To the common eye, traces on a motherboard run amok, twisting their way through fibreglass and copper traces. Some decidely snake around after running about in circles, others follow a straight path with an occasional turn to facilitate the data highway.

In truth, motherboard design is far more complicated than it seems - designers could throw the circuit at a PCB layout program, and force it to lay the traces automatically. Chances are, the board would not work as intended. Motherboard design is an Art, which probably explains why this Fengshui Motherboard will never make it past POST (if one was ever made).

When copper traces are laid side by side, various things can occur at once. Magnetic and capacitative coupling immediately come to mind. What about the number of layers of PCB? Lengthy traces and resistive routes? At high frequencies that components run today, ringing is taken care of through impedance matching while noise is attenuated via differential signalling and grounding techniques. Signalling at high frequencies remain an engineering challenge, and the most successful to date, has to be the Gunning Transceiver Logic (GTL) invented by then Xerox researcher - William Gunning.

Signalling - GTL

GTL signalling on current generation Intel platform are no longer the original, which was specified for operation between tens to hundreds of MHz - insufficient by today's standards. Simply said, the GTL signalling system sets a "window" for logic circuits to determine if the tranceivers are reading a logic high, or logic low. By virtue of the innocent inputs/outputs of chipsets, overshoots and undershoots of these signals are often misread. These misreads are what we know as errors, that is, corrupt data.

When overclocking, the frequencies of signals within the system are raised. These signals, with their increasing state-of-change, interacts with the 3 basic electrical characteristics of resistance, capacitance and inductance. This, is what we know of as reactance. Due to imperfections (design tolerance, thermal degradation of components during wave soldering, trace layout, grounding planes, impurities of the PCB substrate, inconsistent PCB thickness, flux, termination voltage fluctuations etc.), ringing signals can arise from overclocking. Ringing, as they will tell you in engineering school - is not good at all for signal integrity.

As an overclocker, competitive or otherwise, moral integrity is paramount - no cheating, no timer tricks, no pixel-nulling. The next important integrity, would be that of the signals that tick your clocks. Earlier on, we had mentioned that reactance on inputs/outputs, with increased frequencies causes ringing, which overshoots and undershoots on the transceiving pin of the chipsets. The "window" of the I/Os get confused and writes in the wrong logic highs and lows. GTLref is a termination voltage (a termination voltage sets the reference point to which a chipset I/O recognises the incoming signal as a logic high or logic low) engaged by Intel in existing systems which sets the "window" for signalling.

Voltage settings used to achieve results on the next page.

Through manual adjustments of the GTLref, users can control the inadvertent ringing resulting from overclocking. On the Striker II Extreme, you'd see the real-world use of GTLref control for overclocking. Not one, but all four GTLref voltages can be individually adjusted on the CPU. Knowing that each core on the die responds uniquely to voltages and cooling, individual GTLref adjustment is the key to acheiving high FSB frequencies. Our last record on the QX9650 was 456MHz on the FSB, let's see what those new options will do for it!