Overclocking Specifics

Our overclocking experiments with EVGA X58 SLI LE mainboard got off to a flying start. I have to say right away that it ended also on a positive note, although a lot of questions arose in the process.

Let’s start with the memory subsystem, is it capable of working at its nominal frequency of 1867 MHz on EVGA X58 SLI LE? Yes, no problem. Just select the X.M.P. (Extended Memory Profile) profile in the BIOS corresponding to the desired frequency. Although, this is when we uncovered the first issue. Most BIOS parameters such as frequencies, timings and voltages are set to “Auto” by default. The board knows how to react to our actions: when we change the memory frequency, the Uncore bus frequency will be changed accordingly, when we select an X.M.P. profile or enable automatic overclocking function, the voltages will be bumped up as well. However, we don’t see the expected results to all these actions. Almost all parameters remain at Auto and only after rebooting the system we can see the result of our changes. Not the most convenient implementation.

Once we restarted the system we found out that if we select the first X.M.P. profile, EVGA X58 SLI LE mainboard, just like all other boards supporting Extended Memory Profile technology increases the memory voltage to 1.65 V and adds 300 mV to UnCore bus voltage (CPU VTT Voltage), which brigns this parameter to 1.5 V. we have no problems with the memory voltage, as it is the maximum allowed setting for DDR3 SDRAM on an LGA1366 platform and a nominal setting for our specific memory modules. However, 1.5 V for UnCore bus is way too much. By raising this voltage we push the CPU temperature higher up, so the lower it is the better. We don’t know a single mainboard that required this voltage setting to be at 1.5 V. As a rule, you have to raise CPU VTT Voltage to 1.35 V, but EVGA X58 SLI LE required 1.425 V. It is not a tragedy, but a little too high.

After that we lowered the processor clock frequency multiplier to the minimal 12x and find the highest base frequency, when the board remains stable. 215 MHz is a good result, which is typical of most mainboards suitable for successful CPU overclocking.

This is when we uncovered another unpleasant peculiarity about the BIOS of EVGA X58 SLI LE mainboard: we can’t control Intel Turbo Boost technology. We can’t disable it and if it is disabled, we can’t turn it back on. Formally, there is a parameter in the “CPU Feature” sub-section called “Turbo Mode Function”. By default the technology is enabled and we can change this setting from Enabled to Disabled. However, when we return to this sub-section we will still see the Enabled setting in place. The opposite is also true. As we have already said, EVGA X58 SLI LE mainboard, just like other good overclocking mainboards knows how to adequately react to our actions. We lowered the processor clock multiplier in order to find out the maximum operational base frequency. In this case the board disables Intel Turbo Boost technology by itself, which is an absolutely correct action in this situation. However, after that, when we return the multiplier to its nominal value, we can’t enable Intel Turbo Boost technology anymore, just like we couldn’t disable it previously.

Luckily, after we rebooted the system, this technology appeared turned back on, but I have to admit that this incident upset us a lot and had a serious negative effect on initially great impression from EVGA X58 SLI LE mainboard, because the problem appeared to be not only with the “Turbo Mode Function” parameter, but with the entire BIOS. We saw very similar things when we tried to use BIOS settings profiles. When we loaded the previously saved profile, we noticed that not all the settings corresponded to what they were supposed to be. After system reboot, all settings magically turned out exactly what we needed. It turns out that the BIOS is sometimes simply unable to display the correct information and reflect the actual state of things.

We know that even minor BIOS changes may seriously affect the resulting performance and overall system stability. Imagine how significant the effect from Intel Turbo Boost is, which adjusts the CPU clock frequency multiplier. We take any BIOS changes very responsibly and want to be certain that our actions will produce an anticipated result. To our great disappointment, EVGA X58 SLI LE mainboard doesn’t give us this chance, it is unreliable and unpredictable. It may start or not start, the setting may increase or not increase, a function may work or not work… How can we trust a board in this case? I don’t think we can, which is sad.

