- Joined
- Mar 17, 2009
- Messages
- 6,246
What is a sensible everyday use overclock in terms of processor efficiency ?
This is a subject that has been discussed numerous times within the threads and something I have intended to test for quite a while. To really measure the gain one has to measure the percentage overclock from stock and compare that to the percentage gain and the percentage increase in power consumption. Admitted, there are those that don't care about the power consumption, but I suppose these tests are for those people that wish to ensure that the percentage increase remains proportional.
For these tests, Prime95 was used to load the CPU 100% for the power consumption measurements. 7Zip benchmark was used to measure compression and de-compression of the default 32mb dictionary size.
A Maplin Plug-in power and energy monitor was used for the power consumption and a Digital multimeter for the V Core measurements.
The Z68 system in my signature was used.
My CPU is able to OC to 4,5 with all settings on Auto in the BIOS. Anything above that requires manual intervention on the core voltage to get the CPU stable. Unfortunately I do not have a "golden sample" CPU and I need ~1,49V core for a stable 4,9GHz OC under full load. Previous tests at higher than 4,9GHz were not done as I had no intentions of deteriorating my CPU any further. My initial tests indicated a 1,5V + requirement to go beyond 4,9GHz and the intention was purely to test the CPU within practical and reasonable parameters.
The tests were done at 400MHz increments. The increased idle Watts at 4900 MHz can only be contributed to the manual Voltage intervention.
For the analysis I am using 3,3GHz (2500K stock) as the base to show the percentage increases:
? 3,3GHz to 3,7GHz
? 3,3GHz to 4,1GHz
? 3,3GHz to 4,5GHz
? 3,3GHz to 4,9GHz
From the analysis it can be seen that everything sort of remains proportional up to 4,5GHz OC. The system power consumption basically follows the percentage increase in processor OC. At 4,9GHz though you will see the sudden jump.
The eye opener is the power requirement to achieve a 1,2GHz increase in processor clock; ~50W additional power consumption at 4,5GHz, which is still within reason, but then, for another 400MHz (additional 12%) requires another ~50W. That is just insane. If measured as percentage increase from 4500 to 4900 then it is only an effective 8,8% above 4,5GHz.
To run your CPU up to a certain OC, and within the manufacturers voltage range, can indeed have a substantial performance gain. Obviously, there will always be an associated increased power requirement. Every CPU series will have a different point at which it becomes an ineffectual exercise to go beyond for everyday use as the power consumption of the CPU just becomes prohibitive.
This is a subject that has been discussed numerous times within the threads and something I have intended to test for quite a while. To really measure the gain one has to measure the percentage overclock from stock and compare that to the percentage gain and the percentage increase in power consumption. Admitted, there are those that don't care about the power consumption, but I suppose these tests are for those people that wish to ensure that the percentage increase remains proportional.
For these tests, Prime95 was used to load the CPU 100% for the power consumption measurements. 7Zip benchmark was used to measure compression and de-compression of the default 32mb dictionary size.
A Maplin Plug-in power and energy monitor was used for the power consumption and a Digital multimeter for the V Core measurements.
The Z68 system in my signature was used.
My CPU is able to OC to 4,5 with all settings on Auto in the BIOS. Anything above that requires manual intervention on the core voltage to get the CPU stable. Unfortunately I do not have a "golden sample" CPU and I need ~1,49V core for a stable 4,9GHz OC under full load. Previous tests at higher than 4,9GHz were not done as I had no intentions of deteriorating my CPU any further. My initial tests indicated a 1,5V + requirement to go beyond 4,9GHz and the intention was purely to test the CPU within practical and reasonable parameters.
The tests were done at 400MHz increments. The increased idle Watts at 4900 MHz can only be contributed to the manual Voltage intervention.
For the analysis I am using 3,3GHz (2500K stock) as the base to show the percentage increases:
? 3,3GHz to 3,7GHz
? 3,3GHz to 4,1GHz
? 3,3GHz to 4,5GHz
? 3,3GHz to 4,9GHz
From the analysis it can be seen that everything sort of remains proportional up to 4,5GHz OC. The system power consumption basically follows the percentage increase in processor OC. At 4,9GHz though you will see the sudden jump.
The eye opener is the power requirement to achieve a 1,2GHz increase in processor clock; ~50W additional power consumption at 4,5GHz, which is still within reason, but then, for another 400MHz (additional 12%) requires another ~50W. That is just insane. If measured as percentage increase from 4500 to 4900 then it is only an effective 8,8% above 4,5GHz.
To run your CPU up to a certain OC, and within the manufacturers voltage range, can indeed have a substantial performance gain. Obviously, there will always be an associated increased power requirement. Every CPU series will have a different point at which it becomes an ineffectual exercise to go beyond for everyday use as the power consumption of the CPU just becomes prohibitive.