Although the VIA KT400 announced in autumn wasn't first in its family, its release was partially spoiled. The campaign of fast promotion of the DDR400 was to be given up because the JEDEC committee failed to be on the same wavelength regarding the prospects of the new standard with supporters of overclocking of the current memory generation. What does VIA have left? AGP 8x is a beautiful figure and it will please makers of video cards and video chips, but it's very difficult to draw the benefit from this technology - it's like convincing the customers that operation with a hard drive accelerates with the ATA133 protocol. Moreover, the situation with these AGP buses was quite tangled - now we have two updated versions of the KT333 chip: KT333 CF revision (with AGP 4x, but with DDR400) and VIA KT400. As for a south bridge, the tandem of KT400 and VT8235 with an integrated USB 2.0 controller doesn't look like an advantage of the new chipset as all latest VIA south bridges are pin compatible and can be used with any north bridge. 

The situation was aggravated by the first tests of the KT400 carried out with the DDR400. The first modules of this memory type had much inferior latency compared to the DDR333 (as they were often assembled on the same chips used for the DDR333 with overstated latency to reach 400 MHz). Consequently, the KT400 fell behind the KT333 (which was tested with the DDR333) and it made everyone think that the new chip had the inferior characteristics. 

But what chance did the KT400 have not to fail in the test? All earlier (before the autumn) AMD models had a frequency of the processor bus equal to 133(266) MHz, that is why the maximum data rates from the memory the processors could swallow didn't exceed 2.1 GB/s (it corresponds to the DDR266 bandwidth). When we tested the KT333 we showed that there was some benefit with the DDR333 used instead of the DDR266. First of all, because beside a CPU the system has other devices that require memory access, and the additional bandwidth can be taken over by them (for instance, a video card in games, a hard drive in office operations). Secondly, a flow rate is not the only parameter which defines efficiency of the memory subsystem: a random address access time can also be a key factor (in specific applications). 

Note that the benefit from the DDR333 used with the KT333 wasn't proportional to the increased bandwidth, there is only a slight speed-up. So, even if with the DDR400 modules at CL=2, the KT400 is able to only keep up with the KT333, at greater latencies the new chipset would definitely lag behind. To clear up the situation and to find out the real difference between two latest VIA chipsets we carried out a simple experiment the results of which are shown below. 

Test conditions

Testbed: 

• Processor: AMD Athlon XP 2100+ (13x133=1733 MHz), Socket 462(Socket A) 
• VIA KT400 based Mainboards: 
• ABIT AT7-MAX2 (BIOS B7) 
• ABIT KD7-RAID (BIOS B8) 
• ASUS A7V8X (BIOS 1004) 
• Chaintech 7VJL Deluxe (BIOS of 01.08.2002) 
• DFI AD77 Infinity (BIOS of 02.09.2002) 
• Elitegroup L7VTA (BIOS of 20.09.2002) 
• Gigabyte 7VAXP (BIOS F5) 
• MSI KT4 Ultra (BIOS 1.0) 
• QDI KuDoz 7X (BIOS 1.0) 
• Soltek 75FRV (BIOS AN1) 
• VIA KT333 based Mainboard: 
• Gigabyte 7VRXP (BIOS F6) 
• Memory: 2x256 MB PC3200(DDR400) DDR SDRAM DIMM Kingmax, CL 2.5 






• Video card: Palit Daytona GeForce4 Ti 4600 
• Hard drive: IBM IC35L040AVER07-0, 7200 rpm

Software: 

OS and drivers: 
• Windows XP Professional SP1 
• DirectX 8.1b 
• VIA 4-in-1 4.42 
• NVIDIA Detonator v40.41 (VSync=Off) 

Test applications: 

• VirtualDub 1.4.10 + DivX codec 5.02 Pro 
• WinAce 2.2 
• BAPCo & MadOnion SYSmark 2002 (Internet Content Creation + Office Productivity) 
• MadOnion 3DMark 2001 SE build 330 
• Gray Matter Studios & Nerve Software Return to Castle Wolfenstein v1.1 
• Croteam/GodGames Serious Sam: The Second Encounter v1.07 

Would you now take a look at the characteristics of the today's contestants:

Test results

Apart from traditional parameters of the boards based on the on chipset in question we have additional data to compare performance of the chipsets. The previous VIA's solution come on the fastest (see this comparison) board Gigabyte 7VRXP, and for direct comparison we have taken results of one of the KT400 based boards coupled with the DDR333 (based on PC3200 modules working at 333 MHz with minimal timings). Note that the boards on the new chipset are tested with the DDR400, because we are interested in the relative performance and stability of operation at 400 MHz which is not officially supported. Briefly, we have no complains about stability in this mode. 

