DFI LANParty UT ICFX3200 T2R/G

Overclocking Settings to consider:

It's true that the BIOS needs an education in itself in order to use it to it's full potential. The key areas of the BIOS to consider when overclocking the board are as follows:

• Burst Rate: 4/8, with 8 offering more performance but 4 is likely necessary when clocking the memory high.
  
  
• Command Rate: Three is unnecessary; two should be ample for all clocks. DFI have stated that Micron D9 chips and particularly highlighted GSkill PC6400 as being capable of 1T.
  
  
• Memory Drive Strengths: These are important in helping to obtain 1T and higher clocks, but too high will cause errors. 0 is low, 15 is high. Stock is 5, and we found higher than 13 caused instability.
  
  
• ADDR/CMD Driving Pre-AMP: Is essential to be set to disabled for 1T.
  
  
• CAS Timing (tCAS); Read RAS to CAS Delay (tRCDR); Write RAS to CAS Delay (tRCDW); RAS Precharge (tRP); RAS Cycle Time (tRAS); Row Cycle Time (tRC): These values are your most critical and usually found on many other motherboards. They are the “quoted” memory timings like 3-3-3-9-18. Obviously depending on the memory you buy gives you the ability to run lower timings. It’s been quoted that tRCDW should possibly be one less than tRCDR for better performance, and tRAS should be 25 max. tRAS and tRC have a large effect on bandwidth at the cost of stability, so try to keep pushing as low as possible.
  
  
• tWR: Try aiming for 3 for high bandwidth, however we only managed to obtain 4.
  
  
• FSB IO Strength: 4xN and PMOS are the most important and it’s been suggested trying +/-3 either way. FSB is CPU to northbridge remember, so try this should not affect memory speeds. Try combining with GTLRef voltage adjustments.
  
  
• CPU VID is the core CPU voltage, and obviously increasing this means a higher clock should be available, but with extra heat.
  
  
• CPU Special ADD is an additional fixed voltage to the VID: where 121.25% x 1.6V gives you a maximum of 1.94V Vcore. How do you want your CPU? Crispy, fried or both?
  
  
• CPU VTT 1.2V: This is important for FSB overclocking and should be used in conjunction with increasing the northbridge voltage and GTLRef Voltages.
  
  
• DDR2 Memory Voltage: Over 2.5V you better have a good warranty with your memory, along with some great cooling. We run 2.3V in the lab with Corsair 8500C5, but 2.4-2.5V can used if the quality of your memory chips are good enough so do some research beforehand.
  
  
• NB PLL 1.2V and 1.8V: Increasing also helps with FSB overclocking.
  
  
• NB 1.2V Convert from: Raising the convert from voltage increases efficiency and stability of the “1.2V” voltage, but remember if you increases the NB 1.2V voltage above for PLL/Core/PCI-E then you will need to increase this again.
  
  
• NB PCI-E 1.2V: Useful for overclocking graphics cards and the PCI-E bus.
  
  
• SB Core voltage: This only needs to be adjusted a little. Since the northbridge-southbridge link is PCI-Express based, increasing the speed of the PCI-Express frequency means there’s more bandwidth available for components linked to the southbridge and potentially more overall performance.
  
  
• Clockgen voltage: Leave set at 3.4V and enable Software reset as well to enable a better chance of recovery from a bad overclock.
  
  
• GTLRef Voltage (Gunning Transceiver Logic: doesn't that sound cool!) is the reference voltage the CPU uses to determine high or low signals (on or off) for data on the front side bus (FSB). The standard settings are set to “optimum” by DFI, but try adding to it in small quantities if you hit a wall with everything else. Kentsfield Quad Core CPUs are supported and require the four GTLRef voltages which DFI have integrated in 1&2 and 3&4. Pay particular attention to these settings if you're overclocking a Quad Core CPU.
  
  The typical GTL switches between 0.4V and 1.2V when sending data (off and on), so increasing the value means the signal to noise ratio is stricter: allowing less difference between the data signal and that of the increased noise from increased voltage along the FSB traces between northbridge and CPU. If the standard GTLRef is 0.8V and is increased to 0.9V it means there's a greater difference between the stock 0.4V and reference voltage. But typically when overclocking the 1.2V VTT signal voltage will be increased (Intel rates 1.55V as an upper limit to VTT for it's processors), increasing the upper limit too. There is a greater overall difference allowing for more noise and a stricter data edge on the clock when increasing the frequency: 0.4V (off) - 0.9V (Ref) - 1.3V (on). It’s all about obtaining a balance between the two to filter the actual signals propagating through.
  
  The CPU Core GTLRef values range from 1 to 255, with each value representing approximately 0.00153V, so the stock value of 48 represents a cut off voltage of 0.70V, out of typically 1.21V. Increasing this offers values of typically: 72 = 0.74V, 96 = 0.77V, 120 = 0.81V.
  
  The Northbridge GTLRef values also range from 1 to 255 but each value represents 0.00146V, so a stock value of 128 = 0.86V. Increasing this offers values of typically: 144 = 0.88V, 160 = 0.91V, 176 = 0.93V.

Try here if you want to understand a greater depth of GTLRef Voltage, beyond the scope of this article.

Also kudos to the guide produced by Tony at OCZ.