The 8N-SLI Royal has all the features we've come to expect from high end nForce boards; lots of I/O, decent audio, good memory bandwidth and an easy to tweak BIOS - however I must admit that I did not expect it to be as overclockable as it turned out to be. The 1066Mhz front side bus is a feature that we are going to come to expect from all future boards. I liked the clean layout, and the color coding is actually useful.

As I discussed in the review, the 8N-SLI was a surprisingly good over clocking board. It was the first board we saw that just worked - out of the box - with a 1066 MHz FSB; and this was with a P4 560 processor that is spec'd for an 800Mhz FSB. I had so much fun tweaking that board during the review that I promised a followup tweaking article - and here it is!

First, some background on tweaking systems:

The first thing you have to realize is that for significant overclocking you will likely have to raise at least the processor Vcore voltage; and may also need to raise the chipset and DRAM voltages.

Please realize that you are risking your motherboard and processor by doing these more extreme overclocking experiments. When I push matters, I am always careful to stay reasonably close to the manufacturer suggested default voltages.

Personally I have yet to experience problems when exceeding the default core or DRAM voltages by up to 10%, but you may not be so lucky. Please don't forget that the amount of heat generated goes up drastically with increased voltages and clock speeds.

Since the board was rock solid at 1066MHz, I went for the gold - a 1200MHz FSB. I really did not expect this to work, as it would run the processor at 4.2GHz (1200MHz FSB, 14x300MHz CPU). I was right. It did not work. It might work with extreme cooling measures, or if someone has a particularly overclockable P4 - this speed really could damage your processor or motherboard.

(In case you are wondering why I tried for a really high speed overclock first, then for the rest of the article I went incrementally - it is because I wanted to show you charts interspersed throughout the article)

After reading so far, you deserve to see some of the results:

(The bars are labeled as FSB_MHz DDR_MHz / mem_timing therefore 1120 620 / 3225-1T means that the processor FSB was at 1120MHz, DDR2 was at 620MHz, and we were using 3-2-2-5-1T memory timings)

I like Sandra for a quick memory bandwidth test.

Looking at this chart it is immediately obvious that the default settings produce very poor results compared to what can be achieved by overclocking and tweaking the system, as we get a 30% to 43% increase in memory read speeds with our tweaked results. This proves that there are still good performance gains to be had by tweaking your system.

If we consider the 1066/666/4337-2T as the baseline for comparison - as we might, given that the 1066MHz FSB with DDR2 memory is going to be the next Intel high end platform - we see that by tweaking we can get a further 1% to 9% increase in memory throughput. Now a 1% improvement is basically insignificant, but we can't say that about 9% as that would definitely provide a noticeable boost to memory bandwidth bound applications.

The 1084/600/3225-1T setting jumps out as being a desirable goal, as it provides a 38% increase over the "stock" settings for a 560 processor with DDR2 memory, and while we did not get it 100% stable at this time it seems like with more agressive cooling techniques it might be possible to be stable at this setting. We would also likely have a better chance of getting this FSB setting (or even a higher one) stable with a processor that is rated to run at a 1066MHz FSB - we must remember that we were testing with a processor rated for only 800MHz FSB.