But let’s get back to our CPU overclocking experience. We already know that the board can work fine at pretty high base clock settings and we also know that it can have the memory working at higher frequencies, too. Let’s try and find the most optimal combination of processor and memory overclocking. We increase the base frequency to 181 MHz, set the memory frequency at 1810 MHz and timings at 8-8-8-22-1T. The system passes the stability test successfully.

Unfortunately, we one more time face a widespread problem: increase in the CPU Vcore under heavy load. The screenshot above from Lavalys Everest CPUID shows 1.305 V, while the nominal setting for our particular processor is 1.225 V. However, we didn’t increase it in the BIOS, but just enabled protection against voltage drop. EVGA X58 SLI LE mainboard has formally passed the stability tests in this case, but remember that besides increased Vcore we also had to increase CPU VTT Voltage more than usual to ensure that the memory would work at high frequency with low timings. As a result, after 8 test cycles in LinX utility the CPU core temperature reached 93 °C. During long-term stability tests the total system power consumption got as high as 400 W. Of course, if the system case is not properly cooled, or if the room temperature increases, we won’t be able to guarantee stability during this overclocking.

It is always bad to excessively increase the voltages, but EVGA X58 SLI Classified mainboard managed to turn this issue into an advantage. As you may remember, without formally increasing the CPU core voltage, i.e. with all Intel processor power-saving technologies up and running, we managed to increase the base clock to 186 MHz and hence push CPU frequency to 3.9 GHz. Unfortunately, it didn’t work for EVGA X58 SLI LE and we had to stop at 181 MHz base clock. However, all power-saving technologies remained intact lowering the voltage as well as the multiplier in idle mode.

And why don’t we try and lock the processor core voltage manually in the mainboard BIOS at the nominal setting of 1.225 V? Yes, in this case Intel processor power-saving technologies will be partially disabled and the CPU Vcore will no longer drop in idle mode. However, it won’t increase that much under heavy processor load, which will offer us better thermal conditions for processor cores. Our attempt, however, led to very unexpected and I would even say paradoxical results. The processor core voltage in idle mode turned out much higher than 1.07-1.09 V when all power-saving technologies kicked in. Now it equaled 1.15 V, which is still way below the BIOS setting anyway.

When we got down to stability tests and loaded the CPU, the voltage increased, but not as much as before, when the mainboard was in control and VDroop was enabled.

In reality, things were even worse than before, because the maximum registered CPU Vcore was 1.32 V vs 1.31 V with Auto setting and VDroop enabled. However, the core voltage increased to 1.32 V only between the test cycles for a few seconds when the load was taken off. On average it varied somewhere between 1.235 and 1.247 V, which is not that far from the nominal 1.225 V. However, when voltage protection was enabled, Vcore was constantly equal to 1.29-1.31 V under heavy load. The outcome is quite logical: not only all tests were passed successfully, but the core temperature after the same eight runs of LinX utility reached only 83 °C instead of 93 °C as before.

I don’t think I can explain why EVGA X58 SLI LE behaved the way it did. Logically, when the processor core voltage was set a fixed value in the BIOS, Intel processor power-saving technologies had to partially stop working and it had to remain constant all the time. In reality we see some kind of EIST imitation: pretty low voltage in idle mode (although still higher than with power-saving technologies working normally), which somehow increases under heavy load. Does it mean that despite the voltage changes Intel processor power-saving technologies remain up and running? Or maybe it is the result of another BIOS bug: we disabled VDroop, but in reality the protection continued to work? We don’t know for sure, but the result is really nice: thanks to this mysterious feature or drawback successfully turned to the board’s advantage, we managed to overclock our CPU and memory with acceptable power consumption and processor thermal readings. However, we are going to talk more about power consumption later on and now let’s sum everything related to EVGA X58 SLI LE overclocking in the following table: