CPU Overclocking

Our open testbed was built with the following components:

• Gigabyte GA-EP35-DS4 mainboard, rev. 2.1, BIOS F3;
• 2 x 1024MB Corsair Dominator TWIN2X2048-9136C5D;
• NVIDIA GeForce 8800 GTS 320MB graphics card;
• Seagate Barracuda 7200.10 HDD, ST3320620AS, 7200RPM, 16MB, SATA 320GB;
• Zalman CNPS9700 LED CPU cooler;
• Antec NeoPower HE 550 PSU (550W).

The first part of our overclocking experiments on Gigabyte GA-EP35-DS4 mainboard was performed with an Intel Core 2 Duo E8400 CPU (3.0GHz, FSB 333MHz, 6MB, Wolfdale, rev. C0). This processor can be overclocked to 4.1GHz frequency with Vcore increase and FSB set at 455MHz. Of course, any good mainboard can cope with this task. That is why we decided to lower relatively high clock frequency multiplier of this CPU – x9 – and try to find out the maximum bus frequency for our mainboard to remain stable. We know for a fact that this processor sample will not limit our overclocking up until 540MHz FSB.

This time the processor clock frequency multiplier was set to the minimal value of x6, and so was the memory frequency. We did it to ensure that neither CPU nor the memory would become a bottleneck and prevent us from finding out what our mainboard was actually capable of. CPU Voltage Control was increased to 1.3V, DDR2 OverVoltage Control was set at 2.1V, FSB OverVoltage Control and MCH OverVoltage Control were set as +0.2V.

We tried starting at 520MHz FSB wit these settings and succeeded. We decided to check the stability at this sped later and proceeded to conquering 530MHz FSB. However, in vain, as the mainboard started but couldn’t boot Windows. Increasing the voltage of the chipset North Bridge and FSB bus by +0.35V didn’t help hit even 525MHz. But we were really puzzled to discover that when we got back to 520MHz FSB, the board wasn’t working at this speed either!

Looks like the mainboard lost its ability to boot Windows OS at 520MHz FSB once we messed with FSB OverVoltage Control and MCH OverVoltage Control parameters. So which one is the one to blame? Further experiments revealed that increasing one or both these parameters by +0.3V or higher was very harmful for overclocking: the mainboard immediately lost stability. The maximum harmless value for both these parameters is +0.25V.

At the same time we discovered that Gigabyte didn’t cure one of their older issues: the mainboard would still reset all parameters without notifying the user in case of over-overclocking. We didn’t have to Clear CMOS even once, the board didn’t hang or freeze, didn’t loop on boot-up. The protection against over-overclocking worked perfectly well: every time the board detected over-overclocking it immediately rebooted with the FSB and memory settings dropped to their nominal values. Moreover, we didn’t have to waste our time trying to reboot with non-operational settings for a few additional times (usually, three times): parameters were reset immediately after the first failed boot-up.

Well, things seem to be perfect, but there is one unpleasant issue. Gigabyte mainboards do not notify you that the FSB and memory frequencies have been reset to their nominal values and just continue booting. It would be much better if the boards didn’t act willfully, but stopped, informed the user of reset settings and waited for the user’s response, like other manufacturers’ mainboards do. The user may continue booting with the default settings, or access BIOS Setup and correct the reset parameters. I assure Gigabyte engineers that in most cases the users will prefer the latter option, and not the former one.

So, Gigabyte GA-EP35-DS4 can boot the OS at 520MHz FSB with the processor clock multiplier lowered to x6. However, this achievement is of pure theoretical interest to us. The most important thing is to find all possible multiplier and bus frequency combinations when the mainboard can guarantee maximum overclocking and stable functioning of our processor at 4.1GHz.

As a result, Gigabyte GA-EP35-DS4 mainboard coped very well with overclocking our Intel Core 2 Duo E8400 to the maximum possible frequency of 4.1GHz with the default x9 multiplier. The FSB frequency in this case was raised to 455MHz, and the voltage of the chipset North Bridge and FSB were increased by +0.25V. With lower FSB OverVoltage Control and MCH OverVoltage Control settings the mainboard would often reset all parameters as we have just described above.

We encountered no problems when we reduced the clock multiplier to x8.5: the board remained stable at 483MHz FSB. However once we dropped it to x8, first problems occurred. To hit 4.1GHz frequency with this multiplier we have to increase the bus speed to 512-513MHz. However, Gigabyte GA-EP35-DS4 mainboard could only work stably at 500MHz, not any higher. Not too good for a CPU that can work at 540MHz FSB.

Gigabyte GA-EP35-DS4 also didn’t impress us with overclocking results for the youngest quad-core 45nm CPU - Intel Core 2 Quad Q9300 (2.5GHz, FSB 333MHz, 6MB, Yorkfield, rev. M1). We could only achieve stability at 455MHz FSB. This is not bad of course, but still pretty average. Frankly speaking, we expected Gigabyte GA-EP35-DS4 to do much better than that.

There are a few possible reasons for not very high overclocking potential revealed by Gigabyte GA-EP35-DS4 mainboard. The most widely spread and the most common one is BIOS imperfection. We did use the latest available BIOS version at the time of tests – BIOS F3, which is one of the first ones. So hopefully they will address the issues in the upcoming BIOS revisions. It is possible, but we should keep in mind that although Gigabyte GA-EP35-DS4 is a new mainboard, it uses 2.1 PCB revision and the long-term experience of the same boards without the “E” in the model name. It seems more probable that this board is simply not very good at overclocking 45nm CPUs that appeared not that long ago.

Of course, we could also hold Silent-Pipe chipset cooling system accountable for our not very impressive overclocking experience, because it fails to provide sufficient cooling for the chipset North Bridge once its voltage increases more than by +0.25V. I would like to remind you that this increase leaves the NB voltage at a pretty high level of 1.5V. Take, for example, abit IP35 pro mainboard that is great for CPU overclocking. We only had to increase this voltage setting to 1.47V on abit board to ensure that our Intel Core 2 Quad Q9300 would run stably at 475MHz bus frequency.

I don’t know which of the above described reasons the right one is. It could even be not the best mainboard sample that we got. Please get me right: Gigabyte GA-EP35-DS4 is a very nice board to work with, setting overclocking parameters is easy and convenient. Quiet resetting of the parameters in case of over-overclocking is a minor inconvenience. And the board does overclock CPUs just fine, although not impressively.