<!–#set var="article_header" value="VIA's DDR-Runner For Athlon –
The Apollo KT266 Chipset” –>
Introduction
In German we have a saying that goes “Gut Ding will Weile haben”, meaning something like that good things should not be rushed. This seems to always have been VIA’s motto. I still remember the announcement of VIA’s first Pentium Pro chipset in 1997, back then already called ‘Apollo Pro’. This chipset never made it into production, so that two years later another chipset for Pentium II and Pentium III processors was able to carry the very same name. It wasn’t any different with VIA’s support of AMD’s Athlon processor. It had been the plan to release the Apollo KX133 chipset together with the first Athlon processors. In the end the first Athlon supporters had to go for AMD’s 750 chipset and it took no less than 5 months until Apollo KX133 was finally ready for the market.
The story with Athlon combined with DDR-memory doesn’t seem much different. AMD’s launch of the first Athlon processors with 133 MHz FSB and the first DDR-chipset, the AMD760, is also about 6 months ago. We have to be fair and admit that it took quite a while until motherboards with 133 MHz FSB support and AMD760 chipset finally hit the market, but there’s hardly any excuse why VIA has still not managed to release their highly anticipated DDR-chipset Apollo KT266.
Finally, KT266 is becoming available now. What we are wondering though is if VIA is indeed following the above-mentioned German saying. This saying is not only about the fact that something may take a while longer. It is talking about ‘good things’. This would indeed be the best excuse. You can afford to be late, but then you better be damn good. This article will now find out if Apollo KT266 is indeed the ‘good thing’ that it was worth waiting for. AMD’s 760 chipset has so far proven to be very capable to stand its ground. We have already seen one other Taiwanese chipset maker failing to live up to the qualities of AMD’s DDR-chipset. The performance of ALi’s MaGiK1 chipset is anything but as ‘magic’ as its name would like to make you believe. Then there is VIA’s Apollo KT133A chipset. It comes with ‘old fashioned’ PC133-support, but works with Athlon processors that run their processor bus at ‘266’ MHz (in reality 133 MHz ‘double-pumped’). Even AMD’s 760 chipset had problems to provide significantly better performance than this ‘beefed up’ version of VIA’s original Apollo KT133 chipset. Will the DDR-product from the same ‘breed’ be able to leave its older sibling far behind?
The Specifications Of VIA’s Apollo KT266
As you can see from the picture above, the Apollo KT266 chipset consists of the VT8366 north bridge and the VT8233 south bridge, which we already know from VIA’s Apollo Pro 266 Pentium III DDR-chipset. The main catch of Apollo KT266 in comparison to its predecessor KT133(A) is the new south bridge and its connection to the north bridge via VIA’s “V-Link”, a 8-bit wide data path clocked at 133 MHz, double-pumped and thus capable to transfer data at up to 266 MB/s. Previous chipsets used the PCI-bus as the link between north and south bridge, which only provides 133 MB/s, that needed to be shared with all the other PCI-devices. V-Link is obviously VIA’s version of Intel’s “hub-architecure”, which uses a similar connection between north and south bridge.
Besides this new feature KT266 offers support of DDR-SDRAM as well as support of the already known memory types SDR SDRAM (up to PC133) and virtual channel SDRAM (VCM). This wide support of memory types can also be found in ALi’s “unmagic” MaGiK1 DDR-chipset, while AMD’s 760 is a true DDR-only solution, that does not care for any slower memory type.
The VT8233 south bridge supports all the goodies found in today’s south bridges, such as ATA100, AC-97, MC-97, six USB-ports and integrated network MAC for an external PHY that offers 100/10 Ethernet connectivity. Finally it comes with support of AMD’s PowerNow! Technology, which makes it useful for mobile use, as e.g. in notebooks.
From the technological point of view, VIA’s Apollo KT266 gives a very good impression, while making it look a worthy competitor to AMD’s 760 chipset. However, the proof is in the pudding and we will have to see if KT266 can also provide proper performance besides all those nifty features.
The Sample Motherboard – MSI’s K7T266 Pro MS6380
The Taiwanese motherboard maker MSI has lately proven to be a very reliable provider of excellent platforms, which can easily compete against the products of earth enemy ASUS in terms of performance, reliability, quality as well as special features. MSI’s excellent performing, though still unreleased K7 Master motherboard with AMD 760 chipset has therefore been able to push ASUS from its throne of being the permanent provider of Tom’s Hardware reference motherboards.
MSI K7Master-S, our present AMD760 reference board
In case of VIA’s Apollo KT266 it seems to go in the same direction. While Asus has still not got its KT266-product ready, MSI is already able to supply the press with reliably performing samples of its upcoming K7T266 Pro motherboard.
