The first CrossFire board to become available in North America is DFI’s LANParty UT RDX200 CD-DR. Designed by Oskar Wu, creator of the legendary Abit BP6, the RDX200 pairs ATI’s CrossFire chipset with a myriad of extra peripherals, enough BIOS options to satiate even the most demanding overclocker, and DFI’s usual dose of LANParty flair. The RDX200 is also the only Socket 939 CrossFire motherboard that’s certified by ATI. Can its features and performance compare with the high-end nForce4 boards already available on the market? Read on to find out.
|CPU support||Socket 939-based Athlon 64 processors|
|North bridge||ATI Radeon Xpress 200 CrossFire Edition|
|South bridge||ATI SB450|
|Interconnect||PCI Express (1GB/sec)|
|Expansion slots||2 PCI Express x16
1 PCI Express x1
|Memory||4 184-pin DIMM sockets
Maximum of 4GB of DDR266/333/400 SDRAM
|Storage I/O||Floppy disk
2 channels ATA/133
4 channels Serial ATA with RAID 0, 1 support
4 channels Serial ATA with RAID 0, 1, 0+1, 5 support via Sil 3114
|Audio||8-channel HD audio via SB450 and Realtek ALC882 codec|
|Ports||1 PS/2 keyboard
1 PS/2 mouse
6 USB 2.0 with headers for 2 more
1 Firewire via VIA VT6307 with header for 1 more
1 RJ45 10/100/1000 via Marvell 88E8053
1 RJ45 10/100/1000 via Marvell 88E8001 1 analog front out
1 analog bass/center out
1 analog surround out
1 analog rear out
1 analog line in
1 analog mic in
1 coaxial digital S/PDIF output
1 coaxial digital S/PDIF input
Header for 1 serial port
|Bus speeds||HT: 200-500MHz in 1MHz increments|
|Bus dividers||HT: 1x, 2x, 3x, 4x, 5x
DRAM: 1/2, 3/5, 2/3, 7/10, 3/4, 5/6, 9/10, 1/1
|Voltages||CPU: auto, 0.825-1.55V in 0.025V increments
DDR: auto, 2.4-4.03V in 0.05V increments
HT: 1.2-1.3V in 0.1V increments
NBanalog: 1.2-1.5V in 0.1V increments
NBcore: 1.2-1.5V in 0.1V increments
SB: 1.8-1.9V in 0.1V increments
LDT: 1.2-1.5V in 0.1V increments
|Monitoring||Voltage, fan status, and temperature monitoring|
|Fan speed control||CPU, chassis, north bridge|
The RDX200 uses an all-ATI chipset, pairing the Radeon Xpress 200 CrossFire north bridge with the SB450 south bridge. On the north bridge, the Xpress 200 offers a total of 22 lanes of PCI Express. Four of those lanes are reserved for PCI-E peripherals and x1 slots, while 16 are dedicated to graphics. Like NVIDIA’s original SLI chipset, CrossFire splits 16 PCI-E lanes evenly between a pair of graphics cards, giving each card eight lanes to play with. That configuration may seem less than ideal with NVIDIA’s new nForce4 SLI X16 chipset delivering a full 16 lanes to each of a pair of graphics cards in SLI, but our testing hasn’t shown that the extra lanes do much for performance.
If you’ve been keeping up with the math, you’ll notice that we’ve only accounted for 20 of the Radeon Xpress 200 CrossFire north bridge’s 22 PCI Express lanes. That’s because the remaining two lanes are used as an interconnect to the SB450 south bridge. The interconnect consumes both of the south bridge’s PCI-E lanes, so you won’t find any extras available for other peripherals or PCI-E x1 slots. You also won’t find support for 300MB/s Serial ATA transfer rates or Native Command Queuing (NCQ) in the SB450. RAID support is limited, as well, with only RAID 0 and 1 arrays supported for Serial ATA drives.
To help make up for the SB450’s mediocre RAID support, DFI equips the RDX200 with an auxiliary Serial ATA RAID chip from Silicon Image. The Sil 3114 serves up an additional four Serial ATA ports and can handle RAID 0, 1, 0+1, and 5 arrays, but you won’t find support for 300MB/s Serial ATA transfer rates or NCQ here, either. You will find a pokey PCI interface, forcing the Sil 3114 to compete for bus bandwidth with the RDX200’s Firewire chip and its PCI-based Gigabit Ethernet controller. It’s surprising to see a PCI-based GigE solution on a $200 motherboard, and frankly, a little disappointing. Fortunately, the RDX200 also sports a PCI Express GigE controller from Marvell that shouldn’t be constrained by PCI bus bandwidth.
