Simply being a viable alternative to the nForce2 wasn’t good enough for VIA, it seems. The company has rolled out a new KT880 chipset for Socket A processors, and this one features a dual-channel memory controller. Can the new KT880 snatch the Socket A performance crown from our reigning champion, the nForce2 Ultra 400? Read on to find out.
The chipset
When compared with the KT600, the KT880’s only new feature is its DualStream64 dual-channel memory controller. DualStream64 is more than just an extra 64-bit memory channel, though. The memory controller also features improved branch prediction, an enhanced data prefetch protocol, and other tweaks to improve performance. The KT880 is designed to work with DDR400 and DDR333 memory, and like all dual-channel designs, two DIMMs must be used to achieve optimal performance.
Interestingly enough, the KT880’s new memory controller can address up to 8GB of memory. The KT880 north bridge also features an AGP 8X graphics interface and supports a 400MHz front-side bus.
The KT880 north bridge links into the chipset’s VT8237 south bridge with VIA’s V-Link interconnect. This particular incarnation of V-Link offers 533MB/sec of bandwidth between the chipset’s north and south bridge components, which should be enough for most users.
On the south bridge, the VT8237 sports eight USB 2.0 ports, a six-channel audio controller, 10/100 Fast Ethernet, two ATA/133 channels, and four Serial ATA ports with support for RAID 0, 1 and 0+1 arrays. Although the VT8237 natively supports up to four Serial ATA devices, an external PHY chip is required to access two of the south bridge’s four SATA ports. Mobo manufacturers have traditionally shied away from using external PHYs to tap the VT8237’s extra two SATA ports, so I wouldn’t hold out for a KT880 board with four SATA ports hanging off the south bridge.
Even when only two of its Serial ATA ports are tapped, the KT880 still has a leg up on NVIDIA’s current nForce2 chipsets, whose MCP south bridge chips lack any Serial ATA functionality. However, NVIDIA has pledged to bring Serial ATA support, among other features, to the nForce2 chipset with a new MCP, so VIA’s Serial ATA advantage may not last.
Our testing methods
As ever, we did our best to deliver clean benchmark numbers. Tests were run at least twice, and the results were averaged.
Our test systems were configured like so:
Processor | Athlon XP 3200+ 2.2GHz | |
Front-side bus | 400MHz (2 x 200MHz) | |
Motherboard | DFI Lanparty NFII Ultra | VIA reference |
North bridge | nForce2 Ultra 400 SPP | KT880 |
South bridge | nForce2 MCP-T | VT8237 |
Chipset drivers | NVIDIA ForceWare 3.13 | VIA Hyperion 4.51 |
Memory size | 512MB (2 DIMMs) | |
Memory type | Corsair XMS3200C2 DDR SDRAM at 400MHz | |
Hard drive |
Western Digital Raptor WD360GD Serial ATA hard drive |
|
Graphics | ATI Radeon 9800 XT with Catalyst 4.3 drivers | |
OS | Microsoft Windows XP Professional | |
OS updates | Service Pack 1, DirectX 9.0b |
Thanks to Corsair for providing us with memory for our testing. If you’re looking to tweak out your system to the max and maybe overclock it a little, Corsair’s RAM is definitely worth considering.
We ran all the systems with 2-6-3-2 memory timings. Our KT880 system ran with a 1T command rate, but no such setting was available in the NFII Ultra 400’s BIOS.
Unfortunately, the audio and networking components of our VIA reference board were not functional, so we won’t be looking at the KT880’s audio or Ethernet performance. The VT8237 isn’t a new chip, anyway. We’ve already seen its performance with VIA’s PT800, PT880, KT600, and K8T800 chipsets.
The test systems’ Windows desktops were set at 1280×1024 in 32-bit color at a 75Hz screen refresh rate. Vertical refresh sync (vsync) was disabled for all tests.
We used the following versions of our test applications:
- Cachemem 2.65MMX
- SiSoft Sandra 2004
- Compiled binary of C Linpack port from Ace’s Hardware
- Cinebench 2003
- PICCOLOR v4.0 build 451
- SPECviewperf 7.1
- ScienceMark 2.0 beta (06SEP03-A build)
- Sphinx 3.3
- LAME 3.93.1 (build from mitiok.cjb.net)
- Xmpeg 5.0.1 with DivX Video 5.05
- FutureMark 3DMark03 build 330
- Comanche 4 demo
- Quake III Arena v1.32
- Serious Sam SE v1.07
- Unreal Tournament 2003 demo
- HDTach v2.61
Our tests and methods are generally publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.
Memory performance
Although the nForce2 Ultra 400 and KT880 both feature dual-channel memory controllers, the different memory controllers won’t necessarily offer the same amount of real-world bandwidth.
In fact, the KT880 actually offers more memory bandwidth than the nForce2 in the Sandra and Cachemem bandwidth tests.
Interestingly enough, the KT880 actually comes out behind the nForce2 at the end of the Linpack test, where memory bandwidth is the limiting factor with larger matrices. The performance of the two chipsets is still quite close, though.
In Cachemem’s memory latency test, the results are far from close. Here, the KT880’s memory access latency is nearly 20% lower than the nForce2.
I’ve color-coded the various cache levels (L1 is yellow, L2 amber, and main memory is orange) to make for easier reading. The graphs are presented in rough order of overall access latency, from highest to lowest.
The KT880 manages lower access latencies than the nForce2 with all but the largest step size.
Winstone tests
Although the KT880 wins the Business Winstone test by a point, it just trails the nForce2 in the Multimedia Content Creation test.
