VIA’s K8T800 Pro chipset

Mated with VIA’s VT8237 south bridge chip, the K8T800 offers support for AGP 8X, Serial ATA RAID, USB 2.0, six-channel audio, and the usual litany of checklist features common to today’s core-logic chipsets. Short of an all-out conversion to PCI Express, what more could anyone want out of an Athlon 64 chipset?

The K8T800 Pro
Two things, apparently. VIA has decided to make a pair of useful modifications to its north bridge before AMD’s new Socket 939 processors arrive in force, and the K8T800 Pro is the product of those changes. First, VIA has made the AGP and PCI clocks on the K8T800 Pro run independently of the clock for the HyperTransport link. (There is no traditional front-side bus on an Athlon 64, but the north bridge talks to the CPU via HyperTransport.)

For everyday operation at stock clock speeds, this new clocking scheme doesn’t mean much. However, by making these clocks run asynchronously, VIA is delivering a big, juicy gift to overclockers everywhere, whose Athlon 64 overclocking ambitions have been frustrated by motherboards incapable of running at higher bus speeds. Now, newer versions (including Socket 939 editions) of excellent boards like the Abit KV8 MAX3 and MSI K8T Neo should offer much more flexibility for those of us looking to subject an Athlon 64 2800+ to sweet, egregious abuse.

Second, the K8T800 Pro supports HyperTransport clock speeds of 1GHz. That means the north bridge talks to the CPU even faster than before. Now, follow along with me here. The K8T800 had a pair of 16-bit, 800MHz HyperTransport links—one running in each direction—between the CPU and the north bridge, for theoretical peak bandwidth of 6.4GB/s. The K8T800 Pro has a pair of 16-bit HT links between the CPU and north bridge that can run at 1GHz clock speeds, giving it a theoretical peak bandwidth of 8GB/s.

The trick to the 1GHz HyperTransport link, by the way, is a 5X multiplier on the HyperTransport clock. The HyperTransport link will still show up in the BIOS as a 200MHz “LDT bus” or “front side bus,” but the available multipliers won’t top out at 4X any longer. The nominal clock rate will remain 200MHz. The 5X multiplier for HyperTransport isn’t yet officially supported by AMD, but our rather ancient Athlon 64 FX-51 ran with nary a complaint using a 5X multiplier on the K8T800 Pro reference board. Rumor has it AMD will give 1GHz HyperTransport its official blessing when Socket 939 processors arrive.

Trouble is, nothing much is taxing the K8T800’s 800MHz HyperTransport links as is. 6.4GB/s is more than enough bandwidth for communication between the north bridge and CPU at present, because AMD has moved the memory controller onto the Athlon 64 itself. All that’s left is I/O traffic and AGP communications. I suspect 1GHz HyperTransport will matter more when PCI Express arrives, but for now, the K8T800 Pro has this feature in case it’s needed. (Ok, so 1GHz HyperTransport could be useful now in a multiprocessor Opteron chipset, too, but I don’t believe the CPU-to-CPU links are dictated by chipset considerations.)

So that’s the skinny on the K8T800 Pro chipset. There’s nothing too earth-shattering involved, but it is an evolutionary product that will likely be at the heart of scads of Socket 939 motherboards.

What’s missing from the K8T800 Pro? Either nothing or lots, depending on your perspective. Coming soon on the chipset front are all kinds of new features, including PCI Express, the new High-Definition Audio standard, more and faster Serial ATA RAID channels. NVIDIA has also added Gigabit Ethernet and a firewall to the nForce3 250Gb, the K8T800 Pro’s main competition. This chipset has none of those things. But in high tech, I suppose every product is part of the journey, none a destination.

Some prose on Pros
This section of the review is intended to undo the work of the marketing people, specifically with respect to the “Pro” or “Professional” tag. VIA has added the “Pro” tag to the end of the K8T800 name in order to identify this chipset as the one with the two tweaks described above. This has nothing whatsoever to do with whether the chipset is intended for use in Athlon 64 or Opteron systems. Both the K8T800 and K8T800 Pro are equally at home in sweet, sweet dual Opteron motherboards and low-end Athlon 64 boards.

