- Mar 16, 2014
- 183
- 0
- 10,760
Introduction.
I asked myself the same question a while ago. There isn’t a large quantity scientific, detailed analysis on this subject out there, so I decided to try it out for myself. In theory, a Push-Pull configuration should normalise the pressure in the tower, and therefore reducing the resistance caused by obscured airflow. In plain English, the airflow through the tower will not be as affected by the cooling fins, since these cause resistance. The airflow will have an easier time travelling though the cooling tower, easily said.
In my testing system, I’m using a Noctua NH-U12S CPU cooler. It has compatibility for one or two NF-F12 fans, one of which comes as standard. The NF-F12 fans uses Noctua’s FocusedFlow-technology, to have a better and more focused static pressure than other fans in the same category. These fans are designed for heatsinks and radiators specifically. The CPU in the system is an Intel Core i7-4820K, sitting on a Republic of Gamers Rampage IV Gene motherboard, which is a motherboard with X79, of which this system is based. The CPU is overclocked to 4200MHz, and all of the tests are executed with 4200MHz on the CPU cores. The GPU is an Nvidia GeForce GTX 760OC from ASUS. The RAM are two sticks of HyperX Savage, running at 1866MHz with timings 9-10-11-27 with a total capacity of 16GB. Finally, the PSU in the system is an EVGA SuperNova G2 750W.
This system has a presumed negative pressure. I have two Corsair AF120 Quiet Edition as intake fans in the front, one Corsair AF120 Quiet Edition in the rear and two Corsair AF140 Quiet Edition in the roof. All three 120 millimetre fans are driven through a Fractal Design standard 5 volts, 7 volts and 12 volts fan controller. The 140 millimetre fans are PWM driven. The pressure isn’t tested, hence why I will not guarantee a negative pressure in this case, since there are some areas of dust patches formed around some of the nooks and crannies inside and outside the case. The case is a Fractal Design Arc R2 Mini, with a custom built panel for the drives to make sure no airflow from the intake fans is obscured.
If you are very interested in this subject, I have found one or two very well written explanations for the physical properties of airflow and how case pressure and case obscurance affect cooling performance. Very little of this is necessary to know in order to understand this small experiment. Neither is this article written by a professional, which these in-depth guides about airflow are. I presume.
Execution.
When gathering the temperature information during these tests, I have taken an average of four values, tested and written down each five minutes. This will create an accurate value which we can use as our platform when experimenting with this setup. I got the method from another test on Overclock.net. Though, my averages are not as in-depth as those. It’s the same thing, but in other words, I’m a bit lazy.
These values has been tested in a few different circumstances. I took the temperatures during system idle, during a stress-test with Prime95 and during a stress-test with Prime95 and Valley Benchmark. The Prime95-test was run using the blend setting, to test our CPU and some of the RAM at the same time. I ran the Valley Benchmark test together with Prime95 to measure total system performance, to see how much it affected the ambient temperature in the case, causing other components to heat up. Also, to see if this put a bigger stress on the CPU, causing it to heat up on its own way, rather than the ambient temperature in the case increasing. Although, I would presume that it is a combination of the two, but mostly due to the ambient temperature in the case increasing.
Experiment.
During these tests, the ambient temperature in the room was 18° Celsius. The temperature remained the same during the entire experiment, but you should factor in a margin of error of +/- 1 degree Celsius. To replicate the day-to-day scenario, and to make full use of the case pressure, the side panel has remained on during all of these sequences of the experiment. Less writing, more workshop!
When the Noctua NH-U12S was equipped with one Noctua NF-F12 fan, the temperature when idle was 27°C. The fan was, on average, spinning at 500 RPM. This is fairly normal air-cooled CPU temperature when running idle. Trying to get lower temperatures on idle is unnecessary, though. It’s when the temperature is rising that we find the important and interesting values.
When stress-testing the CPU with Prime95, the average CPU temperature was 50°C. Nothing spectacular. But remember, the CPU is overclocked to 4200MHz. The CPU fan span at an average of 900 RPM. Still, with the NF-F12 fans, very little noise. Excellent job, Noctua!
Finally, the Prime95 and Valley Benchmark test. The CPU temperature rose to an average of 52°C. The CPU fan was still, on average, spinning at 900 RPM. As mentioned earlier, the reason for the rise in temperature is probably explained due to increased heat building up in the case, due to nearly the entire system being stress-tested. This graph below should show the values, slightly easier to read. I hope.
