Home Preview: VIA’s K8T890 chipset
Reviews

Preview: VIA’s K8T890 chipset

Scott Wasson
Disclosure
Disclosure
In our content, we occasionally include affiliate links. Should you click on these links, we may earn a commission, though this incurs no additional cost to you. Your use of this website signifies your acceptance of our terms and conditions as well as our privacy policy.

WITH A PARSIMONIOUS 133 megabytes per second of shared bandwidth, PCI has definitely become the short bus of PC expansion standards. Nearly every other port, slot, and link inside a modern computer is faster than the bus shared by a collection of 32-bit, 33MHz PCI expansion slots and a couple of onboard devices on the typical PC motherboard. A single Serial ATA connection can burst up to 150MB/s, saturating an otherwise-empty PCI bus.

Fortunately, help arrived recently for Pentium 4 motherboards with the introduction of Intel’s 915 and 925X chipsets. These chipsets replace the tired PCI bus with the much faster, more modern PCI Express standard. Shared PCI slots give way to PCI Express X1 slots, which offer 250MB/s of dedicated bandwidth in each direction, or 500MB/s total, to a single device. PCI Express also replaces the PCI-derived AGP standard with a new PCI Express X16 slot that offers up to 8GB/s of total bandwidth, or nearly four times the bandwidth AGP 8X.

All of this is well and good, but there’s been a catch. You see, the Athlon 64 has been hammering the Pentium 4 in many types of performance benchmarks, particularly the gaming ones, for nearly a year now. What’s more, the latest versions of the Pentium 4 pull lots of power and generate lots of heat, making them not the most attractive CPU option. Right now, anyone buying a new system would have to choose between an Athlon 64 with AGP/PCI slots and a Pentium 4 system with PCI Express. That situation is about to be corrected, because chipset manufacturers are prepping core logic chipsets for the Athlon 64 with support for PCI Express. The first of those chipsets that has made its way into Damage Labs is VIA’s K8T890. Read on to see our hands-on report.

The K8T890 north bridge
The K8T890 north bridge chip is the heart of VIA’s new PCI Express chipset. This chip’s internals look very different from a traditional north bridge. The K8T890 has no memory controller because the Athlon 64 has its own memory controller built in, and the K8T890 has no front-side bus or AGP interface. Instead, the K8T890 has links to a couple of other chips: HyperTransport to the CPU and an Ultra V-Link interconnect to the south bridge. Beyond that, the K8T890 is all about PCI Express. 20 lanes of PCI Express, to be precise, stand ready to connect the K8T890 to the rest of the system. In a typical configuration, K8T890 motherboards will devote sixteen PCI Express lanes to graphics, leaving four more to power PCI-E X1 expansion slots. Because this north bridge is so PCI-E-rich, the chip can be mated with VIA’s existing VT8237 south bridge and still drive a full-featured PCI-E motherboard.


The K8T890 on VIA’s prototype motherboard

Of course, the K8T890 inherits most of the features of the K8T800 Pro that came before it, including asynchronous operation. In this case, that means the K8T890’s system bus (or HyperTransport link) can be overclocked without affecting the clock speed of the PCI Express lanes. Overclockers should be able to enjoy the same freedom to crank up CPU speeds that they’ve enjoyed on the K8T800 Pro without fear of forcing their PCI-E graphics cards or other peripherals to run out of spec.

 

PCI Express dips below the equator
VIA’s Ultra V-Link interconnect will join the K8T890 north bridge with one of a couple of different south bridge chips. Initially, the K8T890 will be paired up with the VT8237 south bridge, as it was in the prototype motherboard we tested. Soon, though, VIA’s new VT8251 will hit the market and quickly become the preferred companion to the KT890 north bridge.


A block diagram of the K8T890 chipset. Source: VIA.

Many of the VT8251’s features mirror those of the ICH6 south bridge of the Intel 900-series chipsets. The VT8251 will include two PCI Express lanes, just enough to outfit a motherboard with a pair of PCI-E X1 expansion slots without overwhelming the 1.06GB/s Ultra V-Link interconnect to the north bridge. VIA has added support for Intel’s High Definition Audio standard, so the new south bridge will be able to stream eight channels of audio data at resolutions up to 24 bits and sample rates to 192KHz. Also, like the ICH6R, the VT8251 will support the AHCI standard for Serial ATA drives. AHCI includes provisions for hot-plugging devices and the performance-enhancing Native Command Queuing, raising the very real possibility that Serial ATA will kill off SCSI, even in high-end servers.

Because Serial ATA is growing in popularity, VIA has endowed the new south bridge with four ports of SATA connectivity without the need for an external PHY chip. All four of the drives connected to these ports will be able to participate in RAID arrays, with RAID levels 0, 1, 0+1, and JBOD on offer. Two ATA/133 ports remain, as well, for older hard drives and other storage devices.