LAME MP3 encoding
We used LAME 3.92 to encode a 101MB 16-bit, 44KHz audio file into a very high quality MP3. The exact command-line options we used were:
lame –alt-preset extreme file.wav file.mp3
DivX video encoding
The KT880 continues to barely trail the nForce2 in our MP3 encoding test, but leads in DivX encoding by a frame and a half per second.
Sphinx speech recognition
Ricky Houghton first brought us the Sphinx benchmark through his association with speech recognition efforts at Carnegie Mellon University. Sphinx is a high-quality speech recognition routine that needs the latest computer hardware to run at speeds close to real-time processing. We use two different versions, built with two different compilers, in an attempt to ensure we’re getting the best possible performance.
There are two goals with Sphinx. The first is to run it faster than real time, so real-time speech recognition is possible. The second, more ambitious goal is to run it at about 0.8 times real time, where additional CPU overhead is available for other sorts of processing, enabling Sphinx-driven real-time applications.
Sphinx shows the KT880 out ahead, but only by the slimmest of margins.
Cinebench 2003 rendering and shading
Cinebench is based on Maxon’s Cinema 4D modeling, rendering, and animation app. This revision of Cinebench measures performance in a number of ways, including 3D rendering, software shading, and OpenGL shading with and without hardware acceleration.
The results of Cinebench 2003’s software tests are all quite close, but the KT880 manages to pull out to a slight lead in the OpenGL hardware shading test.
SPECviewperf workstation graphics
SPECviewperf simulates the graphics loads generated by various professional design, modeling, and engineering applications.
The KT880 edges the nForce2 Ultra 400 in two thirds of viewperf’s workstation graphics tests. Again, scores are pretty close across the board.
ScienceMark
ScienceMark measures performance with several common scientific computing type algorithms, including single-precision and double-precision floating-point matrix math using the popular BLAS routines. First up are some physics problems.
The KT880 crunches ScienceMark’s physics problems faster than the nForce2, but the performance gap is only notable in the Molecular Dynamics test.
ScienceMark’s Blas matrix tests show the KT880 out ahead of the nForce2, too. However, the results remain close.
picCOLOR image analysis
We thank Dr. Reinert Muller with the FIBUS Institute for pointing us toward his picCOLOR benchmark. This image analysis and processing tool shows us the results of a number of simple image manipulation calculations.
The KT880 continues to eke out victories in picCOLOR’s image processing tests. Here’s a look at the results for picCOLOR’s individual tests:
With the exception of a couple of tests, the KT880 and nForce2 Ultra 400 turn in identical performances in picCOLOR.
Quake III Arena
Wolfenstein: Enemy Territory
Serious Sam SE
Unreal Tournament 2003
Comanche 4
3DMark03
The KT880 comes out ahead in all but two of our gaming tests, wiping away memories of the KT600’s comparatively poor gaming performance against the nForce2 Ultra 400.
USB 2.0 transfer rates
Our USB transfer rate tests use HD Tach 2.61 and a USB 2.0 hard drive enclosure to measure performance.
Although the nForce2’s USB write speeds are competitive with the KT880, the VIA chipset offers much faster read speeds.
Unfortunately, those faster read speeds carry a higher CPU utilization penalty. The KT880’s CPU utilization during our USB transfer rate tests is more than twice as high as the nForce2 Ultra 400, so there’s a tradeoff for faster performance.
Disk I/0 performance
Our disk controller performance tests use a Maxtor 740X-6L 7,200RPM hard drive for “parallel” ATA (PATA) and a Western Digital Raptor WD360GD 10,000RPM hard drive for Serial ATA (SATA). Because we use different drives for PATA and SATA, scores aren’t comparable between the two. PATA scores should only be compared with each other. The same goes for SATA scores.
Because the nForce2 doesn’t feature an integrated Serial ATA controller, its results were obtained using the Lanparty NFII Ultra motherboard’s bridged Serial ATA port, which hooks Serial ATA drives into the nForce2’s ATA/133 controller.
That bridging trick doesn’t work out so well in HD Tach’s read burst speed test, where the KT880’s native Serial ATA controller clearly offers better performance than the nForce2 Ultra 400.
Scores in HD Tach’s transfer rate and access time tests are much closer, though. Here, we’re more than likely limited by the speed of our Raptor WD360GD hard drive.
Unfortunately, HD Tach reveals extremely high CPU utilization for the KT880’s ATA/133 controller. The chipset’s Serial ATA controller doesn’t suffer the same fate, though; CPU utilization with SATA drives is much lower than with “parallel” ATA drives.
It took a while for VIA to bring a dual-channel memory controller to its Socket A chipset family, but the KT880 finally makes it happen. The results are impressive, too. The KT880 was slightly faster than the nForce2 Ultra 400 in almost all of our application benchmarks, making it easy to recommend the chipset to enthusiasts looking to wring as much performance as possible from the Socket A platform.
From a features standpoint, the KT880’s native Serial ATA RAID support gives the chipset an edge against the current nForce2. However, the KT880’s high CPU utilization for USB and “parallel” ATA transfers is a little disheartening, and the VT8237’s integrated audio controller is still a far cry from NVIDIA’s SoundStorm APU. NVIDIA has an updated nForce2 MCP in the works with Serial and “parallel” ATA RAID, Gigabit Ethernet, and a hardware-accelerated firewall, but there’s no telling when that chipset will arrive—or whether it will be available with the SoundStorm APU, now absent from the nForce3 line. An updated nForce2 south bridge won’t necessarily improve the chipset’s performance in our application benchmarks, where the KT880’s lower memory access latencies may still reign supreme.
At the end of the day, the KT880 is a faster and generally more capable chipset than the nForce2 Ultra 400. Socket A may be dead in the minds of those enamored with AMD64, but the KT880 looks like a great option for performance enthusiasts looking to capitalize on inexpensive processors like the Athlon XP 2500+, currently selling for only $66.