Moving strictly on to two fans. When running idle, the average CPU temperature was 24°C. Both fans ran at the same speed, on average 300 RPM. All other factors are still the same, which gives us an average reduction in temperature of 3°C. 24°C is actually quite great for an air cooled idle temperature. Although, idle temperatures is nothing to brag with.
When stress testing with Prime95, the average temperature on the CPU was 45°C. The two CPU fans was spinning at 500 RPM on average. Quite impressive temperatures, for an overclocked and air cooled CPU. Seeing this, if the fans are taken into manual control, we might be able to take some more degrees of that value. After a while though, we reach the absolute point where we can’t deduct more units of temperature, with this setup, since the temperature is starting to normalise. This all has to do with the defined rules of nature, with heat conduction and so on. I will not go further than that, here and now.
Now, the final test with Prime95 and Valley Benchmark. Again, the CPU temperature increased, by not by much. On average, the CPU temperature was 47°C. The average fan speed was the same as when only running with Prime95. The reason for this is that the actual CPU indirectly heats up. It’s the air, and therefore the airflow in the case becoming warmer, due to the GPU emitting more heat, simply. The Valley Benchmark scores in this example is not important, however I can tell you that there was a 3 point difference in the final scoring. Not strange, since it's a GPU benchmarking program. Again, I hope this graph below will make our lives slightly easier.
Conclusion.
Forcing a Push-Pull scenario to an air-cooled CPU setup makes a difference. Though, the difference between only one fan and having two fans is small. It’s a marginal difference. These tests are carried out in a “best possible” scenario, with only the benchmarking and monitoring software running. In the real world, I’m sure that the results would be a bit more spread out, since different types of software and games are not a synthetic benchmark. They use the resources in the computer more dynamically.
Is it worth it? That’s quite personal to decide. The results might vary in other systems than mine. Perhaps, if your average CPU temperature is higher than mine, or you have a higher overclock, it might make a bigger difference, and a bit more sense. I think then that this could work as a benchmark. A template to show if it makes any significant difference. Difference it makes, but not such a big one.
A massive Thank You for reading! Please, if you see any obvious errors in the scientific data, or any other obvious fault, I’d appreciate if you helped me to correct it!
Thank You.
I asked myself the same question a while ago. There isn’t a large quantity scientific, detailed analysis on this subject out there, so I decided to try it out for myself. In theory, a Push-Pull configuration should normalise the pressure in the tower, and therefore reducing the resistance caused by obscured airflow. In plain English, the airflow through the tower will not be as affected by the cooling fins, since these cause resistance. The airflow will have an easier time travelling though the cooling tower, easily said.
In my testing system, I’m using a Noctua NH-U12S CPU cooler. It has compatibility for one or two NF-F12 fans, one of which comes as standard. The NF-F12 fans uses Noctua’s FocusedFlow-technology, to have a better and more focused static pressure than other fans in the same category. These fans are designed for heatsinks and radiators specifically. The CPU in the system is an Intel Core i7-4820K, sitting on a Republic of Gamers Rampage IV Gene motherboard, which is a motherboard with X79, of which this system is based. The CPU is overclocked to 4200MHz, and all of the tests are executed with 4200MHz on the CPU cores. The GPU is an Nvidia GeForce GTX 760OC from ASUS. The RAM are two sticks of HyperX Savage, running at 1866MHz with timings 9-10-11-27 with a total capacity of 16GB. Finally, the PSU in the system is an EVGA SuperNova G2 750W.
This system has a presumed negative pressure. I have two Corsair AF120 Quiet Edition as intake fans in the front, one Corsair AF120 Quiet Edition in the rear and two Corsair AF140 Quiet Edition in the roof. All three 120 millimetre fans are driven through a Fractal Design standard 5 volts, 7 volts and 12 volts fan controller. The 140 millimetre fans are PWM driven. The pressure isn’t tested, hence why I will not guarantee a negative pressure in this case, since there are some areas of dust patches formed around some of the nooks and crannies inside and outside the case. The case is a Fractal Design Arc R2 Mini, with a custom built panel for the drives to make sure no airflow from the intake fans is obscured.
If you are very interested in this subject, I have found one or two very well written explanations for the physical properties of airflow and how case pressure and case obscurance affect cooling performance. Very little of this is necessary to know in order to understand this small experiment. Neither is this article written by a professional, which these in-depth guides about airflow are. I presume.
Execution.
