All our tests were run in an open testbed with the following configuration:
- Mainboard: EVGA nForce 750i SLI FTW, BIOS SZ1B;
- Intel Core 2 Duo E8400 (3.0GHz, 333MHz FSB, 6MB, Wolfdale, rev. C0);
- Intel Core 2 Quad Q9300 (2.5GHz, 333MHz FSB, 6MB, Yorkfield, rev. M1);
- Memory: 2x1024MB Corsair Dominator TWIN2X2048-9136C5D;
- Graphics card: NVIDIA GeForce 8800 GTS 320MB;
- HDD: Seagate Barracuda 7200.10 (ST3320620AS) 7200RPM, 16MB, SATA 320GB;
- CPU cooler: Zalman CNPS9700 LED;
- PSU: Antec NeoPower HE 550 (550W).
First I have to remind you of a few peculiarities of Nvidia based reference mainboards that actually were the reason we had to retest EVGA nForce 750i SLI FTW mainboard. If you try to achieve higher FSB speeds, for instance, by lowering the processor clock frequency multiplier in the BIOS of an Intel based mainboard, the board will work with the multiplier you set. Pretty logical, isn’t it? We actually expected mainboards on Nvidia chipsets to act the same logical way. However, it turned out that in this case the mainboard will only start with the changed clock frequency multiplier. In Windows the multiplier will be lowered in to the minimal x6 in idle mode and returned back to its nominal value in case of higher CPU utilization. To actually lock the multiplier at the desired value, you have not just to set this value in the mainboard BIOS, but also disable the Enhanced Intel SpeedStep (EIST) and C1E technologies.
Note that we don’t call this a bug. It is no bug, but a peculiarity that you should keep in mind and use to your advantage. It actually gives the users of Nvidia based mainboards more freedom. If you lower the clock multiplier and disable only EIST (C1E will remain active), the selected multiplier value will be the maximum, while in idle mode it will keep dropping to the minimal x6. On Intel based mainboards, things are much simpler: if you lowered the multiplier, it will be what you set it to be, without dropping to any minimal values.
We didn’t take into account this peculiarity when we were testing EVGA nForce 750i SLI FTW for the first time. To find out the maximum FSB frequency this board can handle, I did everything as usual: lowered the processor clock frequency multiplier to the minimal setting of x6, increased the FSB and NB voltages and tried to boot at a relatively low frequency of 450MHz. Our Intel Core 2 Duo E8400 processor is not the best overclocking choice: it remains stable only at frequencies no higher than 370MHz FSB if its Vcore stays at nominal 1.225V. The CPU can overclock to 450-455MHz FSB with its nominal multiplier but Vcore needs to be increased to 1.55V. Of course, the system wouldn’t boot, because the processor clock multiplier increased to the nominal x9, but we didn’t increase the Vcore, because we knew we set the multiplier at x6 in the BIOS. As a result, we could only boot the operating system at 425MHz FSB that’s why we were not very happy about our experience with EVGA nForce 750i SLI FTW mainboard.
Now that I knew about this peculiarity of boards on Nvidia chipsets, I not only lowered the processor clock frequency multiplier and increased the voltage, but also disabled EIST in the mainboard BIOS. As I have expected, the EVGA nForce 750i SLI FTW mainboard proved much more overclocking-friendly, because it booted the OS at 520MHz FSB. In order to check out the mainboard’s performance at high FSB frequencies and with maximum CPU overclocking, we increased the processor clock frequency multiplier to x8, raised Vcore to 1.55V and set the bus frequency at 512.5MHz (2050MHz in quadrupled units). Once we increased the FSB voltage to 1.35V and the chipset North Bridge voltage to 1.5V, the system passed all stability tests.
After this evident success we definitely need to apologize to EVGA for the first wrong conclusions about their mainboard. So we continued the practical experiments with our EVGA nForce 750i SLI FTW.