As one might expect, the crab makes an appearance on the RDX200, as well. The board uses Realtek’s ALC882 High Definition Audio codec, which gives it a leg up on Socket 939 nForce4 boards that are limited to plain old AC’97 sound. I wouldn’t get too excited about the HD Audio support, though. We haven’t had any luck getting DVD-Audio to play back on motherboards that use the SB450, and DVD-A is where most folks will be getting their high-definition audio fix.
Despite being based on the red team’s chipset, the RDX200 doesn’t deviate from the color palette used by DFI’s LANParty nForce4 products. The black board is peppered with bright yellow hardware, and a couple of orange DIMM slots are thrown in for good measure. DFI even includes in a set of matching yellow Serial ATA and rounded IDE cables. It’s a bold look, and the brightly-colored ports, slots, and cables glow under UV light, leading at least one TR editor to throw a couple of cold cathodes into his new LANParty-based system.
Looking at the board from above, it’s easy to see that the RDX200’s layout doesn’t play follow the leader. That’s a generally good thing, and we particularly like the fact that the board’s power connectors are clustered together in the top right-hand corner of the board. This location should better accommodate cases that mount the power supply below the motherboard, and with a couple of zip ties and careful routing, it should be easy to keep power cables from crowding the CPU socket in traditional ATX cases.
We should note that unlike most Socket 939 motherboards, the RDX200’s auxiliary 12V power connector is of the eight-pin variety. According to the board’s manual, a PSU with an eight-pin 12V connector is preferred, but not necessary.
Normally we’d find a set of DIMM slots next to the power connectors on the right side of the board, but the RDX200 bucks this tradition and mounts the DIMM slots over to the left, with the CPU socket on the right. This arrangement should give the DIMM slots access to better cooling via most enclosures’ rear exhaust fans, but moving the CPU socket away from the exhaust fan could result in higher CPU temperatures. The relocated socket won’t necessarily line up with the CPU ducting available in some cases, either.
Given the layout of DFI’s LANParty nForce4 boards, we shouldn’t be surprised to see a non-standard DIMM slot and socket configuration on the RDX200. The LANParty NF4 series had its DIMM sockets mounted parallel to the top edge of the board, with the socket below.
By now, you’ve no doubt noticed the silver heat sinks that surround the RDX200’s CPU socket. The heat sinks are mounted on the board’s voltage regulator modules, and their low-profile design shouldn’t interfere with larger CPU coolers.
Speaking of clearance, we applaud DFI for leaving extra room between the board’s first PCI Express x16 and x1 slots. This allows double-wide graphics cards to be used without compromising the x1 slot, although using a second double-wide card in a CrossFire configuration will cost you one of the board’s PCI slots.
To ensure clearance for longer PCI and PCI Express cards, the RDX200 employs a pair of low-profile chipset coolers for keeping Radeon Xpress 200 and SB450 temperatures in check. We’re not all that keen on active chipset cooling, but the swanky mag-lev fan DFI uses to cool the north bridge has proven to be quieter and more reliable than most chipset fans. A completely passive design would be preferred, though.
It’s a little surprising to see north and south bridge chipset components mounted so close together on a motherboard. The two are only about an inch apart, and the area looks even more crowded thanks to the board’s eight Serial ATA ports. However, the bottom right-hand corner of the board really is the best location for these ports, as it puts them close to the hard drive bays in most enclosures.
From this angle, we can also see the RDX200’s onboard power and reset buttons. These are handy little extras to have, especially for those of us who run open test systems. It’s hard to imagine the average user finding them useful, though.
With the exception of PS/2 mouse and keyboard ports, the RDX200’s port cluster is legacy-free. The lack of parallel and serial ports leaves plenty of room for other goodies, including six USB ports, two GigE jacks, a Firewire port, and coaxial S/PDIF input and output ports.
The board’s remaining audio ports are found on the included Karajan audio riser, which plugs directly into the board and neatly fills the gaping hole in the port cluster. The Karajan riser moves the audio codec chip and all its analog input and output ports off of the motherboard in order to isolate them from board-level noise. This arrangement should improve audio quality. We’ve seen mixed results from similar implementations in the past, including several iterations of the Karajan riser, so it will be interesting to see how this version pans out on the RDX200.
