Installation of Win98 was painless - you'll need the driver CD to get full support for the VIA KX133 chipset of course, but it's in the box, nestled snugly with a floppy cable and a single UDMA/66 cable. A standard IDE cable would've been a nice cheap inclusion but Epox elected not to include one. The manual is excellent, very clear and full of screenshots and diagrams. There's also a full copy of Norton Ghost included.

Overclocking and Stability
Those of you who read my K7M review will recall I concluded that Athlon overclocking was still firmly in the hands of the soldering-iron wielders. This was due to the K7M's inability to run at higher than 110MHz FSB (and, to be honest, it wasn't entirely stable even there). A few reviewers pointed the finger at the AMD 750 chipset used on that board so I was eager to see if the VIA KX133 used on the Epox would be more flexible. I could get an Athlon 650 to run stably at 110MHz but it would crash at 115MHz. I thought this might be a limit of the core, though, so I got Jim's Athlon 500 (which has been proven via a gold-finger device to be stable at 650) and cranked it up to 115MHz. It ran the Unreal Flyby with SETI in the background for a good couple of hours at this FSB, for a core speed of 575MHz. Not an exhaustive test, but it shows the chipset to be much more flexible in this regard than the AMD 750. However, I can already hear the BX-board owners scoffing at a meagre 15% FSB increase. Hopefully this is a good omen for the chipset and other implementations will allow this to be pushed further. Since the K7M review a plethora of gold-finger devices have burst onto the market so you can put your soldering iron away - multiplier-based Athlon overclocking is now a no-brainer - but stable FSB adjustment gives you much finer control (once you've set your multiplier to something higher), a few extra MHz and, of course, the advantages of the overclocked memory, PCI, AGP etc. I've been running the Athlon 650 at 715MHz (110 FSB) for a couple of days now and it hasn't blinked once. In fact, this board has not had a single error during testing - only when I tried to run the KingMax at CAS2 @ 148MHz did it complain - certainly not the fault of the board.

Performance
Ok, onto the meat of the matter. The test configuration was as follows:

Athlon 500 fitted with GlobalWin VEK-32 cooler OR
Athlon 650 fitted with GlobalWin FKK-50 (modified as per here) cooler
KingMax PC133 RAM
Diamond Viper V770U at default clock in AGP 2X mode
Leadman LP6100 400W Athlon-approved PSU

I tested the both boards with both processors, and on the Epox I ran an extra set of tests with the memory running at 133MHz. This is a feature of the asynchronous memory bus of the KX133 chipset so of course the K7M can't do it. If you feel this is somehow unfair then ignore the marked 133MHz benchmark - but if I was using this board I would configure it like this for sure.

First up, a synthetic benchmark in the form of Sisoft's Sandra. CPU benchmarks first:

Sandra CPU Benchmark

Well, pretty unsurprisingly the board and memory has zero effect on this benchmark, which is good. You would not expect any change in the speed of the CPU core, L2 cache etc. Next up is the memory benchmark which shows a slightly different story:

Sandra Memory Benchmark

These results are a little ominous. You can see the Epox lagging behind the older K7M at the same memory speed. Only when the Epox "cheats" by using the memory at 133MHz does it beat the K7M. However, keep in mind that the current version of Sandra has (according to the website) "Preliminary AMD Athlon (K7) chip support only!" so these results may not be entirely accurate. I thought it important to include them, though - anyone familiar with my previous motherboard reviews would question the lack of Sandra benchmarks and I think it's still a valuable tool. Besides, we're comparing apples with apples, so you'd think any Athlon-specific oddity in Sandra would affect both boards. Anyway.

Let's see how they stack up in a more real-worldly benchmark, MadOnion.com's 3DMark2000. This is a punishing test of video, AGP bandwidth, memory and CPU and is designed to approximate many modern gaming environments. You Q3-heads, driving-sim addicts and wannabe Top-Gun's pay attention:

That big graph might take you a second or two to get your head around, but basically it shows two major points. Firstly, at the same clock speed, the Epox is definitely faster than the K7M. Presumably this is because of the different chipsets. Secondly, the asynchronous memory feature of the Epox makes quite a difference to performance. You can see in both 1024x768 tests the Athlon 500 in the Epox with 133MHz ram was faster than the Athlon 650 in the K7M!

Conclusions
Well, I said in the beginning of this review that poor motherboards were keeping me from adding an Athlon system to my stable. I've been telling my friends to wait for the KX133 boards and I feel well justified for that attitude now. If this Epox is an respresentative of the KX133 boards from the other manufacturers then the Athlon's time has truly come at last. However, off the top of my head I can't think of many ways they could improve on the Epox except perhaps a wider/higher selection of FSB's and voltage adjust in the BIOS. The board's layout is excellent for cooling, it has all the modern features and standards, it has shown itself to be rock-stable and speed-wise it's a very impressive performer. The Epox EP-7KXA has now become my recommended motherboard for the AMD Athlon processor.