One of the more interesting decisions VIA made with the VT8251 was to leave out Gigabit Ethernet support and stick with 100Mbps Ethernet. VIA seems to have determined that with PCI Express, Gigabit Ethernet would best be left to an external controller chip. In fact, VIA’s press materials on the VT8251 cite higher performance with PCI-E-based GigE than with NVIDIA’s integrated GigE solution. I suspect VIA will have success with this approach, because it is likely to please the penny pinchers at Taiwan Inc.’s motherboard manufacturers. By contrast, NVIDIA has pledged to offer more and better hardware acceleration of networking tasks in its future chipsets. I’ll be interested to see whether NVIDIA can continue to differentiate its product (and charge more for it) by offering its integrated GigE and firewall.

More north bridges planned
VIA has two more 890-series north bridges in the works for the Athlon 64. The most exciting is probably the K8T890 Pro, equipped with what VIA is calling DualGFX Express. This is VIA’s SLI-ready north bridge with support for dual graphics slots. Judging by everything I’ve seen and heard, I don’t believe VIA has worked out yet just how the allocation of PCI Express lanes to each of the two slots will work, but the company does appear to be committed to the idea of building an SLI-capable chipset. The diagram of the DualGFX Express architecture VIA supplied is deliciously vague:

Realistically, I tend to think the vanilla K8T890 north bridge should be able to drive a perfectly acceptable dual PCI-E graphics implementation. Eight lanes of PCI Express per slot should be plenty for a good while yet. I’m curious to see how VIA decides to handle it. They will probably have to match NVIDIA lane for lane on both slots in order to sell their solution. Already, NVIDIA has been informing the world that the next nForce will be the only “certified” SLI chipset for the Athlon 64. That confidence speaks volumes about the openness of the certification process.

VIA’s roadmaps say the K8T890 Pro should be sampling right about now, with production coming later this year.

Another K8T890 variant VIA’s cooking up is a mobile/value version of the north bridge with integrated graphics. The reasonably decent DeltaChrome graphics technology from S3 will finally find a suitable home in the K8M890. If S3 follows through on its unified driver plans, the K8M890 could be VIA’s most competitive integrated graphics chipset in ages. The KTM890 should also arrive some time this year, after it samples in October, if all goes as planned.

 

The K8T890 in action
We had the chance to play with a K8T890 reference board for a little while, and we were able to extract some benchmark scores from it. This puppy was obviously nothing close to a production motherboard. When we test chipsets, we like to test performance twelve different ways from Sunday, including lots of south bridge I/O tests; those are becoming much more important now that the Athlon 64’s integrated memory controller has robbed north bridge benchmarking of its usual drama. Unfortunately, the VIA reference board was neither designed nor intended to serve as a showcase of south bridge performance. What’s more, the reference board didn’t have the new VT8251 on it, just the familiar VT8237 we’ve already tested extensively.


VIA’s K8T890 reference mobo

On top of all that, the reference board wasn’t up for the task of running at really tight RAM timings or engaging in dubious overclocking adventures. We were able to get it stable with relatively conservative memory settings, but we just couldn’t push too hard. VIA told us they didn’t want to put too much effort into developing their reference boards because full production motherboards should be available from mobo makers like Abit in the very near future. In fact, this board from Abit is apparently ready to roll as soon as VIA gets volume chipset production going.

Given all these considerations, we decided to confine our testing to the thing that matters most on the K8T890 north bridge: PCI Express graphics. (We’d have tested other types of PCI-E cards, too, if we could find any.) Part of our goal was simply to verify that the K8T890 works as advertised, and part of it was to see whether the performance was acceptable. Based on our experience with the Intel PCI-E chipsets, we knew better than to expect immediate or dramatic performance increases in games and graphics.

We hit one snag immediately: the reference board’s PCI-E X16 slot wouldn’t accommodate a GeForce 6800 GT card. The CMOS battery and a capacitor were situated in the way of longer PCI-E graphics cards, so we had to confine our testing to a Radeon X600 XT. The X600 XT isn’t an exact match for the Radeon 9600 XT on the AGP side of things, but it’s close enough for comparison. The GPUs are essentially the same, with the same core clock speed, but the X600 XT has a higher memory clock speed than the 9600 XT. In order to even things up a little bit, I overclocked the memory on our 9600 XT to 720MHz. That’s 20MHz short of the X600 XT’s 740MHz RAM, but it was the best the poor card could do. In any event, we tested at low resolutions so that video memory bandwidth shouldn’t be too much of an issue.

We will, of course, have a fuller look at K8T890 performance and compatibility when production motherboards based on the chipset arrive. For now, let’s see what the reference board had to show us.