DFI’s LANParty line caters to overclocking enthusiasts, so it’s no surprise that the RDX200’s BIOS is filled with plenty of goodies.
Unfortunately, Memtest86 isn’t one of them. DFI includes a version of Memtest86 in the BIOS for its LANParty NF4 boards, but the memory stress testing app doesn’t make an appearance on the RDX200. It’s an unfortunate loss, as Memtest86 is one of the best ways to stress test overclocked memory or aggressive timings.
The RDX200’s BIOS does include a wealth of memory timing options, including the ability to set a 1T command rate. A number of memory dividers are also available, allowing users to run the memory bus at a fraction of the HyperTransport clock. However, those dividers are limited to values at or less than one, so you can’t run the memory bus faster than the HyperTransport clock.
You can run the HyperTransport clock much faster than stock, though. The RDX200 BIOS’s HT clock goes all the way up to 500MHz, with lower dividers available to help keep the board’s HyperTransport link to the CPU from getting flaky at high clock speeds. The BIOS provides a wealth of CPU multipliers, and it also provides a Cool’n’Quiet MAX FID option that lets the user set the maximum CPU multiplier that AMD’s clock-throttling technology will use. This option allows Cool’n’Quiet to coexist peacefully with HyperTransport-based overclocking&3151;a useful and noteworthy feature missing from most Athlon 64 motherboards.
As far as voltages go, the RDX200 is loaded. CPU voltages are available up to 1.55V, and the CPU VID “Special Control” feature allows users to crank the CPU voltage by a further 2.4% to 36%. Memory overvolting options are also plentiful, with the board offering DDR voltages up to a whopping 4.03V. The BIOS also has several voltage options for the board’s chipset components and HyperTransport link.
In addition to offering an impressive array of overclocking and overvolting options, the RDX200’s BIOS also gives users control over the board’s CrossFire configuration. Users can disable CrossFire from within the BIOS, so there’s no need to use a terminator card if the second PCI Express x16 slot is unoccupied.
Fan speed control has become an increasingly important BIOS feature, and the RDX200 delivers a decent complement of temperature thresholds for the board’s CPU, system, and north bridge fan headers. Users can also set a CPU shutdown temperature, a handy safety feature that’s surprisingly missing from many enthusiast-oriented motherboards.
With so many options, the RDX200’s BIOS is ripe for fine tuning, overclocking, or other meddling. Mucking around with settings doesn’t always work out, though. Fortunately, the BIOS’s CMOS Reloaded feature allows users to save up to four BIOS configurations. These configurations can be saved and loaded from within the BIOS and loaded automatically at boot time by holding a hot key.
While the RDX200 undoubtedly has one of the best-equipped BIOSes around, it’s a shame that none of its tweaking or overclocking options are available in Windows. ATI doesn’t offer an equivalent to NVIDIA’s nTune system utility, and DFI doesn’t bundle a hardware monitoring or tweaking application with the board. Seasoned enthusiasts may actually prefer to do most of their tweaking in the BIOS, but less experienced users may find the interface a little daunting. Then again, if you find a BIOS interface daunting, maybe you shouldn’t be playing around with overclocking or tweaking options in the first place.
All tests were run three times, and their results were averaged, using the following test systems.
|Processor||AMD Athlon 64 FX-53 2.4GHz|
|System bus||HyperTransport 16-bit/1GHz|
|Motherboard||DFI LANParty UT NF4 Ultra-D||Asus A8N32-SLI Deluxe||DFI LANParty UT RDS200 CF-DR|
|North bridge||NVIDIA nForce4 Ultra||NVIDIA nForce4 SPP 100||ATI Radeon Xpress 200 CrossFire|
|South bridge||NVIDIA nForce4 SLI||ATI SB450|
|Chipset drivers||ForceWare 6.70||ForceWare 6.82||CATALYST 5.11|
|Memory size||1GB (2 DIMMs)||1GB (2 DIMMs)||1GB (2 DIMMs)|
|Memory type||OCZ PC3200 EL Platinum Rev 2 DDR SDRAM at 400MHz|
|CAS latency (CL)||2||2||2|
|RAS to CAS delay (tRCD)||2||2||2|
|RAS precharge (tRP)||2||2||2|
|Cycle time (tRAS)||5||5||5|
|Hard drives||Western Digital Raptor WD360GD 37GB SATA|
|Audio driver||Realtek 3.75||Realtek 3.78||Realtek HD 1.96|
|Graphics||NVIDIA GeForce 6800 GT with ForceWare 81.85 drivers|
|OS||Microsoft Windows XP Professional|
|OS updates||Service Pack 2, DirectX 9.0c|
We used the following versions of our test applications:
- SiSoft Sandra Standard 2005 SR3
- WorldBench 5.0
- TCD Labs HD Tach v3.01
- Futuremark 3DMark05 Build 120
- DOOM 3
- Far Cry v1.3
- Unreal Tournament 2004 v3323
- RightMark Audio Analyzer 5.5
- RightMark 3D Sound 2.1
- Cinebench 2003
- Sphinx 3.3
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. Most of the 3D gaming tests used the Medium detail image quality settings, with the exception that the resolution was set to 640×480 in 32-bit color.
