AMD just had their Financial Analyst day today and they just spoiled us with a ton of information for all of their upcoming products. So let’s get into some of it!
First, there's the entire core architecture roadmap. AMD basically gave us a list of architectures that are going to be released in some way in the next 2 and half years. We got Zen 4, obviously, Zen 4 V cache, interesting but we’ll get back to it soon, and Zen 4C which is in their Bergamo Epyc CPU’s. Those would be utilizing TSMC’s 5 and 4 nm process nodes. Past that, theres also Zen 5, Zen 5 V cache and Zen 5C. Those on the other hand are on 4 and 3 nm’s.
going back to zen 4 though, mark paper master answered a question that was on everybody’s mind about that next-generation architecture. What are the instructions per clock, AKA the IPC gains? Well, their testing found somewhere between 8 to 10%. This wasnt Just cinebench though. This was thoroughly tested with SpecInt 2017, Cinebench and geekbench. So that number is fairly final.
Add to that the insane clocks we saw during the last event, and its a pretty healthy boost in performance if you ask me. In fact, AMD claims over a 25% improvement in performance per watt compared to Zen 3 and over a 35% increase in their cinebench multi-threaded score.
They also talked about the next generation of mobile APUs. We talked about that a couple of weeks ago, but it's now official. Zen 4 APU’s codenamed Phoenix Point will come with RDNA 3 graphics. Not only that, but it will be based on the 4nm node. There was no mention of the beefier dragon range APUs, but AMD did confirm it would come so we’ll just have to wait and see. Zen 5 mobile APU's on the other hand would have RDNA 3+ graphics. Both of these would feature AI extensions.