When gathering the temperature information during these tests, I have taken an average of four values, tested and written down each five minutes. This will create an accurate value which we can use as our platform when experimenting with this setup. I got the method from another test on Overclock.net. Though, my averages are not as in-depth as those. It’s the same thing, but in other words, I’m a bit lazy.
These values has been tested in a few different circumstances. I took the temperatures during system idle, during a stress-test with Prime95 and during a stress-test with Prime95 and Valley Benchmark. The Prime95-test was run using the blend setting, to test our CPU and some of the RAM at the same time. I ran the Valley Benchmark test together with Prime95 to measure total system performance, to see how much it affected the ambient temperature in the case, causing other components to heat up. Also, to see if this put a bigger stress on the CPU, causing it to heat up on its own way, rather than the ambient temperature in the case increasing. Although, I would presume that it is a combination of the two, but mostly due to the ambient temperature in the case increasing.
Experiment.
During these tests, the ambient temperature in the room was 18° Celsius. The temperature remained the same during the entire experiment, but you should factor in a margin of error of +/- 1 degree Celsius. To replicate the day-to-day scenario, and to make full use of the case pressure, the side panel has remained on during all of these sequences of the experiment. Less writing, more workshop!
When the Noctua NH-U12S was equipped with one Noctua NF-F12 fan, the temperature when idle was 27°C. The fan was, on average, spinning at 500 RPM. This is fairly normal air-cooled CPU temperature when running idle. Trying to get lower temperatures on idle is unnecessary, though. It’s when the temperature is rising that we find the important and interesting values.
When stress-testing the CPU with Prime95, the average CPU temperature was 50°C. Nothing spectacular. But remember, the CPU is overclocked to 4200MHz. The CPU fan span at an average of 900 RPM. Still, with the NF-F12 fans, very little noise. Excellent job, Noctua!
Finally, the Prime95 and Valley Benchmark test. The CPU temperature rose to an average of 52°C. The CPU fan was still, on average, spinning at 900 RPM. As mentioned earlier, the reason for the rise in temperature is probably explained due to increased heat building up in the case, due to nearly the entire system being stress-tested. This graph below should show the values, slightly easier to read. I hope.
Moving strictly on to two fans. When running idle, the average CPU temperature was 24°C. Both fans ran at the same speed, on average 300 RPM. All other factors are still the same, which gives us an average reduction in temperature of 3°C. 24°C is actually quite great for an air cooled idle temperature. Although, idle temperatures is nothing to brag with.
When stress testing with Prime95, the average temperature on the CPU was 45°C. The two CPU fans was spinning at 500 RPM on average. Quite impressive temperatures, for an overclocked and air cooled CPU. Seeing this, if the fans are taken into manual control, we might be able to take some more degrees of that value. After a while though, we reach the absolute point where we can’t deduct more units of temperature, with this setup, since the temperature is starting to normalise. This all has to do with the defined rules of nature, with heat conduction and so on. I will not go further than that, here and now.
Now, the final test with Prime95 and Valley Benchmark. Again, the CPU temperature increased, by not by much. On average, the CPU temperature was 47°C. The average fan speed was the same as when only running with Prime95. The reason for this is that the actual CPU indirectly heats up. It’s the air, and therefore the airflow in the case becoming warmer, due to the GPU emitting more heat, simply. The Valley Benchmark scores in this example is not important, however I can tell you that there was a 3 point difference in the final scoring. Not strange, since it's a GPU benchmarking program. Again, I hope this graph below will make our lives slightly easier.
Conclusion.
Forcing a Push-Pull scenario to an air-cooled CPU setup makes a difference. Though, the difference between only one fan and having two fans is small. It’s a marginal difference. These tests are carried out in a “best possible” scenario, with only the benchmarking and monitoring software running. In the real world, I’m sure that the results would be a bit more spread out, since different types of software and games are not a synthetic benchmark. They use the resources in the computer more dynamically.
Is it worth it? That’s quite personal to decide. The results might vary in other systems than mine. Perhaps, if your average CPU temperature is higher than mine, or you have a higher overclock, it might make a bigger difference, and a bit more sense. I think then that this could work as a benchmark. A template to show if it makes any significant difference. Difference it makes, but not such a big one.
A massive Thank You for reading! Please, if you see any obvious errors in the scientific data, or any other obvious fault, I’d appreciate if you helped me to correct it!
Thank You.
Facebook
Twitter
Instagram