As often happens in these discussions, possible causes of the lock-up problems I've described started breeding like mink; everything from the power supply (I'm using 300W Sparkles) to T-bird problems (I'm using Athlon classics) was suggested. However, I'd also asked to hear from satisfied owners of KX133-based systems, and I did hear from them. The positive response wasn't overwhelming, but it was solid enough to convince me there are people out there whose KX133 rigs are quite stable.
So I decided to revise my view a bit, based on several facts:
1) The EPoX EP-7KXA I'd used for testing is its own problem, independent of any chipset issues.Further digging around brought a reminder that Tom Pabst had already addressed the problems with KX133 boards in Windows 2000. Though I thought Via would have corrected the issues Tom raised back in early May, it appears the latest 4.03 Via AGP drivers don't replace the default Windows 2000 viaagp.sys, which is problematic. (An updated version is available at Tom's.) Beyond that, neither the default nor updated viaagp.sys file is truly stable. It appears Via has yet to deliver a stable KX133 AGP driver for Windows 2000.
2) Barring the EPoX, all of my KX133 testing has been conducted on a pair of Abit KA7s in Windows 2000 using a GeForce2 GTS card.
I have to wonder how much such stability issues are caused by poor drivers, versus how much Via's driver development is a delicate exercise in compensation for inherently problematic hardware. Though I did get positive feedback from some satisfied KX133 owners, a handful wrote to say they had seen the same AGP lock-up problemand not just from Win2000 users. I honestly don't know what's going on hereor whether we're looking at anything more that the usual assortment of tech problems that folks run across every daybut it leaves me with a bad taste in my mouth. I'll keep an eye on this one. Let me know if you hear anything.