Next up we began testing with Hyper Pi to test the motherboard and CPU combination at very high usage levels. Hyper Pi runs one instance on each core of the processor simultaneously, both virtual and physical cores. In this test we are running 12 instances of 32M to test the maximum performance of the processor. Each of the 12 tests will calculate Pi to 32 Million decimal places!
With 12 instances, we are looking at 384,000,000 million Pi decimal places being calculated. Essentially a race to finish the simultaneous calculations, the lower time is better.
The 3820 actually outperforms the 3930K in this benchmark, especially with the stock settings. The 3820 has a higher clock speed at the stock settings with 3.6 compared to 3.2 for the 3930K. The Turbo speed of the 3820 is only .1 higher at 3.9 compared to 3.8 for the 3930K.
Since the benchmark runs a single instance on every single core, the score given is the average of the score for each individual cores completion time. Some cores will take longer than others to complete the calculation, even when calculating the same data, and at the same clock speed. This is a common story with all computer hardware, as all silicon, even individual cores within one CPU, will not perform exactly the same. The results of the 3930K, having more cores, does stand the chance of having more cores finish at a slower speed, thus possibly causing the slower overall score. This test was conducted multiple times to make sure that the results were reproducible.
An interesting result of these tests is that a 6-core processor will not always outperform a 4-core in all situations. In the vast majority of common user tasks, there will be very little difference. One must also keep in mind that there is also the fact that the 3820 is an entirely different die than the 3930K, so there may be some architectural differences that also explain the results.
HYPER PI POWER TESTING
Again we are testing the power draw of the processors during the Hyper Pi testing. These power draw measurements are much higher than one would experience in normal usage. Here we can see a recurring theme: the 3820 consumes less power at a 5.0 overclock than its 6-core counterpart does with stock speeds.
These results again point back to the fact that the 3820 can excel in some areas, while still consuming significantly less power. With higher base clocks at stock settings, the 3820 for a casual user with no intention of overclocking is going to be a very clear choice. If one were to mix in the fact that the 3820 actually outperformed the 3930K in overclocked results, with a 1 second advantage, and less power than a stock 3930K, one could make a very clear case for the 3820 winning this round of testing.
Cinebench is a very CPU intensive benchmark. It is based upon CINEMA 4D which is a commonly used program for 3d graphic creations for studios and production houses. Using the CPU for rendering is a good way to really push the performance up a notch. The higher score is better.
The great thing about this testing is that it can be representative of the higher loadings that one would expect in a professional environment. We can also use the very same test to isolate single core performance with the same data testing, which is particularly useful for our purposes.
The Stock and Overclocked ‘single’ categories are with a single CPU core active, and the ‘Multi’ categories are with all processor cores being put under load.
A large percentage of consumer applications are woefully single threaded, usually only stressing one core. The 3820 matches the 3930K in both single core test scenarios. This is very illustrative of the performance similarities that will be experienced with these CPUs in single threaded (majority) of applications.
This multi-core testing stresses all the cores of the processor to the maximum, and showcases a bit of the power advantage for the 3930K when it comes to multithreaded applications. Professional level graphics and video/audio encoding programs are much more intensive, designed to maximize the performance by utilizing all cores simultaneously.
In the Multi-core testing the 3920K handily outperforms the 3820 as the full power of its cores are brought to bear.
CINEBENCH POWER TESTING
Once again the 3820 at overclock levels draws nearly the same power as a 3930K at stock when used in a multicore scenario. The power levels are very close when testing only one CPU. This is a nice power measurement for the 3930K as well, as with these tests it sips power when using only one core, but can turn up that power on demand.