to decipher cryptic beep codes if the system refuses to boot. The POST code display is tucked away in a corner of the board right below the third PCI Express x16 slot.
Yes, it’s no longer enough for high-end motherboards to feature two PCI Express x16 slots; three is the new standard, ostensibly to support future physics products. The ICFX3200’s third PCIe x16 slot only gets two lanes of electrical connectivity, though, so don’t get your hopes up for a CrossFire three-way.
DFI has done a good job with the ICFX3200’s slot configuration, leaving room below the top x16 slot for double-wide coolers while filling out the rest of the stack with standard PCI slots. PCI Express may be the future, but the persistent scarcity of PCIe peripherals has many of us unwilling to give up older PCI cards.
The ICFX3200’s port cluster is about what you’d expect from DFI. You won’t find serial or parallel ports here, but you do get six USB ports, one Firewire port, and coaxial S/PDIF input and output ports. An additional Firewire and four USB ports can also be accessed via onboard headers, should you be so inclined.
That rather large gap in the LANParty’s port cluster is filled by DFI’s Karajan audio riser, which houses all of the board’s analog audio ports in addition to its ALC885 codec chip. This riser is designed to help isolate the codec chip from board-level noise, improving analog output quality in the process, but we’ve had mixed results with it in the past.
We don’t normally cover manuals and driver CDs, in part because there’s little to talk about. However, the ICFX3200’s manual and driver CD come in a delightfully Hello Kitty shade of pastel pink that just doesn’t fit with the rest of the board. Let’s hope this palette doesn’t represent a new artistic direction for DFI; if you’re going to pair pink with a high-end enthusiast board, it might as well be a hot, eye-searing magenta—not a pastel better suited to My Little Pony.
BIOS options and tweaking software
Barbie also invades the ICFX3200’s BIOS with an alarmingly pink splash screen that just about slaps you across the face when the board first boots.
At least it’s another pastel pink, so the slap is pretty weak. You can easily disable the splash screen in the BIOS, as well. In fact, that may be one of the easiest things to do. You see, the ICFX3200’s BIOS is not for the weak at heart. This is quite possibly the most hardcore set of tweaking and overclocking options we’ve ever seen available on a motherboard, and that can be either a blessing or a curse, depending on just how much patience you have for fiddling.
Things start out innocently enough, with the BIOS yielding control over the CPU multiplier and providing front-side bus speed options up to 511MHz. You can also adjust the memory bus speed between 133 and 658MHz and even tweak the speed of the north bridge PCI Express links.
Then things get complicated.
On the voltage front, you can crank the CPU up to 1.6V, and then add up to an additional 21.25% with a secondary boost, yielding a maximum CPU voltage of 1.95V. Memory voltages are also available up to 3.01V, conveniently through a single setting. The next set of voltages you might want to manipulate would be those corresponding to the chipset, and there is certainly no shortage of options to choose from. You have a total of five different north bridge voltages at your disposal, plus voltage control for the south bridge and for the board’s clock generator
Things get even more interesting when we look at memory tweaking. In addition to the usual array of memory timings, you can dig deep into the memory controller and manipulate settings that even the board’s manual suggests be left at their default values.
Curiously, though, this almost obsessive attention to detail hasn’t extended to the BIOS’s fan speed control section. Sure, you can define high and low temperature targets for three onboard fan headers, but that’s about it. For a BIOS that goes to great lengths to give overclockers far more tweaking options than they probably need, it would have been nice if DFI had spent at least some time improving user control over its fan speed options.
Like its fan speed options, DFI’s CMOS Reloaded profile management system hasn’t changed much over the years. CMOS Reloaded hardly needs an upgrade, though; its ability to save and load up to four BIOS configuration profiles has yet to be matched, especially when you consider that any profile can automatically be invoked during the boot process by holding down a predefined hotkey.
The ICFX3200’s BIOS is clearly tailored for hardcore overclockers and savvy tweakers, but if you find it a little intimidating, the board also comes with a copy of AMD’s System Manager software.
This System Manager software lets users manipulate a number of system variables from Windows, including the front-side bus, memory, and even PCIe speeds.
System Manager also yields control over system voltages, although not nearly as many as you get through the BIOS. The software also confirms that the AMD RD600 is, in fact, an ATI CrossFire Xpress 3200 for Intel.
