As we come to the end of 2017, we can reflect back to see that AMD has made a very successful comeback to the high-performance CPU market with their “Ryzen” brand of CPUs. They launched providing HEDT-like performance with mainstream price tags to a great response. The new “Zen” core is a complete rework from the ground up, replacing the aging “Bulldozer” family of processors. AMD surpassed their own target and managed to achieve an impressive 52% uplift in Instructions Per Clock (IPC) over the “Excavator” core. Combined with advances in clock gating and power efficiency and a new 14nm FinFET process, Ryzen is a remarkably efficient yet very powerful core.
Performance per clock of the “Zen” core is high enough to allow it to be competitive with Intel’s best CPU cores of the Broadwell and Skylake family, but just how close is it? Last month Intel released their response to the Mainstream Ryzen CPUs in the form of Coffee Lake, lead by the six-core, twelve thread i7 8700K. But the underlying architecture of the new processor cores remains the same as the Skylake parts from 2015.
Today I am going to take a quick (rather academic, as I will explain later) look at Ryzen’s core to core, clock to clock performance against a Coffee Lake CPU with the same number of cores, threads, cache and running at the same frequency. I will be pitting a Ryzen 3 1200 and i3 8350K against each other in a series of benchmarks and game tests, with both processors running at 3.5 GHz and using 2400 MHz, C15 RAM.
The CPUs
Up first is the Ryzen 3 1200. This is a quad-core, four-thread part based on the same Zeppelin die that makes up all of the “Zen” CPU parts from the R3 series to the highest end Threadrippers and even Epyc server processors. It is a modular design with 8 physical cores clustered into two groups called “Core Complexes”(CCX). Each CCX contains four closely-linked cores and an 8MB L3 cache. The two groups are connected together using AMD’s Infinity Fabric.
The Ryzen 3 has two cores per CCX disabled and also loses 4MB of L3 per cluster, too. Giving it four cores and 8MB of L3 (split into two 4MB chunks). This is an effect way of re-using dies with defective cores or L3 cache.
The “Zeppelin” die, clearly visible are the two Core Complexes.
On the Intel Side, we have the i3 8350K. This is also a quad-core, four-thread processor. It is based on a native quad core die consisting of a four core group connected with a ring-bus topology and an 8MB block of shared Last Level Cache. Although in the “8th Generation” branding, the 8350K is likely a Kaby Lake die rebadged into 8th gen clothing.
A die shot of Kaby Lake-R, the chip used for the 8th generation i3 series.
Test Setup
For this test both CPUs are clocked to 3.5 GHz and are using 2400 MHz RAM with 15-17-17 timings. An Nvidia GeForce GTX 1070 Ti video card was used to allow the bottleneck to be the CPU in the video game tests.
A little note
One thing I will get out of the way; I am not comparing these parts on a value or overall performance basis. The i3 8350K is 60% more expensive than the Ryzen 3. That’s a lot, the real competitor to this part would be the i3 8100. But since I only have an 8350K on hand, it will do (and has the same amount of L3 Cache as the Ryzen 3). Neither am I going to overclock each part to its limit. This is merely an academic look at the clock to clock, core to core performance of two processor architectures. In addition, I am fully aware of the fact that nobody in their right mind in 2017 would play a game at 800×600 or rarely even at 720p; but for the sake of shifting the bottleneck to the CPU and seeing how many frames each part can push, such resolutions were used.