Do Intel's Sandy Bridge CPUs perform better when they're in a colder environment? That's the theory, and we put it to the test.
During the launch phase of Sandy Bridge we kept hearing that the new processors would perform better in a colder environment. Puzzled by this statement we asked Intel if a processor in Antarctica would outperform one located in Hawaii, and the reply was that ambient temperature definitely plays a part in the performance of Sandy Bridge. Armed with this small piece of information and nothing more, we set out to uncover the truth behind the Sandy Bridge temperature claims.
A performance gain due to the temperature at which the processor operates could come in two distinctly different ways. One being a frequency increase, since temperature plays a big part in overclocking and the maximum frequency a component can achieve. The other would be a performance increase at the same frequency; this would mean an HD video transcode in Antarctica would complete quicker than in Hawaii. We set up a series of simple experiments to test both theories.
How do environmental temperatures affect Sandy Bridge performance?
In the first experiment we started with our processor at 40 degrees Celsius and ran 3DMark Vantage. We then decreased the temperature to 30 degrees and repeated the process until we reached -20 degrees. Over our seven 3DMark runs at different temperatures we experienced no significant change in score: this would indicate that there is no performance gain at the same frequency when temperatures are different.
Next we set up the same system and set the temperature to 40 degrees Celsius. This time we tried to achieve the maximum overclock possible before decreasing the temperature in 10-degree intervals. Starting at 40 degrees, we achieved a frequency of 4,900MHz; at 30 degrees we found it was possible to reach 5,000MHz; and we reached a peak frequency of 5,100MHz at 10 degrees Celsius. This proves that temperature does
affect the maximum performance, but this has been the case for nearly 10 years (and surely the Intel spokesperson meant more with their comment).
Going back to the drawing board we decided the Turbo Boost feature was worth investigating. The Turbo Boost works by increasing the CPU frequency when there is demand under heavy load, then decreasing the frequency when the temperature becomes too high. With this in mind we set up a third experiment.
After enabling the Turbo Boost functionality and setting our processor to 4,800MHz we were ready to go. Starting at 40 degrees Celsius we ran the Prime 95 test to generate a massive heat load and then observed the platform. After a few seconds the frequency dropped to 4,600MHz, then dropped again to 4,400MHz, all the way down to 3,900MHz. After we decreased the temperature to 30 degrees Celsius it took longer for the frequency to drop; this pattern continued repeatedly right down to -10 degrees.
Intel was spot on, it seems temperature does
play a big part in Sandy Bridge CPU frequency, but only under certain circumstances. Those wanting to overclock their systems will benefit from investing in serious air cooling or even high-end water cooling systems. What we can draw from this analysis is how Sandy Bridge was designed to work. During periods of heavy usage the Turbo Boost provides a temporary speed boost, and when the temperature becomes too high the frequency scales back. While you can gain a small performance boost by running the system in a colder environment, we still think the cost of moving to Antarctica probably isn’t worth the return on investment.