Memory tweaking is possible with the AMD System Manager, as well. Some timings, such as CAS latency, can only be manipulated through the BIOS. There are plenty of others to play with, though.
Overall, the AMD System Manager is a pretty basic system utility. It’s better than nothing, which is what we’re used to getting with
ATI AMD chipsets. However, AMD apparently has big plans for its system management software, so they may yet match the functionality already available through Nvidia’s excellent nTune system utility.
Our testing methods
We’re comparing the performance of the LANParty UT ICFX3200 T2R/G to that of a collection of Core 2 motherboards, including Asus’ P5N-E SLI and Striker Extreme, and EVGA motherboards based on Nvidia’s nForce 680i SLI and 680i LT SLI reference designs.
Although we had no problems running a 1T command rate on the Asus and EVGA boards, the DFI wasn’t stable at 1T, so we had to back it off to a 2T command rate for testing.
All tests were run at least twice, and their results were averaged, using the following test systems.
|Processor||Core 2 Duo E6700 2.67GHz|
|System bus||1066MHz (266MHz quad-pumped)|
|Motherboard||Asus P5N-E SLI||Asus Striker Extreme||EVGA 122-CK-NF68||EVGA 122-CK-NF67||DFI LANParty UT ICFX320-T2R/G|
|North bridge||Nvidia nForce 650i SLI SPP||Nvidia nForce 680i SLI SPP||Nvidia nForce 680i SLI SPP||Nvidia nForce 680i LT SLI SPP||AMD RD600|
|South bridge||Nvidia nForce 430i SLI MCP||Nvidia nForce 680i SLI MCP||Nvidia nForce 680i SLI MCP||Nvidia nForce 680i LT SLI MCP||AMD SB600|
|Chipset drivers||ForceWare 8.26||ForceWare 9.53||ForceWare 9.53||ForceWare 9.53||Catalyst 7.2|
|Memory size||2GB (2 DIMMs)||2GB (2 DIMMs)||2GB (2 DIMMs)||2GB (2 DIMMs)||2GB (2 DIMMs)|
|Memory type||Corsair TWIN2X2048-8500C5 DDR2 SDRAM at 800MHz|
|CAS latency (CL)||4||4||4||4||4|
|RAS to CAS delay (tRCD)||4||4||4||4||4|
|RAS precharge (tRP)||4||4||4||4||4|
|Cycle time (tRAS)||12||12||12||12||12|
|Audio codec||Integrated nForce 430i/ALC883 with Realtek HD 1.57 drivers||Integrated nForce 680i/AD1988B with 126.96.36.19970 drivers||Integrated nForce 680i/ALC885 with Realtek HD 1.57 drivers||Integrated nForce 680i LT/ALC885 with Realtek HD 1.57 drivers||Integrated SB600/ALC885 with Realtek HD 1.57 drivers|
|Graphics||GeForce 7900 GTX 512MB PCI-E with ForceWare 93.71 drivers|
|Hard drive||Western Digital Caviar RE2 400GB|
|OS||Windows XP Professional|
|OS updates||Service Pack 2|
Thanks to Corsair for providing us with memory for our testing. 2GB of RAM seems to be the new standard for most folks, and Corsair hooked us up with some of its 1GB DIMMs for testing.
Also, 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.
We used the following versions of our test applications:
- SiSoft Sandra Standard XI SP1
- WorldBench 5.0
- TCD Labs HD Tach v3.01
- Futuremark 3DMark06 Build 1.02
- Splinter Cell Chaos Theory v1.05
- RightMark Audio Analyzer 5.5
- RightMark 3D Sound 2.3
- Cinebench 9.5
- Sphinx 3.3
- Quake 4 1.30 with trq4demo1 demo
- F.E.A.R. 1.08
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.
The LANParty gets off to a poor start in our memory subsystem tests, falling a little behind in Sandra and trailing by quite a bit in Cachemem. Cachemem’s latency test results are particularly alarming for the ICFX3200, which pulls up close to 30 nanoseconds short of the slowest EVGA board.
