VIA's KT133A - Overclocking w/ the 133MHz FSB: The "Issue"

The Problem

After reading the emails we received, as well as the messages that you guys posted in our forums, the problem seems to occur when you are dealing with AMD chips that are designed to run at a 100MHz (200MHz DDR) FSB.  More specifically, it seems to happen only on boards that have a 100/133MHz jumper.

Generalized from the information we gathered, we’ve come up with a theory that explains what may be happening.  We will use an example to illustrate our theory. Take an Athlon “B” 1GHz that’s designed to run at 100MHz FSB and mate it with a KT133A motherboard featuring a 100/133MHz jumper.  We’ll also assume that the multiplier settings are implemented in the BIOS, as is the case for most KT133A boards.  The chip has a clock multiplier of 10, so  you put it on the motherboard, set the 100/133 jumper to 100MHz , and boot up at 1000MHz, and everything runs great.

However, in order to take advantage of the 133MHz FSB offered by the KT133A and get a free performance boost, you may want to lower the multiplier to 7.5 and raise the FSB speed to 133MHz, yielding the same 1GHz speed for which the chip was intended.

Two steps need to be completed in order to accomplish this - increase the FSB and lower the multiplier. Of course, these have to be one at a time, but the order definitely matters.  

Obviously, increasing the FSB speed to 133MHz first poses a problem because the board will try to boot at 10 x 133MHz = 1.33GHz before you can even enter the BIOS to lower the multiplier. 

So the correct approach, in theory, would be to first lower the multiplier to 7.5, and then raise the FSB to 133MHz.  So you boot up 10 x 100MHz, then lower the multiplier to 7.5,  and then boot again at 750MHz.  Now you shut down the system, change the jumper from 100MHz to 133MHz, and finally boot up at what should still be 1000MHz, but now at 7.5 x 133MHz.

Unfortunately, things did not go so smoothly for many people.  Whatyou may have found is that the system would never POST under such conditions. 

In order to confirm the problem, we took an Athlon 800MHz (8 x 100MHz) that does not run at 1GHz (10 x 100MHz).  We used two KT133A motherboards that have the 100/133 jumper, the EPoX 8KTA3 and the Iwill KK266, as our test beds.  With the jumper in the 100MHz position the systems booted up at 800MHz without any problem.  We then lowered the multiplier ratio to 6 and the systems booted up at 600MHz without a hitch as well.  However, when we then shut down the systems and changed the jumper to 133MHz, both systems failed to POST, despite the fact that the CPU should run fine at 6 x 133MHz = 800MHz.  What’s going on here? Overclocking the FSB of a CPU has never been limited by the CPU itself, but rather by the chipset or multiplier associated with the CPU.