Overclocking an Intel Core i7-2700K (Page 3)

Results and Conclusion

For this article, I am only testing the stock vs overclocked performance, to see how well it scales and how easy this chip is to overclock. In future articles I will be able to compare processors I review against each other, giving a more meaningful comparison.

Stock4.6GHz4.8GHz5.0GHz
CPU Voltage1.24V1.3V1.38V1.45V
Loaded Temp (Air)40C66C76Cn/a
Loaded Temp (Water)n/an/a56C61C

As you can see, above 4.6GHz, this processor requires a CPU voltage increase of approximately 0.04V for every 100 MHz frequency increase. A quick test showed that this scales up to the maximum multiplier that works on my processor, which is 55x. It took 1.63V to benchmark at those settings, and my cooling kept the processor below 70C for all tests.

I didn't stress test at that speed as it would have needed more voltage and generated too much heat for my cooling. If you want to try running your processor with settings that high, I recommend that you check Intel's new overclocking insurance as some processors are at risk of failing with that much voltage.

Linpack benchmark results

First up, we have some Linpack results, as measured using Intel Burn Test 2.53. This measures how many raw GFLOPS performance is available.I always set it to the number of threads the processor can execute, so the processor is thoroughly tested.

SuperPi 1M benchmark results

SuperPi 32M benchmark results

SuperPi measures single threaded performance while calculating the value of Pi. In the test on the left, the value is calculated to 1 million decimals. The test on the right calculates 32 million decimals. In both these tests, the performance scales well with the speed on a single core.

wPrime 32M benchmark results

wPrime 1024M benchmark results

wPrime calculates the square roots of the first n numbers. In the results on the left, square roots of the first 32 million numbers are calculated, and on the right we have 1,024 million. In this test you can set the number of threads used, so I set it to use 8 threads. The numbers scale nicely in both of these as well.

ucbench benchmark results

Unrar-crack benchmark (otherwise known as ucbench) tests the number of password attempts that can be tried per second in a rar password cracking app, using different numbers of cores and different instruction sets. The results her scaled nicely as well, however the best results it gave were using SSE2 instructions. I would have expected Sandy Bridge processors to operate faster with later instructions.

Fritz Chess  benchmark results

Fritz Chess benchmark measures how quickly a computer can execute a set of chess moves. The processor doesn't scale evenly at higher speeds when running this benchmark. The result at 5.0 GHz is only about 0.3% higher than at 4.8 GHz, showing that on this processor not all applications will benefit from the higher speeds and the extra power consumption that goes with it.

This is a very easy processor to overclock, and combined with the Maximus IV Gene-Z and it's UEFI BIOS, reached 4.8 GHz with air cooling and 5 GHz with water cooling. I was also able to easily reach speeds up to 5.5 GHz, but not with safe voltages. Given adequate cooling, the potential of the Core i7-2700K is as good as any quad core processor on the market today, and as long as you can keep temperatures well below 70C under load, it is easy to get speeds in excess of 5 GHz (which I used only for short runs because of the voltage involved).

On all but Fritz Chess Benchmark, the performance scaled pretty evenly with the clock speed, giving a good indication that most applications which fully load one or more cores will benefit from an overclock at most speeds.


<-   Page 1: Introduction
<-   Page 2: BIOS overclocking

Comments: 0
Terms and Conditions · Privacy Policy · Contact Us (c) Copyright 2003 - 2010 Gennadiy Shvets

Search CPU-World

Search site contents:

Identify part

Identify CPU, FPU or MCU:

Related CPU families

Quick Links

Jump to CPU / Family

Type in partial CPU manufacturer, family, model and/or number.