Motherboards don’t always handle four-DIMM configurations well, so we popped in an extra couple of memory modules to look for problems. We had to slow the P5N-E’s DRAM command rate to 2T to get four DIMMs running stable on that board. The others had no problem running a tighter 1T command rate with four DIMMs, with the exception of the ICFX3200, which ran at 2T throughout our testing.
Adding a couple of DIMMs doesn’t make the LANParty any more competitive with the rest of the pack. The ICFX3200’s memory access latencies aren’t just problematic with a 8192KB block and 512-byte stride—they’re slow pretty much across the board.
Fortunately, memory subsystems don’t always dictate performance in real-world applications. The LANParty does reasonably well in WorldBench, turning in a score just two points short of the fastest board we tested.
The ICFX3200 is competitive in our first round of gaming tests, too, although it stumbles a little in Quake 4. Doom-engine games tend to be sensitive to memory performance, so that result isn’t entirely surprising.
Multi-GPU gaming performance
Our first round of gaming tests was conducted with lower in-game detail levels and display resolutions, but we’ve cranked things up for a second round. These tests use high resolutions, high detail levels, and anisotropic filtering and antialiasing. We’ve tested each board with a single GeForce 7900 GTX. The Asus and EVGA boards were also tested with two 7900 GTXs running in SLI. Since the ICFX3200 doesn’t support SLI, we tested it with a Radeon X1900 XTX in single-card and CrossFire configurations.
Our purpose here is not to compare the merits of SLI versus CrossFire, or even the GeForce 7900 GTX with the Radeon X1900 XTX. Instead, we’re looking at how performance scales when a second graphics card is installed.
CrossFire doesn’t scale as well as SLI in some applications, notably F.E.A.R. and Splinter Cell. That’s likely a result of graphics architectures and drivers as much as any shortcomings inherent to the AMD RD600 chipset or the LANParty board, though.
Cinebench scores don’t vary much.
Sphinx speech recognition
However, Sphinx is particularly sensitive to memory subsystem performance, and that puts the ICFX3200 at the back of the pack.
Audio codec drivers tend to determine performance with positional audio, but the ICFX3200 somehow manages to consume slightly fewer CPU resources than other Realtek-equipped boards. The DFI board’s CPU utilization is notably higher than the Asus Striker Extreme, which uses a codec chip from Analog Devices.
We should also note that some of Realtek’s HD audio drivers fail to correctly implement EAX occlusions and obstructions, rendering some games virtually unplayable with EAX effects enabled.
We used an M-Audio Revolution 7.1 sound card for recording in RightMark’s audio quality tests. Analog output ports were used on all systems. To keep things simple, I’ve translated RightMark’s word-based quality scale to numbers. Higher scores reflect better audio quality, and the scale tops out at 6, which corresponds to an “Excellent” rating in RightMark.
Karajan rising? The ICFX3200 does rather well in RightMark Audio Analyzer, managing to lead the field in dynamic range and total harmonic distortion. DFI only loses ground in the frequency response test, and then only to the EVGA nForce 680i SLI.
ATA performance was tested with a Seagate Barracuda 7200.7 ATA/133 hard drive using HD Tach 3.01’s 8MB zone setting.
ATA performance is pretty even between these boards, but the LANParty is a tad slower in HD Tach’s burst speed test.
Serial ATA performance
Moving to Serial ATA, we tested performance with a Western Digital Raptor WD360GD SATA hard drive. Again, we used HD Tach 3.01’s 8MB zone test.
The LANParty’s Serial ATA burst speeds are also slower than those of its competitors. It’s not just the SB600, either; the onboard Promise controller has the slowest burst speeds of the lot, and it also stumbles in the write speed test.
Our USB transfer speed tests were conducted with a USB 2.0/Firewire external hard drive enclosure connected to a Seagate Barracuda 7200.7 hard drive. We tested with HD Tach 3.01’s 8MB zone setting.
Older ATI chipsets were plagued with poor USB performance, but the now AMD-branded SB600 can at least hold its own. Transfer rates aren’t as speedy as the competition, and CPU utilization isn’t quite as low, but performance is at least reasonable.
Our Firewire transfer speed tests were conducted with the same external enclosure and hard drive as our USB transfer speed tests.
VIA may have dropped off our chipset radar, but the company’s Firewire controllers are proving to be quite competitive on enthusiast boards.
