Overclocking with Gigabyte P35C-DS3R and Intel Q6600

Overclocking/Overclocking Tests by KeithSuppe @ 2008-03-27

Overclocking successfully requires a mix of the right components, we take a closer look at the pinnacle of affordability and performance product from Gigabyte, the P35C-DS3R, testing with a Quad Core Q6600 we see how we can fly with a water cooled CPU and some spicy DDR2 memory modules.

  • prev
  • next

System Specs / Overcloking

Overclocking


Madshrimps (c)


Overclocking is trial and error. How severely the error part effects the motherboard is determined by the effectiveness of BIOS recovery utilities. These do, in part, what manually clearing the CMOS once did. Ironically clearing CMOS manually is more difficult on modern motherboards as it now requires removing the battery, jumping the pins and replacing the battery. This process is made more difficult since many motherboards have re-located the CMOS jumper and battery just below the PCI-E 16X slot requiring the removal of the video card. Gigabyte uses a simple twin pin design which only requires a jumper if manual CMOS re-setting becomes necessary. When overclocking recovery utilities have become invaluable, think of what it would be like to manually clear CMOS every time your settings didn't "take."

When Gigabyte build the P35C-DS3R they may have underestimated demands 65nm quad core would place on processor power circuitry in the hands of Overclockers. While their supported processors lists the Q6600 (requiring BIOS F4 for compatibility), nowhere does it state; "overclocks Q6600 to such and such a speed...." Nonetheless I could see how Enthusiasts would look at this board and presume a six-stage processor power supply would suffice. Given the panegyric praises sung by Gigabyte in celebration of their Ultra Durable 2 system the presumption is justified. Unfortunately, even with Ultra Durable the best I came up with was an 8x400FSB or 3.20GHz at default Vcore. If your wondering how the highest OC in this test was accomplished under default voltage your beginning to see it was not the processor itself which held us back. Increasing Vcore as high as 1.5V I was able to post at 4.0GHz (9x445FSB) however the system would hang just prior to the splash screen and increasing (G)MCH , PCI-E or FSB voltages did not remedy the scenario. In thumbnails below are several default FSB/CPU compared to overclocked processor configurations. All benchmarks were based on these.

9x266FSB - DDR2-800 @ 800MHz | 8x400FSB - DDR2-800 @ 960MHz

Madshrimps (c) Madshrimps (c)

9x266FSB - DDR3-1600 @ 1066MHz | 9x333FSB - DDR3-1600 @ 1333MHz | 7x400FSB - DDR3-1600 @ 1600MHz

Madshrimps (c) Madshrimps (c) Madshrimps (c)


My philosophy for Overclocking is that there is a difference between overclocking and overvolting, Vcore (or any voltage adjustments) are always the last tool in my arsenal. Besides increasing heat, damaging transistors and shortening the life of the processor, overvolting is often an unnecessary shortcut and its misuse leads to complacency among other bad habits. Many overclockers begin increasing Vcore or VID before they've tried the alternative. During these tests I did use Vcore as a last resort, however; settings where it was "necessary" were not stable regardless. Let’s try and explain this phenomenon.

Initially I attributed problems overclocking the Q6600 to the processor itself. Albeit designed a 65nm process and having a 95W TDP, this is still a power hungry chip. I looked to the VRM circuitry as a potential problem area and my first instinct was to find a Number 2 pencil for a Vdroop mod. However after reading the Anadtech article Overclocking Intel's New 45nm QX9650: The Rules Have Changed everything fell into place. The article brings to light specific Intel processor power regulation circuitry many have overlooked and I had dismissed or at least minimized. The quote below presumes you know something about motherboard CPU power circuitry (if not read the article in full), it explains why Vdroop is a necessary component of Intel processor power regulation.

During periods of high CPU demand, the VRM circuit works hard to supply the current required by the processor. However, as soon as that load is gone, the VRM circuit must act quickly in order to reduce the current supply to the level needed to match the new demand. Because it's impossible for the VRM circuit to respond instantaneously, the larger the load change the greater the maximum potential peak overshoot voltage. Controlling the magnitude of these peak values is critical for maintaining system stability. By positioning the processor's no-load (idle) voltage level higher during periods of light loading, it's possible to sustain a larger negative voltage dip without crossing the processor's lower specified voltage limit. In addition, "drooping" the load voltage as a function of supply current allows the VRM to effectively limit the maximum positive peak overshoot voltage (experienced during a heavy to light load transient) to a value below the maximum allowable CPU voltage. This resulting control system ensures the processor supply voltage, regardless of CPU load, never violates a specified limit... ....source


Voffset and Vdroop work together to protect the processor from exceeding maximum voltage and reduce "settling time" between Idle and Load voltage transients, respectively. These transients are proportional to the load placed on the processor, especially during overclocking. If overvolting your processor seems to increase the instance of Vdroop then Intel power regulation circuitry is doing its job. Under these circumstances a Vdroop mod would do more harm then good. A better resolution to the problem in the case of our P35C-DS3R might be an 8-pin 12V ATX connector with slightly more robust power circuitry. My "purist philosophy" of overclocking aside, I've simply had better results overclocking without overvolting in recent years and in fact in this test I was able to maintain a 9x400FSB OC without the need for Vcore, which was 100% stable. Raising Vcore made matters worse at these settings. In many of my previous articles I not only deter overvolting I suggest overclocking under the AUTO setting as opposed to manually adjusting Vcore.

Over the years I've seen the Vcore value fluctuate in CPU-Z which I always attributed to a software error. In fact those readings were most likely accurate since Vcore should be fluctuating. Looking at Chris Hares Processor Electrical Specification sheet for the Q6600 (CPUID 06FB 95W TDP) Vcore ranges between 0.85V ~ 1.5V. Interestingly the P35C-DS3R low Vcore setting starts at 0.90V. Every Intel processor on that site has Vcore listed in ranges. This is an important distinction for Intel architecture, and I hope this explanation assist some of you out there. Onto our test system.

Madshrimps (c)


Intel Test System:
Processor Intel Q6600 SLACR Socket-775
Mainboard Gigabyte GA-P35C-DS3R
Memory Patriot Viper DDR3-1600 7-7-7-20 (2x1GB)
GEIL DDR2-800 CL4-4-4-12 (2x1GB)
Graphics BFG 8800Ultra
Storage Seagate Barracuda 80GB SATA Perpendicular
Optical Plextor PX755-SA DVD/RW
Power Supply Tuniq Ensemble 1200W
Cooling Swiftech Storm / Laing Vario D38 / 3x Sunon 98CFM / Danger Den Res. / Black Ice XtremeIII (1/2")
Operating System Windows XP
Chassis Thermaltake Mozart TX (open chassis)


Water-cooling with Swiftech Storm, a triple radiator and three Sunon 98CFM fans dropped processor temps relatively close to my ambient room temp of 19C. In my test area the PC is just a few inches from sliding glass doors which explains the low temperatures.

Onto benchmarks and test results...
  • prev
  • next
Comment from thorgal @ 2008/03/27
Nicely written Keith ! Good thing you linked to Anand : the vdroop article by Kris is one of the best OC articles I've ever come across!

Did you play around with the tRD (static tRead value) setting to try and increase FSB above 445 ? You should also have a look at Kris' follow up article to the vdroop article, and see what's the impact of tRD.

Only pity in the review imho is the fact that DDR3 was not tested at 400x8 as well, then we could have comparable results to DDR2. At 266FSB DDR3 on this board seems slightly faster.
Comment from Massman @ 2008/03/27
Good job, Keith, liked the extensive text and the test

 

reply