The K7T266 Pro is equipped very nicely. Its three DDR-DIMM slots can be equipped with up to 3 GB of DDR-memory. It comes with 5 PCI slots, one AGP-Pro slot and one CNR-slot. Besides the two ATA100 IDE channels provided by the VIA south bridge, you will also find a Promise RAID0/1 ATA100 IDE-controller, full cable support of the six USB-ports as well as a PC2PC USB-network connection. The BIOS setup of the MS6380 offers a wide selection of overclocking features, so that everybody should be pleased with this motherboard. With the board we used brand new and excellent performing PC2100 / CL2 DDR-SDRAM from Infineon, which performed at its best even at automatic settings.
We have not become completely unfaithful to Asus however, as we are putting up the A7V133 as our reference board with VIA’s Apollo KT133A chipset against the two DDR-platforms from MSI.
Expectations
VIA’s Apollo KT266 represented by MSI’s K7T266 Pro will have to stick up against AMD’s 760 (MSI K7Master-S) and VIA’s own Apollo KT133A in form of Asus’ A7V133. What we would like to see is that VIA’s new chipset is clearly able to beat its older sibling KT133A and that it is able to at least reach the performance of AMD’s 760 chipset.
There is enough criticism of DDR-memory already, because even AMD’s 760 chipset is not able to provide the expected major performance boost over SDR-SDRAM solutions such as Apollo KT133A. VIA’s own Athlon-DDR solution cannot really afford to be slower than AMD’s chipset, because the least a user deserves to gain from DDR-memory is the 5-10% performance increase provided by AMD760.
ALi’s MaGiK1 chipset has already been the first DDR-flop, which justifies its continued existence only with the marketing-effect of the magic term ‘DDR’, but certainly not with any respectable performance. We don’t want to see another flop, especially not produced by one of the biggest supporters of ‘TeamDDR’.
Benchmark Setup
System | VIA Apollo KT266 | AMD 760 | VIA Apollo KT133A |
Processor | Athlon ‘C’ 1.33 GHz, overclocked to 1.466 GHz |
||
Motherboard | MSI MS-6380 “K7T266 Pro”, rev. 1.0 Pre-Release BIOS |
MSI MS-6341 “K7 Master-S”, rev. 0D Pre-Release BIOS |
Asus A7V133, rev. 1.01 BIOS 1004 |
Memory | 256 MB Infineon PC2100 DDR-SDRAM, CL2, Setting CL 2, 4-way Interleave | 256 MB Micron/Crucial PC2100 DDR-SDRAM, CL2, Setting 8-8-5-2-2-2-2 | 256 MB Micron/Crucial PC133 SDRAM, CL2, Setting 2-2-2 |
Graphics Card | NVIDIA GeForce 2 Ultra Reference Card, 64 MB, Driver 6.67 (Win98/Win2k) |
||
Hard Drive | IBM DTLA-307075, 75 GB, 7200 RPM, ATA100, FAT32 Win98, NTFS Win2k |
||
DirectX | 8.0a |
||
Desktop Resolution for BAPCo’s Sysmark 2000 and Webmark2001 | 1024x768x16x85 |
||
Quake 3 Arena | Retail Version no sound |
||
Unreal Tournament | Version 4.28 (patched) no sound |
||
Evolva | Rolling Demo v1.2 Build 944 Standard command line = -benchmark Bump Mapped command line = -benchmark -dotbump |
||
3D Mark 2000 | Build 335, Default Benchmark |
||
SiSoft Sandra Standard | Version 2001 |
||
Desktop Resolution for SPECviewperf 6.1.2 | 1280x1024x32x85 |
||
FlasK Settings | Video Codec: DivX 3.11 alpha, Fast-Motion, keyframe every 10 seconds, compression 100, data rate 910 kbps Audio Codec: audio not processed Video Resolution: 720×480, 29.97 fps, interlaced Resizing: Nearest Neighbor |
Memory Bandwidth
This chart gives already a strong hint of what we will have to expect of VIA’s Apollo KT266 in the following benchmarks. Sandra’s Stream-algorithms are once using the normal integer data formats (up to 32-bit) and once the larger FPU data formats (up to 80-bit), which also explains why the FPU/memory results are always higher than the integer results. This algorithm does not just measure peak bandwidth, but is also impacted by memory latency. It is clear to see that Apollo KT266 with DDR is way behind AMD’s 760 chipset and only little faster than the Apollo KT133A with PC133 SDRAM.
Office Application Performance Under Windows98
The good old Sysmark2000 application benchmark might not be too CPU and memory-intensive, but it already gives a pretty good idea what we will have to expect of the following results. AMD’s 760 is leading ahead of a KT266 DDR-platform that is barely able to beat the KT133A-motherboard with the cheaper PC133 SDRAM.
3D Gaming Benchmark Results
Nothing has changed. The fastest 3D-platform for Athlon processors remains AMD’s 760 chipset, while KT266/DDR is at least a few percent faster than KT133A with PC133.
At the higher resolution the impact of the graphics card grows, so that the results are closer together. However, AMD760 is clearly in the lead, while KT266 has a marginal edge over KT133A.
3D Gaming Benchmark Results, Continued
The NV15-demo is very CPU intensive and doesn’t care too much about memory performance. The succession is still the same though. AMD760 remains untouched.
The picture only changes marginally at the higher resolution.
3D Gaming Benchmark Results, Continued
Unreal Tournament is also showing the same again. AMD760 can indeed make a difference over KT133A, while KT266 seems more and more of a waste of DDR-memory.
The higher resolution is again not changing much. KT266 is hardly able to beat KT133A with the cheaper PC133 SDRAM. AMD 760 rules.
3D Gaming Benchmark Results, Continued
In Evolva the lead of AMD’s 760 is even more impressive, while KT266 can just about beat KT133A, which remains very close behind.
The higher resolution and the added bump-mapping are putting a high pressure on the GeForce2 Ultra card, so that all chipsets are performing pretty much the same.
There may be a new 3Dmark out already, but for this comparison 3DMark2000 is just fine. All chipsets are very close together, because this benchmark is mainly focused on testing the 3D-card. Still AMD760 comes out ahead of KT266, which leads a tiny bit ahead of KT133A.
MPEG4 Encoding With FlasK MPEG
This memory streaming benchmark is able to show once more how little KT266 is able to get out of DDR-SDRAM. KT133A with PC133 SDRAM is extremely close behind. AMD760 is again showing what is actually possible.
Professional OpenGL
SPECviewperf 6.1.2 is also mainly focused on things like 3D-card and AGP-performance. Still the memory performance of a system has some impact too. The picture is once again the same we have seen before. KT266 is failing to reach the performance level of AMD760.
Conclusion
A KT266 review of the German online magazine ‘CHIP Online’ that had been published very ambitiously in English was recently raising hopes that VIA’s Apollo KT266 chipset would not perform as poorly as what had been seen by many reviewers in the weeks before CeBIT. I was also able to see improvements, but although I was using the very same board and revision, KT266 is still way too slow to challenge or even replace AMD’s 760 chipset.
The step to the new and currently more expensive memory solution DDR-SDRAM should be justifiable with a noticeable performance gain, unless TeamDDR wants to play with the same questionable tactics as the Rambus/Intel combo. Right now, the only platform that could just about justify the usage of DDR-SDRAM is AMD’s Athlon processor plus AMD’s 760 chipset. Many people like to say that future processors at higher clock speeds will be able to take more advantage of DDR-SDRAM, but I would like to demystify this belief. As long as the processor bus specs of Athlon don’t change, it ain’t gonna be able to benefit from memory with higher data bandwidth. PERIOD! What it WILL benefit from however are lower latencies. VIA as well as ALi have presented DDR-platforms that may just as well provide the DDR-typical peak memory bandwidth of 2.1 GB/s, but at latencies that slow Athlon down rather than accelerate it. All the gain of DDR’s higher data bandwidth over PC133 SDRAM is lost again due to sad latency issues. TeamDDR has got to stop those things from happening, unless it wants to risk the marketing-value of the term ‘DDR’.
As a conclusion, I could maybe say the typical words always used in early reviews “let’s hope VIA will finally improve KT266”. However, I have my doubts if this will happen any time soon. My advice to you is to either forget about DDR altogether for the time being, or to go for Athlon plus AMD760 and NOTHING ELSE. ALi and VIA will try their hardest to benefit from the current DDR-hype to cover up the rather disappointing performance of their DDR-solutions.
Taiwanese motherboard maker won’t make this choice easy for you though. As it seems, VIA’s Apollo KT266 is about $10 cheaper than AMD’s 760. This is reason enough to produce more KT266 DDR-motherboards than AMD760-motherboards. There is also the rumor that AMD has problems supplying reasonable volumes of AMD760. As a result you will see hyped DDR-platforms with ALi or VIA chipset all over the place. The best way to handle this fuss is to ask yourself if you really need top performance. VIA’s Apollo KT133A still provides a reliable and very well performing Athlon platform and it only requires the cheaper PC133 SDRAM. If you really want DDR, go AMD760!