Performance - CineBench
Before we begin testing a small note. Over the years we've been able to bring you persistent and consistent performance metrics. A discussion that has been ongoing for years now is when testing processors, there is reference performance as defined by Intel, and you have performance as offered and tweaked by motherboard manufacturers. We have ALWAYS tested processors at reference configuration. Most motherboards, however, will apply a number of tweaks, for example please look below:
Example screenshot of ASUS automatically enabling a tweak/enhancement.
So in this example, if you have Multicore enhancement enabled (which is an ASUS/MSI/ASRock/Gigabyte default), all cores will clock at higher turbo frequencies. Also, they'll be able to do that for a longer time e.g. The Intel reference could be 30 seconds for max two cores at 4.8 GHz, whereas this enhancement could apply all six cores at 4.8 for an indefinite time. For our processor reviews, we disable stuff like that, as we want to show you reference performance. For motherboard reviews, we'll leave tweak enabled as then you are testing the motherboard performance. We obviously do the very same in our AMD Ryzen reviews. Also, let me state, there is nothing wrong with these tweaks as hey, it's extra performance for free. However, for CPUs, we test at reference baseline specifications.
Processor performance: CineBench 15
We are slowly transitioning towards CineBench 15 as this newer version has the option to measure single threads. Apart from that, the rendering software R11.5 to R15 and new footage, the new version now supports systems with up to 256 threads. The performance of processors and graphics cards is, as usual, determined on the basis of 3D scenes. A selection of test results allows a rough classification of the benefit of your own system. For the CPU test is a scene with around 280,000 polygons used, while the GPU test based on OpenGL comes with about a million polygons, high-resolution textures, and various effects. The results will be issued in final points (CPU) and fps (GPU). According to the developers, the software has been "extensively developed to exploit the performance of new hardware as possible." The results are unsurprisingly not comparable with those from earlier versions. You'll notice we still need to add a number of processors, all in due time. You'll notice the single core perf paints a completely different picture here.
It is obvious that single threaded performance at 4.7~5.0 GHz, even in a short burst, will make a difference here.
Instructions per cycle (IPC)
Below I have started a new IPC test, now forgive the lack of results (I ditched the CPU-Z results after they changed the benchmark in the newer revision). This is the single CB15 run with the cores all locked (fixed) at 3500 MHz
This IPC test will build up and get updated over time. Basically, we lock all processor cores at 3500 MHz. We disable turbos and things like XFR. That way you can see the architecture performance of the processor clocked at exactly the same frequency. This is a single thread measurement. For many people, this is the holy grail of CPU measurements in terms of how fast an architecture per core really is. I, however, tend to say there's more to it than that, and that would be higher frequency allowances defining that per core performance.