Intel Core 2 Extreme QX6700 CPU Overclocking
Our article called Multi-Core Confrontation: Core 2 Quad Q6600 vs. Core 2 Duo E6850 proved that even today, when there are very few applications that would use more than two cores, quad-core processors make a lot of sense, despite the first impression you might get. Of course, it will take some time before quad-core CPUs become mass, they are still the prerogative of computer enthusiasts, however, we cannot ignore the constantly growing group of happy quad-core owners. Overclocking a quad-core processor is formally the same procedure as in case of a single-or dual-core CPU, although the mainboard, cooling system and especially the processor voltage regulator circuitry get loaded much heavier in this case. That is why we got an Intel Core 2 Extreme QX6700 engineering sample on Kentsfield core B3 revision with the nominal voltage of 1.35V. Gigabyte GA-P35-DS3L was the first board to be tested with it.
At first we were going to find out how far this CPU can clock without increasing its Vcore, and the system stability in this case was tested with a 15-minute OCCT run. The system passed this stability tests at 310MHz and 320MHz FSB, at 330MHz FSB the tests was passed but the result was a little doubtful. The countdown timer would freeze every now and then, stop for 10-15 seconds and then catch up for the lost time in a hasty fashion. So, we decided to continue our overclocking experiments with increased voltages.
With the Vcore set at 1.45V the system passed at 340MHz FSB, however, failed to hit 350MHz FSB. Since Core Temp reported 70ºC temperature maximum with average being around 65-66ºC, we increased Vcore further to 1.5V. After that we passed the tests at 350MHz FSB (69-71ºC, 75ºC maximum) and moved on to 360MHz surprised that the temperature hardly increased at all, when we noticed that the CPU utilization was only 77%. But of course, the CPU wasn’t running at the utmost of its potential, so the above mentioned results are of no use to us whatsoever.
The second round of tests started at 330MHz FSB and nominal Vcore setting but in Prime95 utility: although it takes longer to run, it seems to be more reliable. The system stability was tested with Large FFT benchmark that ensures the maximum CPU utilization and heating. We revealed stability issues very quickly. Even at 325MHz frequency one of the cores registered an error after 30 minutes of testing. Well, 3.2GHz at nominal Vcore is not a bad result at all for a quad-core processor. Let’s see what it is capable of with increased core voltage.
The system worked fine at 1.5V for 45 minutes loaded with Large FFT test from Prime95 utility, but the processor temperature balanced around 77-80ºC despite the Zalman CNPS9700 LED cooler working at its maximum fan rotation speed. 1.5V seemed to be too much for our cooler, so we decided to check out the performance at 1.45V maximum. The test was passed successfully, but we kept the system running idle for a while to cool down. However, after a few minutes the mainboard suddenly shut down and refused to boot up again. After clearing CMOS the board would boot, but would shut down again on loading Windows. We replaced the CPU with Intel Core 2 Duo E6300, but nothing changed.
But maybe it was not the mainboard that caused the problem? Maybe the PSU failed workload like that? At the time of tests we didn’t have a suitable PSU at hand, so we decided to see if the abit IP35 Pro would work with the same PSU in the same mode. When trying to disassemble the testbed, we couldn’t remove the four-pin ARX12V power plug from the connector on the board. After some significant effort and additional tools, we discovered that one of the plastic pin linings melted and stuck dead to the mainboard connector.
The PSU turned out OK, abit IP35 Pro mainboard worked fine with it and any processors, but our poor Gigabyte GA-P35-DS3L passed away. It didn’t pass away completely, but turned into a severely crippled platform. It sometimes starts with an Intel Core 2 Duo E6300 in nominal mode, and can work for a while, but in most cases it would shut down on booting the Windows OS. While the board is working, we noticed that the ATX12V power supply connector gets extremely hot, although Intel Core 2 Duo E6300 processor wasn’t overclocked and the connector was a plastic one, not made of copper or even aluminum. We all know that it is important to monitor CPU temperature at all times, we pay additional attention to chipset and MOSFET temperatures since recently, but we have never come across a situation when we needed to check on temperature of plastic power supply connectors.
Eight-pin ATX12V connectors like 24-pin connectors instead of 20-pin ones appeared for a reason. Of course, since you can’t use a large cable, you can distribute the power among several smaller cables. However, it is the first time that we experienced issues like that during quad-core processor overclocking on a mainboard featuring 4-pin ATX12V power connector.
Actually, mainboards very rarely get out of service, even during overclocking. The only one we could remember at the time of this review was Asus P4P800 that dies on us a few years ago after being used as a test platform for a year. The Prescott processors killed it. Gigabyte GA-P35-DS3L box claims that it is optimized for quad-core CPUs and boasts ultra durability, which seems quite doubtful in our case.
Today only Gigabyte GA-P35-DS3L and GA-P35-S3L feature simple processor voltage regulator circuitry, all other models from S3 series feature 6-phase circuitry, and mainboards with the P index and two PCI Express x16 slots also have an 8-pin ATX12V connector. Look at the photos in the beginning of our review. ATX12V connector has two pins for the “ground” and the other two for “12V”, however, most of the power or maybe even all of it went through only one of them that melted in the end. Of course, no one is protected against a defective mainboard, we just hope that it is not a mass issue, but rather a single case of our bad luck.