Both the MPEG4 encoding and WinAce archiving with its 4 MB dictionary load the system most of all, and the difference between the chipsets must be the greatest here. The gap between KT400+DDR400 and KT400+DDR333 is just 1.5-2% but it's enough to put the second pair into the first position. And what about KT333? It (remember that we have given scores of the fastest board) almost keeps up with the top KT400 models being maybe just 2-3% behind. So, we can speak about equal scores of both chipsets, with just a trifling advantage of the later model. As for memory for the KT400, the DDR400 modules coming with worse timings bring in some additional delays, that is why such boards should be coupled with quality DDR333 modules. With the results obtained we don't have to continue our extended tests because the second aspect of difference between the chipsets - AGP 8x - is covered best of all in reviews of 3D accelerators in our Video system section. 

As for the KT400 based contestants, we have three leaders: ASUS A7V8X (the only board that managed to outpace the KT333 model even with the DDR400), Chaintech 7VJL Deluxe and ABIT KD7-RAID. The difference between the outsiders and the leaders is quite big: the Soltek 75FRV falls behind by 6%, and Elitegroup L7VTA by 7.5%. It's not typical of boards based on the same chipset; moreover, such a gap is comparable to that between successive processors in a line. 

The SYSmark 2002 traditionally shows that the boards do not differ markedly in real applications, though the leaders and outsiders are the same; an owner of the top board can count on a 3% gain in typical business applications and Internet content creation ones. However, top solutions are bought definitely not to speed up texting in in the MS Word. 

The games, even in low resolutions, don't show any difference between the contestants because the overall performance here primarily depends on a video card and a processor which are the same for all the boards. However, the ASUSTeK, ABIT and Chaintech look a bit better, while the last positions are again taken over by the Soltek 75FRV (2-3%) and Elitegroup L7VTA (3-7%). 

Conclusion

First of all, let me remind you that the KT400 is not worse than the KT333; usage of DDR400 modules with not minimal latency is not rational because this results in lower speeds at a higher price for such memory. Besides, as the DDR400 modules with minimal timings are not available yet, we can't affirm that they can perform better than DDR333. So, on the one hand, the only reason to recommend KT400 based boards is that they are equipped with the VT8235 south bridge (but remember that such combination can be found among KT333 based solutions as well). On the other hand, we can't warn you against KT400 boards, - it seems that the price/accessories ratio will be a determining factor when choosing a right model. 

As for mainboards the layout is as follows: the ASUS A7V8X and ABIT KD7-RAID take the lead in the tests, though the gap from the next models is pretty small. But the outsiders left an unpleasant impression: if the Soltek 75FRV managed to keep to the norm with its 3-4%, the Elitegroup L7VTA lags behind by 7%. That is why if the speed is crucial for you, the latter is not a good choice. 

The ABIT AT7-MAX2, ASUS A7V8X and Chaintech 7VJL Deluxe boards have good accessory packs, while both ABITs and Soltek 75FRV provide the most precise adjustment and overclocking. The special prize goes to the ABIT AT7-MAX2 and ASUSTeK boards which took the premium positions in almost all nominations. The brief descriptions of all the boards, plus their price tags will help you make the choice. 

It should be noted that we obtained the results for the AMD processor coming with a 266MHz bus. But right after the New Year AMD promises processors with a 333 MHz bus (not Hammer, but it's still a good solution, especially with the Barton core with 512 KB L2 cache). Will the situation with the KT266A->KT333 take place again? Will the updated KT400 (the current version doesn't support the asynchronous memory operation mode at FSB 166(333) MHz) get an advantage over the KT333 due to the bandwidth of the DDR400 provided that it supports DDR400 with new processors? Will the KT400's successors (that would support 333MHz FSB and DDR400 at the same time, not like current KT400) outdo the KT333 due to the bandwidth of the DDR400? How will latency of the PC3200 modules affect it? All the questions will be answered in special tests which are going to take place in our lab when the new AMD processors and the new VIA chipsets (if any) become widely available on the market. 
 
  

Write a comment below. No registration needed!