NVIDIA, however, has elected to use the “Professional” label a little bit differently. The nForce3 150 is NVIDIA’s Athlon 64 chipset, while the nForce3 Pro 150 is the Opteron-oriented product. How are they different from one another? The name, mainly, and perhaps the price—but not the silicon.

Just so that’s clear to everyone.

Test notes
We were able to spend a little bit of quality time with a K8T800 Pro reference board, and we have a quick set of benchmarks for you. This was a Socket 940 board, so we’re able to compare it to a K8T800 and to an nForce3 150. We had hoped to compare the K8T800 Pro to NVIDIA’s recently unveiled nForce3 250GB chipset (which we just reviewed) but we couldn’t locate a Socket 940-based nForce3 250Gb mobo just yet. Seems everyone’s waiting for Socket 939.

Additionally, time limits and some limitations of the VIA reference board prevented us from conducting the full range of chipset tests we generally like to perform. Reference boards don’t always have all the features one might wish for implemented and ready for testing, and such was the case with this board. As a result, we’ve just run through our standard CPU test suite to get a look at the K8T800 Pro’s basic system performance. The VT8237 south bridge is nothing new, anyhow, and you can get a sense of its performance by reading our other recent chipset articles, like our nForce3 250Gb review.

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:

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.

The test systems’ Windows desktops were set at 1152×864 in 32-bit color at an 85Hz 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 (9.89)
• Compiled binary of C Linpack port from Ace’s Hardware
• Discreet 3ds max 5.1 SP1
• NewTek Lightwave 7.5
• Cinebench 2003
• POV-Ray for Windows v3.5
• PICCOLOR v4.0 build 472
• SPECviewperf 7.1.1
• ScienceMark 2.0 beta (23SEP03 build)
• Sphinx 3.3
• LAME 3.95.1 (build from mitiok.cjb.net)
• Xmpeg 5.0.3 with DivX Video 5.11
• FutureMark 3DMark03 build 340
• Comanche 4 demo
• Quake III Arena v1.31
• Serious Sam SE v1.07
• Splinter Cell v1.2
• Unreal Tournament 2003 demo v.2206
• Wolfenstein: Enemy Territory v2.55

The tests and methods we employ are generally publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.

Memory performance
I’d best warn you up front that the following set of benchmarks is not likely to evoke a cathartic, life-changing experience. You’re in little danger of forgetting to breathe, either, although you may well doze off for a second and slam your forehead into your keyboard.

Because the Athlon 64 processor has its memory controller built in, performance differences between A64 chipsets are potentially very small, especially when those chipsets are implemented well. Such is the case with all three of the motherboards we’ve tested here. Any significant performance differences we’ll see in these tests will likely come from several sources, including differences in BIOS and motherboard implementations. One of the few notable differences between the nForce3 Pro 150, the K8T800, and the K8T800 Pro is the speed of the HyperTransport link. The nForce3 Pro 150 is kinda pokey on that front with a 16-bit, 600MHz downstream link and an 8-bit, 600MHz upstream. NVIDIA has addressed this limitation with the nForce3 250 series, however.

Since these results are so scintillating, I’ll let them speak for themselves, and limit my commentary on the following pages. I’ll discuss things as needed, then sum up at the end.

As I said, any memory performance differences are likely the result of motherboard or BIOS variations, because the chipset has little to do with memory access on the Athlon 64.

Unreal Tournament 2003

Quake III Arena

Wolfenstein: Enemy Territory

Tom Clancy’s Splinter Cell

See a trend yet? For gaming, at least, the K8T800 Pro is slightly slower than the K8T800, perhaps due to its asynchronous AGP/PCI clocks. However, the nForce3 Pro 150 is a little slower than either VIA chipset.

Comanche 4

Serious Sam SE

3DMark03

The trend holds solid throughout our gaming benchmarks, but the performance differences are so minor, you may not have noticed.

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.

LAME MP3 encoding
We used LAME 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

Yep. Uh huh.

3ds max rendering
We begin our 3D rendering tests with Discreet’s 3ds max, one of the best known 3D animation tools around. 3ds max is both multithreaded and optimized for SSE2. We rendered a couple of different scenes at 1024×751 resolution, including the Island scene shown below. Our testing techniques were very similar to those described in this article by Greg Hess. In all cases, the “Enable SSE” box was checked in the application’s render dialog.

