New Details about AMD’s Ryzen Architecture Were Revealed at ISSCC

More and more details are surfacing regarding AMD’s brand new Ryzen CPU lineup, and the latest piece of news comes straight from the International Solid-State Circuits Conference, which took place earlier this month. What’s truly interesting is that we’re finally able to catch a glimpse of some actual die shots of AMD Ryzen cores, as well as some shots of AMD’s core complex modules, both of which provide some insight into the Ryzen architecture.

Starting off with the AMD Ryzen core, we instantly notice that each core has just one integer and FPU section, which is a very important design choice as it differs greatly from AMD’s previous Bulldozer and Piledriver architectures. While this design choice will definitely help improve the product’s single threaded performance and IPC, we should also take note of the “Branch Prediction” section, which will be tasked with identifying what a branch in the code will do before it actually happens. It’s true that these systems can be wrong at times, in which case the CPU performance is pushed back a bit as it needs to clean up after the failed prediction, but these cases are incredibly rare when it comes to good branch predictor, which is likely why AMD has decided to implement one. The company’s solution was dubbed “Neural Net Prediction,” and it is reportedly much better when compared to previous iterations.

As for the CPU’s CCX or Core Complex, its main asset comes in the form of flexibility, as it will allow AMD to make various adjustments regarding core configuration and  L3 cache in the future. Moreover, AMD hopes that Ryzen will reach “node parity” with Intel, all thanks to a simple and efficient design.

Last but not least, the company wanted to reveal a few details about its “Precision Boost” system that will enable accurate clock control for the processor. This tech could be used for automatic overclocking or for energy saving. Working alongside the Pure Power monitoring tech, Precision Boost adjusts the CPU clock in 25Mhz increments depending on the CPU’s “health and workload,” and according to AMD, this is done without any “halts or queue drains.”

These technologies definitely look impressive at a glance, and most of us already suspected that Ryzen would show great promise. Right now, we’re looking forward to the lineup’s official launch in order to analyze some real world performance.