 

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 64 3800+ 2.4GHz Athlon 64 3800+ 2.4GHz
System bus HT 16-bit/800MHz downstream
HT 16-bit/800MHz upstream
HT 16-bit/800MHz downstream
HT 16-bit/800MHz upstream
Motherboard MSI K8T Neo 2 VIA K8T890 reference
BIOS revision 3.0 070604 Beta
North bridge K8T800 Pro K8T890
South bridge VT8237 VT8237
Chipset drivers 4-in-1 v.1.11 beta (9/7/04) 4-in-1 v.1.11 beta (9/7/04)
Memory size 1GB (2 DIMMs) 1GB (2 DIMMs)
Memory type Corsair XMS3200XL DDR SDRAM at 400MHZ Corsair XMS3200XL DDR SDRAM at 400MHZ
CAS latency 2 2
Cycle time 6 6
RAS to CAS delay 3 3
RAS precharge 3 6
Command rate 2T 2T
Hard drive Maxtor MaxLine III SATA 250GB
Audio Integrated
Graphics Radeon 9600 XT 128MB AGP Radeon X600 XT 128MB PCI-E
OS Microsoft Windows XP Professional
OS updates Service Pack 2, DirectX 9.0c

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:

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.

 

Benchmark results
In order to see how the K8T890 performs, we’ve tested with three of the most graphics-intensive games around. First up is DOOM 3. We tried a couple of different quality modes for this game, to see what would happen. The medium quality mode uses smaller textures and is recommended by id Software for cards with 128MB of memory. The high quality mode is recommended for cards with 256MB, so there should be more texture data flying about in that mode.

In either mode, the K8T890 performs about on par with the K8T800 Pro. It’s a little slower in both cases, but the margin of difference is slim.

Next up is the video stress test of the Source engine that will power Half-Life 2.

The K8T890 Pro actually looks a little faster here, but not by enough to count for much.

Next, we have three different Far Cry demos.

Again, it’s very close, but the K8T890 holds its own. Overall, its gaming performance isn’t bad, but we’d hope production boards will match the AGP-based K8T800 Pro stride for stride.

Finally, we’ll look at a synthetic test that copies an image from the video card’s frame buffer back into main memory. Unlike the gaming benchmarks, we’d expect this one to show us some real differences between AGP and PCI-E.

Not bad. We saw about 369MB/s transfer rates with the Intel 925X chipset, but that was with WinXP Service Pack 1 and a different version of ATI’s drivers, so I hesitate to make to direct a comparison. Obviously, the K8T890 Pro’s PCI-E X16 slot offers quite a bit more bandwidth than an AGP slot.

 
Conclusions
Our quick look at the K8T890 was tantalizing, and we’re waiting expectantly for production motherboards based on this chipset. The thing seems to work right, and the PCI-E implementation does deliver more bandwidth than AGP. Even without the new VT8251 south bridge, the K8T890 north bridge could do a decent job of ushering the Athlon 64 into the PCI Express era.

I wish I could tell you precisely when that is going to happen, but the truth is, I’m not quite sure. VIA is working to get the K8T890 to market as soon as possible, and they clearly want to beat NVIDIA to the punch. I think whoever gets there first stands to capture the hearts of upgrade-hungry enthusiasts who have been putting off a purchase until the right combination of Athlon 64 and PCI Express became available.

We just need to test a production board now and see how it performs in our full suite of tests. We promise to gang tackle the FedEx guy if we think he might be delivering our Abit AX8—every day until it arrives. 

Latest News

Ripple Dumps 240 Million XRP Tokens Amid 17% Price Decline
Crypto News

Ripple Dumps 240 Million XRP Tokens Amid 17% Price Decline

Crypto Expert Draws A Links Between Shiba Inu And Ethereum
Crypto News

Crypto Expert Draws Link Between Shiba Inu And Ethereum

The founders of the second-largest meme coin, Shiba Inu, have been a mystery. However, many people within the crypto industry have made some probable suggestions regarding the coin’s fundamentals. Del...

The Lucrative FTX Bankruptcy Trade and Ongoing Legal Battle
Crypto News

The Lucrative FTX Bankruptcy Trade and Ongoing Legal Battle

The FTX cryptocurrency exchange crash in November 2022 left a trail of aggrieved investors and customers who had their funds frozen due to the complex bankruptcy case.  However, the collapse...

Bitcoin (BTC) Price Set to Enter “Danger Zone” – Time to Back-Off or Bag More Coins?
Crypto News

Bitcoin (BTC) Price Set to Enter “Danger Zone” – Time to Back-Off or Bag More Coins?

SNB to Kick Off Rate Cut Cycle Sooner Than Expected
News

SNB to Kick-Start Rate Cut Cycle Sooner Than Expected

Top Crypto Gainers on 18 March – AVAX and RNDR
Crypto News

Top Crypto Gainers on 18 March – AVAX and RNDR

smartphone security organization
Community Contributions

How to Successfully Tackle Smartphone Security in Your Organization