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 RDX200 doesn’t do much to differentiate itself in our memory subsystem tests. Any performance differences here should stem from how each motherboard tunes the Athlon 64’s on-die memory controller.
WorldBench performance is pretty close, but the RDX200 is just shy of tying the nForce4 boards.
The RDX200 settles for last place in our low-resolution gaming tests, with the exception of Far Cry. Scores are pretty close overall, though.
CrossFire gaming performance
Our first round of gaming tests were conducted with low 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, so they should be more indicative of how gamers play in the real world. We’ve tested each board with a single GeForce 6800 GT, and we’ve also benchmarked the A8N32-SLI with a pair of 6800 GTs running in SLI. We tested the RDX200 with single and dual Radeon X850 XT cards using ATI’s Catalyst 5.11 graphics drivers.
When looking at our CrossFire performance results, pay special attention to the jump in performance from single- to multi-card configurations. We’re not out to compare the Radeon X850 XT’s performance with that of the GeForce 6800 GT; we just want to see how adding a second card can improve overall performance.
The RDX200’s CrossFire performance improves on a single Radeon in three of our four gaming tests, but Unreal Tournament 2004 confirms that ATI still has some work to do on application profiles for its multi-GPU rendering technology. In 3DMark05 and Far Cry, SLI has a slight edge over CrossFire in terms of the percentage increase in performance going from one to two cards. Surprisingly, though, DOOM 3 benefits slightly more from a second Radeon X850 XT than it does from a second GeForce 6800 GT.
Cinebench scores are very close, although the RDX200 trails slightly.
Sphinx speech recognition
The RDX200 is also a little behind in Sphinx, although not by a margin we’ll be too worried about.
Although none of the boards we tested provide true hardware acceleration for 3D audio, Realtek’s HD Audio drivers do support more simultaneous voices than its AC’97 drivers. That allows the RDX200 to complete RightMark 3D Sound’s 32-buffer test, while the AC’97 nForce4 boards are stuck at 24 buffers. Still, there’s little difference between the actual CPU utilization of these audio implementations.
We used an M-Audio Revolution 7.1 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.
The RDX200 comes out ahead in a couple of RightMark Audio Analyzer’s quality tests, although it’s unclear whether the difference is the board’s Karajan audio riser, the presence of a different Realtek codec chip than is on the other boards, or some other factor.
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 across the board, but the RDX200 comes out a little behind in the burst and write speed tests.
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.
South bridge Serial ATA performance is consistent, but the RDX200’s Silicon Image SATA controller scores quite poorly in the burst and write speed tests. We’ve seen poor PCI performance from the SB450 in the past. Given the fact that the Sil 3114 rides the PCI bus, it’s likely that the ATI south bridge is to blame here.
Our USB transfer speed tests were conducted with a USB 2.0/Firewire external hard drive enclosure connected to a 7200RPM Seagate Barracuda 7200.7 hard drive. We tested with HD Tach 3.01’s 8MB zone setting.
ATI chipsets have a reputation for dismal USB performance, and the RDX200 doesn’t disappoint. Burst and sustained transfer rates are quite poor, certainly not worth the high CPU utilization. To be fair, the Asus A8N32-SLI also exhibits high CPU utilization during USB transfers, but at least its throughput measures up.
Our Firewire transfer speed tests were conducted with the same external enclosure and hard drive as our USB transfer speed tests.
Although it’s competitive in most of our Firewire tests, the RDX200’s Firewire write speeds lag considerably. The board uses a PCI-based Firewire chip, so we could be seeing the SB450’s PCI performance issues popping up again.
