X79 Overclocking
| BIOS Frequency and Voltage settings (for overclocking) | |||||
|---|---|---|---|---|---|
| Row 0 - Cell 0 | ASRock X79 Extreme9 | Asus P9X79 Deluxe | Asus P9X79 WS | Gigabyte G1.Assassin2 | Gigabyte X79-UD5 |
| Base Clock | 90-300 MHz (1 MHz) | 80-300 MHz (0.1 MHz) | 80-300 MHz (0.1 MHz) | 80-333 MHz (0.01 MHz) | 80-333 MHz (0.01 MHz) |
| CPU Multiplier | 12x to 60x (1x) | 12x to 57x (1x) | 12x to 57x (1x) | 12x to 59x (1x) | 12x to 59x (1x) |
| DRAM Data Rates | 800-2400 (266.6 MHz) | 800-2666 (266.6 MHz) | 800-2666 (266.6 MHz) | 800-3200 (266.6 MHz) | 800-3200 (266.6 MHz) |
| CPU Vcore | 0.60-1.52 V (5 mV) | 0.80-1.70 V (5 mV) | 0.80-1.70 V (5 mV) | 0.80-1.74 V (5mV) | 0.80-1.74 V (5mV) |
| VTT Voltage | 0.86-1.71 V (13 mV) | 1.05-1.70 V (6.25 mV) | 1.05-1.70 V (6.25 mV) | 0.72-1.61 V (5mV) | 0.72-1.61 V (5mV) |
| X79 PCH Voltage | 0.73-1.91 V (13 mV) | 1.10-1.70 V (6.25 mV) | 1.10-1.70 V (6.25 mV) | 0.87-1.98 V (5 mV) | 0.87-1.98 V (5 mV) |
| DRAM Voltage | 1.20-1.80 V (15 mV) | 1.20-1.99 V (5 mV) | 1.20-1.99 V (5 mV) | 0.83-1.51 V (5 mV) | 0.83-1.51 V (5 mV) |
| CAS Latency | 4-15 Cycles | 3-15 Cycles | 3-15 Cycles | 5-12 Cycles | 5-12 Cycles |
| tRCD | 4-15 Cycles | 4-15 Cycles | 4-15 Cycles | 5-31 Cycles | 5-31 Cycles |
| tRP | 4-15 Cycles | 4-15 Cycles | 4-15 Cycles | 5.15 Cycles | 5.15 Cycles |
| tRAS | 9-63 Cycles | 4-40 Cycles | 4-40 Cycles | 5-63 Cycles | 5-63 Cycles |
The addition of boot straps (chipset to CPU base clock ratios) to the LGA 2011 platform should have made overclocking easier compared to multiplier-locked LGA 1155-based CPUs, and yet every manufacturer appears to have figured out a way to make the process more difficult. ASRock, for example, make it clear where the ratios take effect as you scale up and down the base clock range. Altering Gigabyte’s base clock control by only 0.1 MHz caused boot failures with both motherboards when used in conjunction with higher-than-stock (33x) CPU ratios, leaving 100.00 and 125.00 MHz as our only overclocking options.
Asus’ limits were least-intrusive and appeared to be caused by changes in C1 core stepping power controls (a lower thermal throttling limit is one change that's easy to identify). We tried maxing out all of the settings we knew might help, but to no avail. Even still, 4.4 GHz at a mere 1.35 V core is nothing to be ashamed of.
We retested with a C0 Core i7 and shot straight to 4.7 GHz. We won't bother breaking down the settings we used to achieve that frequency because we don't want anyone to form expectations about a processor they can't actually buy. We'll instead keep testing whatever production-era samples we can get in hopes of finding a better example of Sandy Bridge-E's potential.
Once we figured out how ASRock's automatic adjustment worked, we were able to push our processor to a similar frequency as those achieved on both Asus boards. ASRock had also contacted us to say that it discovered the same C1-oriented multiplier problem in one of the 40 CPU samples it tested, credited a competitor with finding a workaround (we like that kind of honesty), and further stated that its team has developed a similar workaround we should have access to in mid-December. Since that's when the testing starts for our next round-up, we'll have a good opportunity to hold ASRock to that promise and report back to you.
Gigabyte’s results don’t look bad, but we were unable to overclock beyond stock Turbo Boost limits using multiplier adjustment alone. We were instead forced to use the 1.25x boot strap with a 34x multiplier and leave the 100 MHz base clock well enough alone.
Intel’s P67 X79 chipset still has the same 107-108 MHz limitation we've endured from mainstream Sandy Bridge platforms for the past 11 months. Boot straps multiply that range by 1.25x (and higher), though we were not able to reach the 166 or 250 MHz straps with our processor.
Asus has the best DRAM data rates, its P9X79 WS and P9X79 Deluxe taking first and second place.