POV-Ray rendering
POV-Ray is the granddaddy of PC ray-tracing renderers. POV-Ray relies more heavily on x87 FPU instructions to do its work, because it contains only minor SIMD optimizations.

Lightwave rendering
NewTek’s Lightwave is a popular 3D animation package that includes support for multiple processors and is highly optimized for SSE2.

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.

VIA’s faster HT links and perhaps its AGP implementation shows us a little something in the hardware shading test, as the K8T800 and Pro open up a slight lead, for once, on the nForce3 Pro 150. Perhaps viewperf can show us more of the same…

SPECviewperf workstation graphics
SPECviewperf simulates the graphics loads generated by various professional design, modeling, and engineering applications.

The VIA chipsets spice things up a little with a relatively strong showing in viewperf. Perhaps if we had a pro wrestling announcer doing play-by-play for us, these margins of victory would be more exciting.

ScienceMark
I’d like to thank Alex Goodrich for his help working through a few bugs the 2.0 beta version of ScienceMark. Thanks to his diligent work, I was able to complete testing with this impressive new benchmark, which is optimized for SSE, SSE2, 3DNow! and is multithreaded, as well. Unfortunately, we don’t yet have a version of ScienceMark capable of taking advantage of SSE3’s new complex arithmetic instructions.

In the interest of full disclosure, I should mention that Tim Wilkens, one of the originators of ScienceMark, now works at AMD. However, Tim has sought to keep ScienceMark independent by diversifying the development team and by publishing much of the source code for the benchmarks at the ScienceMark website. We are sufficiently satisfied with his efforts, and impressed with the enhancements to the 2.0 beta revision of the application, to continue using ScienceMark in our testing.

The molecular dynamics simulation models “the thermodynamic behaviour of materials using their forces, velocities, and positions”, according to the ScienceMark documentation.

Scientists believe that members of the camel family lived in North America at least 40 million years ago. Before the Ice Age, camels had developed into a distinct species and had moved westward across Alaska to western Asia. In Asia, two groups separated and gradually became the two chief kinds of camels known today. Meanwhile, smaller members of the camel family had moved southward from North to South America. Today, four members of the camel family live in South America: (1) alpacas, (2) guanacos, (3) llamas, and (4) vicunas. By the time Europeans went to North America, no members of the camel family had lived there for many thousands of years. No one knows why they disappeared.

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 is partially multithreaded, and it shows us the results of a number of simple image manipulation calculations.

Moving along…

Conclusions
Like I said, the benchmark results didn’t exactly light up the world. All three chipsets were pretty evenly matched. However, the K8T800 Pro gives up very little performance in gaining the flexibility of an independent PCI/AGP clock. That alone will give the K8T800 Pro a solid position in the Athlon 64 world when Socket 939 arrives.

I wish we could have overclocked the VIA reference board, but our time with it was limited, and as I said, reference boards don’t generally offer a robust suite of overclocking options. Overclocking exploits will probably best be performed on a production motherboard, anyhow.

The K8T800 Pro will have to contend with a couple of flavors of NVIDIA’s nForce3 when Socket 939 debuts. For lower-end mobos, the K8T800 Pro will go head to head with the nForce3 250, a cheaper version of nForce3 that lacks GigE networking. Against the nForce 250, the K8T800 Pro should hold its own just fine.

For higher end enthusiast boards, the K8T800 Pro will face a stiffer challenge against the nForce3 250Gb. The PCI bus is simply inadequate (at least at 32 bits and 33MHz) for Gigabit Ethernet, so south bridge-based GigE is a compelling feature right now. VIA has solid GigE technology of its own, but this tech hasn’t found its way into a VIA south bridge yet. VIA also needs to step up its software development a little bit, because NVIDIA’s RAID, firewall, and tweaking utilities are looking pretty slick. I could see myself picking an nForce board over a K8T800 Pro on the strength of the nForce3 250Gb’s better software and GigE support. VIA could recapture the initiative by pulling its Envy24 audio technology into a new south bridge along with GigE, instantly giving its chipsets an edge over the Soundstorm-deprived nForce3.

For now, the K8T800 Pro north bridge with the VT8237 south bridge will certainly suffice. We should be getting our hands on a production K8T800 Pro motherboard soon, and sweet, egregious abuse of an Athlon 64 processor will ensue. Stay tuned.