After posting an article critical of some of VIA’s practices a short while ago, I was contacted by the company with some additional information, and I thought it would be a good idea to follow up, for those of you who are interested.

First off, and most importantly, I have been assured that the KT400 will absolutely not launch in April. Clearly the prior information I had received was inaccurate, and for that I make no excuses. There will be no April launch. Nevertheless, KT400 is coming, and probably sooner than later, so the real issue remains unchanged.

Our contact with VIA can be summarized as follows: VIA acknowledged that the chipset market is currently in an ‘extremely dynamic’ phase, leading to faster product cycles. This places the company in a difficult position -- an assertion I would agree with. VIA pointed out that, while it will release a DDR400 chipset, it will only do so after extensive validation, and pointed out that this was the case with its KT333 as well, which launched several months after similar chipsets from SiS. VIA further mentioned that SiS and nVidia were demonstrating DDR400 chipsets, and that VIA needs to remain competitive.

But VIA Did Spend Time Validating DDR333

Granted, but I submit that it still wasn’t enough.

It’s certainly true that VIA’s DDR333 chipset was released after SiS’, and that VIA spent more time validating the platform. VIA even mentioned a one-day summit held a short while ago for companies involved with DDR333. A one-day summit is nice, but not enough. DDR333 is not ready. The company says it will only release a DDR400 chipset when it feels it’s ready. I don’t dispute that; I dispute VIA’s idea of what ‘ready’ actually is.

It’s no coincidence that the chipset manufacturer with by far the most stringent and demanding validation procedures hasn’t touched DDR333 or DDR400 with a ten foot pole yet.

But SiS and nVidia Are Demonstrating DDR400

So? Talk about chasing the wrong horse. VIA wants to be a leader, and that’s fine. Every company does. But then chase the leader, not SiS. nVidia isn’t exactly a model of success in the chipset business either.

As I said initially, VIA needs to decide just how serious it is about DDR400. If the company is serious, and wants itself and its products to be taken as such, it needs to do better than it currently is with its validation procedures. It needs to decide whether it’s more important to have a product out right away, or have a more solid product with better silicon and drivers later on. Sometimes you can’t have both, and until now, VIA has always chosen the former more than the latter. If it ever hopes to get beyond the low- to mid-end desktop market, it needs to learn how to choose the latter.

So let SiS release its DDR400 chipset. It’ll probably get about as much attention as SiS’ DDR333 chipsets did. If consumers think VIA’s chipset will be worth waiting for, they will. Give them something better to wait for, not more of the same. Intel gave manufacturers and consumers a better DDR266 chipset to wait for in Brookdale-D, and as a result, virtually no manufacturers even considered the P4X266 (before you claim legalities were the reason behind that, look at how many manufacturers considered SiS’ perfectly-legal P4 DDR chipset). Why would they, after all, when a clearly better chipset was just around the corner? We talk to enough motherboard manufacturers to be able to say with absolute certainty that virtually all of them consider 845D to be a more mature, stable, and reliable product, and well worth the wait.

VIA has produced some good chipsets. There's no reason to believe that VIA isn't capable of producing great chipsets, but in order to do so, it needs to slow down, worry less about others, and concentrate on proper validation and specification. If that doesn't happen, then mediocre chipsets are the best we're going to see.

To nVidia, ATi, SiS, or anyone else who thinks DDR400 is ready to be introduced within the next few months; it’s not, and pushing this too fast will only do more damage, both to the concept of DDR400 (when it really is ready), and to the companies' respective reputations.

Dan Mepham

(If you'd like some more empirical information on some of our own experiences with DDR platforms, such as the VIA and SiS platforms for Pentium 4, you might want to read this article too.)