We evaluated Ethernet performance using the NTttcp tool from Microsoft’s Windows DDK. The docs say this program “provides the customer with a multi-threaded, asynchronous performance benchmark for measuring achievable data transfer rate.”
We used the following command line options on the server machine:
ntttcps -m 4,0,192.168.1.25 -a
..and the same basic thing on each of our test systems acting as clients:
ntttcpr -m 4,0,192.168.1.25 -a
Our server was a Windows XP Pro system based on Asus’ P5WD2 Premium motherboard with a Pentium 4 3.4GHz Extreme Edition (800MHz front-side bus, Hyper-Threading enabled) and PCI Express-attached Gigabit Ethernet. A crossover CAT6 cable was used to connect the server to each system.
The boards were tested with jumbo frames disabled.
The LANParty proves that you don’t necessarily need fancy chipset-level Gigabit Ethernet controller with hardware TCP/IP offloads to offer blistering throughput with low CPU utilization. Thanks to a couple of Marvell GigE controllers, the ICFX3200 offers more consistently strong performance across its two networking options than boards based on Nvidia’s nForce 680i SLI.
We measured system power consumption, sans monitor and speakers, at the wall outlet using a Watts Up power meter. Power consumption was measured at idle and under a load consisting of a multi-threaded Cinebench 2003 render running in parallel with the “rthdribl” high dynamic range lighting demo.
Unfortunately, we couldn’t get the Core 2 Duo’s C1E enhanced halt state working properly on the ICFX3200. C1E was enabled in the BIOS, but the board refused to lower CPU clock speeds at idle unless we invoked SpeedStep via Windows’ minimal power management profile. We contacted DFI about the issue, but it’s yet to be resolved.
With SpeedStep disabled, the LANParty’s power consumption doesn’t look that hot. However, enabling SpeedStep drops idle power consumption by close to 15W, putting the board at the front of the pack. Power consumption under load is no contest.
For our overclocking tests, we swapped our Core 2 Duo E6700 engineering sample for a retail E6300 we’ve had up to 3.3GHz without extra voltage or cooling. To get things started, we dropped the board’s CPU multiplier to 6x, locked the memory bus at 400MHz, and started turning the screws on the front-side bus. System stability was tested with Prime95 and the rthdribl HDR lighting demo.
Getting the ICFX3200 up to 400MHz proved to be easy, but pushing it beyond required fiddling with the myriad of voltage options available in the BIOS. After literally hours of tweaking, trial and error, and combing the XtremeSystems forums, I managed to get the LANParty up to a stable front-side bus speed of 450MHz. That’s short of what we’ve seen from a number of other Core 2 motherboards, although it’s still enough to push an E6300 to 3.15GHz, which isn’t too shabby.
In the right hands, the ICFX3200’s dizzying array of voltage and timing options might yield higher front-side bus speeds than the 450MHz we were able to achieve. However, I can’t get over how much effort it took to for us to get there. This board may overclock well, but it is not, under any circumstances, easy to overclock.
I wanted to like the LANParty UT ICFX3200-T2R/G. On paper, it’s a maverick third-party candidate that bucks convention and blazes its own trail. That says a lot about DFI’s willingness to try something different, and I respect that. But that doesn’t make the ICFX3200 a good motherboard.
There are good elements to the design, of course. The layout is solid, we’re big fans of passive chipset cooling, and it’s nice to see Gigabit Ethernet done right. Even the BIOS’s fine granularity and attention to detail can be appreciated by hardcore tweakers and overclockers, though it’s probably a little much for mainstream enthusiasts.
An overly complicated BIOS isn’t what sinks the LANParty’s appeal, though. Instead, it’s the fact that with all that power at our fingertips, our board’s overclocking performance was unremarkable. What’s worse, its performance at stock speeds was lower than that of the competition, including at least one board that costs much less. The RD600’s memory controller clearly has issues, and that may be why the ICFX3200 is the only board on the market that uses this chipset.
So the ICFX3200 isn’t a mouth-watering blend of flavors from AMD and Intel, and it’s certainly not something we’d recommend given the $200 asking price. However, it does give us faith in DFI’s ability to push the envelope. Now all they need to do is build a Core 2-compatible LANParty board based on a better chipset.