Here we have even more information about AMD’s upcoming Ryzen processor. This comes in the form of actual die shots of actual AMD Ryzen cores, which were presented at the International Solid-State Circuits Conference held earlier this month. First lets start with the die shot of the single AMD Ryzen core.
Looking at the die shot we can see that each core has only one integer and FPU section. This is a major design departure from earlier AMD designs were they split an FPU between two integer units. This change represents AMD’s attempt to improve IPC by moving from a CMT (clustered multi-threading) model to a SMT (simultaneous multi-threading) based model. This design change essentially trades two integer sections per core for a larger, better performing one.
You can see a section of the die labeled “Branch Prediction”. AMD has marketed this as “Neural Net Prediction”, but it is really known as a branch predictor. Its main function is to decide what a branch in the code will do before it actually happens, so the rest of the CPU can get ready for when that happens. Now if this prediction is wrong, the rest of the CPU has to spend time cleaning up after the missed prediction. Now a good branch predictor is not often wrong, so it actually saves more time that it looses by guessing correctly.
Now let’s take a look at a CCX, or Core Complex which is essentially a quad-core module.
AMD has made statements that their CCX module is flexible and allows for several adjustments to be made in the future with regards to L3 cache and core configuration. So if we were guessing AMD has plans for Ryzen will beyond the “Summit Ridge” line of processors. AMD has said they are confident on a “node parity” with Intel, which is something that we haven’t seen in a while.