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 Chaintech’s Zenith 9CJS motherboard with a Pentium 4 2.4GHz (800MHz front-side bus, Hyper-Threading enabled) and CSA-attached Gigabit Ethernet. A crossover CAT6 cable was used to connect the server to each system.
The nForce4 boards were tested with the NVIDIA Firewall and Jumbo Frames disabled.
Gigabit Ethernet performance is mixed for the RDX200. On one hand, the board’s PCI Express-connected Marvell 88E803 boasts impressive throughput. On the other, it exhibits higher CPU utilization than other GigE implementations, including the nForce4’s integrated Gigabit Ethernet. The RDX200’s PCI-based Marvell 88E8001 is a total disaster; throughput is less than half that of the PCI Express GigE implementations, and CPU utilization is nearly 50%. Again, we’re seeing extremely poor performance from a PCI-based peripheral connected to the SB450 south bridge.
We should note that it took DFI’s latest BIOS, version RDXDB23, to squeeze acceptable PCI-E Gigabit Ethernet performance out of the RDX200. Older BIOS revisions exhibited higher CPU utilization and much slower throughput.
For our overclocking tests, we relaxed our memory timings to 2.5-4-4-8, lowered the CPU and HT multipliers, and started cranking on the HT clock. Things progressed smoothly all the way up to 270MHz, but at 275MHz and higher, the board would hang right after the PCI device listing stage of the boot process. We tried everything we could to broach the 270MHz barrier, including extra voltages, lower memory dividers, different DIMMs, and even disabling all of the board’s onboard PCI devices, but none would allow us to boot into Windows with a HT clock above 270MHz. That’s a somewhat disappointing result given the RDX200’s wealth of overclocking options, but 270MHz is still a healthy overclock.
To test the RDX200’s performance with an overclocked HT and memory bus, we ran the board with a 270MHz HT clock, 1:1 memory divider, and a 9x multiplier that kept our Athlon 64 FX-53 processor as close to its default 2.4GHz clock speed as possible. Testing was conducted with relaxed 2.5-4-4-8 memory timings, so the results below differ from those earlier in the review.
Overclocking yields decent performance gains in Sphinx and Unreal Tournament 2004. Thanks to the RDX200’s Cool’n’Quiet MAX FID option, this configuration is also compatible with Cool’n’Quiet clock throttling.
DFI’s LANParty UT RDX200 CD-DR maintains a lot of what we’ve liked about the LANParty line, including an original layout, plenty of BIOS tweaking potential, overclocking options that cooperate with Cool’n’Quiet, and a unique visual flair that sets it apart from the sea of enthusiast-oriented mobos on the market. The RDX200 also has the unique distinction of being the only CrossFire board currently available in North America. However, CrossFire’s cachet is admittedly muted at the moment. We’re still waiting for Radeon X1800 master cards to arrive, effectively limiting CrossFire to older Radeon X800-series GPUs.
CrossFire may be relatively new, but the problems we’ve noted with this board’s I/O performance have plagued ATI south bridges for some time. The SB450’s USB performance is much slower than that of the competition, and there appear to be some very serious issues with the chip’s PCI implementation that are hampering peripheral performance. We’re not talking about performance differences of only a couple of percent here, either.
The SB450 is also missing several key storage-related features that are available in competing chipsets. The south bridge doesn’t support 300MB/s Serial ATA transfer rates or Native Command Queuing, and we’ve found that the latter can definitely have an impact on performance during disk-intensive multitasking. RAID support is also weak, with the SB450 limited to only RAID 0 and 1 arrays. DFI tries to make up for this shortcoming with a Silicon Image SATA controller that adds support for RAID 0+1 and RAID 5, but its performance appears to be crippled by the chipset’s slow PCI implementation.
Were the RDX200 cheaper than its competition, it might be easier to forgive the board’s flaws. However, with a street price hovering around $200, the RDX200 is considerably more expensive than competing nForce4 SLI motherboards, and it’s just $10 cheaper than Asus’s nForce4 SLI X16-powered A8N32-SLI. To DFI’s credit, the RDX200’s audio implementation is superior to that of most nForce4 boards, but High Definition Audio alone can’t make up for the motherboard’s other shortcomings. With nForce4 SLI motherboards, including DFI’s own, available for much less, it’s hard to recommend the RDX200.