VIA Apollo KX133 Athlon Chipset - Part 1

Conclusion

Now that the highly anticipated KX133 is here, there are a number of things that must be said before you can simply base a conclusion on the performance figures mentioned in our tests.

First of all, finding motherboards based on the AMD 750 chipset will become increasingly difficult as the KX133 motherboard solutions become available on the market.  This should happen within the next two months, although you’ll be able to buy KX133 motherboards within a few weeks. 

Secondly, although the AMD 750 with SuperBypass enabled is on-par with, and sometimes faster than the KX133 even when using a 133MHz memory bus, keep in mind that not all revisions of the AMD 750 chipset support this feature.  So while the AMD 750 may be the faster overall chipset, keep in mind that not all revisions of the chipset are created equal.

Third, the KX133 may be slower than the AMD 750 with SuperBypass enabled, but things could be much worse.  For one thing, we noticed no AGP incompatibility problems with AGP 4X video cards and the KX133 chipset (AGP 4X was enabled for our test GeForce card using the Detonator 3.68 drivers) which was received by a huge sigh of relief after expecting quite a few AGP related problems from this chipset (judging by past history).

The performance of the KX133 isn’t noticeably slower than the AMD 750 either, and for a user looking to build a new Athlon system, going with the KX133 chipset will open them up to a variety of motherboards that wouldn’t be options if they were opting for a board based on the AMD 750 chipset.  The true test will have to be when other manufacturers begin shipping their KX133 boards.  It will be interesting to see if, with a few BIOS tweaks, the performance of the KX133 can be improved. 

The AGP 4X support of the KX133 isn’t a big deal, it’s mainly a marketing ploy.  From our tests, the difference between AGP 4X and AGP 2X is negligible.  The biggest performance benefit of the KX133 is its PC133 SDRAM support which can come in handy for applications that are very dependent on memory transfers (such as professional level applications, image editing software, voice recognition, etc…). 

The Virtual Channel SDRAM support of the KX133 looks very interesting on paper, but in reality it means nothing.  The relative unavailability of VC-SDRAM modules to anyone but this mysterious OEM that seems to be purchasing all of the chips is a big downside.  Secondly, the performance advantage, or lack thereof, that VC-SDRAM holds over regular SDRAM isn’t worth the trouble to go searching for it. 

The biggest selling point for the KX133 will be its cost.  Because of the highly integrated 686A south bridge and the 4-layer PCB design of the KX133 reference boards, motherboards based on the KX133 chipset should be noticeably lower priced than those based on the AMD 750.  It is because of its cost that the KX133 chipset gets our recommendation, not because of its AGP 4X support or VC-SDRAM support, but because of the bottom line: its pricetag.

If you’re going to be building an Athlon system in the next few weeks you’ll probably want to go for a KX133 based motherboard.  If you can wait, then in a few months AMD should have their own DDR chipset for the Athlon that will assume the place of the KX133 as the chipset to have and it should offer a nice performance premium over the KX133 and the AMD 750. 

On the topic of dual processor Athlon chipsets, VIA will not be producing one in the foreseeable future leaving AMD as the only option for a dual processor Athlon chipset.

In the end, is it really all that odd that CPU manufacturers seem to know how to make chipsets that allow their CPU to perform optimally?