Related Articles

Overclocking

This board is extremely overclocking friendly.

Even using fairly aggressive memory timings, our P4-560 immediately ran at default voltage settings at a multiplier of 14 with a 1066Mhz FSB - running the CPU at 3.73Ghz.

The memory bandwidth showed an impressive 30% gain in moving from a 3.6Ghz CPU clock with 800Mhz FSB to a 3.73Mhz CPU clock with a 1066Mhz FSB. Given that the processor was only overclocked by 3.6%, and the bus by 33.3%, it is safe to say that the vast majority of the increased performance for Sandra was due to the higher bus bandwidth.

Frankly, that was one of the easiest and productive overclock I've done recently.

A good benchmark to use for testing the effects of overclocking is WinRAR. We normally use it for processor reviews; however I decided to also use it in this article, and the future overclocking article, as it gives a good indication of the increase in performance for programs that are a mixture of processor speed and memory bandwidth bound.

Now with such an easy and successful overclock, I could hardly stop there, could I?

How far did we get?

How does 1084Mhz FSB with 600Mhz memory running at 3-2-2-5-8-1T strike you?

(We did increase CPU Vcore to 1.45V and memory supply voltage by 0.1V to get to 1084Mhz/1T timing)

Its interesting to note that with the 800Mhz FSB Sandra showed "only" 4.9Gb/sec memory bandwidth; an overly optimisting and simplistic calculation shows that there is theoretically close to 6.4Gb/sec bandwidth available on a 64 bit 800Mhz P4 FSB, and close to 8.5Gb/sec bandwidth on a 1066Mhz FSB.

What's even more interesting, is that with a 1066 (and later 1084) Mhz FSB we got a very impressive increase in memory bandwidth, and the Gigabyte Doom 3 benchmarks became competitive with the test results for the other motherboards we were comparing it to.

Conclusion

The Gigabyte 8N-SLI is an excellent all-around performer, and its overclocking capabilities were outstanding.

I had so much fun tweaking the system that you will likely soon see a followup article on tweaking the 8N-SLI here on Neoseeker. The 1066 Mhz front side bus really helps make the nForce4 shine; the memory bandwidth in Sandra increased by almost 40% with careful tweaking of the memory parameters and the bus speed.

The gaming performance was about 10% lower for DOOM3 and Halo than other nForce boards; I am convinced that the problem lies with conservative chipset settings. The bad news is that I was not able to adjust them in the BIOS; the good news is that the board appears to be an excellent board for overclocking, and DOOM3 showed much better performance when the system was pushed. (We'll cover how much better when we do our followup overclocking article.)

The only negative thing I can find to say about this board was that we had problems with clearing its CMOS after changing processors during one of our experiments; even the "CLEAR CMOS" jumper did not help. It took disconnecting all cables, clearing the CMOS and discharging some capacitors before the board would post again. Now this may not be typical of all of the ROYAL boards, we could just have been unlucky, and as I said, overall I am extremely happy with how the 8N-SLI performs.

Let's put it this way. If I wanted to build a P4 system in the near future, odds are I'd use a Gigabyte 8N-SLI ROYAL thanks to its performance, features, and overclockability.