Normally we wouldn’t report on something like this, because ultimately “some guy on a forum said a thing” isn’t exactly newsworthy on its own. However, in this case, the poster in question is Kepler_L2, a well-known fellow who frequently shares insider information and astute analysis of upcoming hardware, particularly for AMD.
Posting on the AnandTech forums in a long-running thread about Zen 5, Kepler_L2 replied to a user asking about specific claimed performance increases on AMD’s next-gen processors in the popular SPECint benchmark. SPECint is a test that measures a CPU’s pure integer math prowess, and while it doesn’t really have much relation to consumer client desktop workloads, it’s still meaningful in the high-performance compute (HPC) arena.
That means that these rumored gains are almost entirely from the major redesign of the CPU core that’s coming with Zen 5. AMD has stated unambiguously (see above slide) that Zen 5 will be the first total rejig of the Zen CPU core since its introduction, and if you’re ever going to see a “greater than 40%” performance gain from core changes, a clean-sheet do-over like this is the time, so it’s not totally unbelievable.
Posters in the thread point out that 40% happens to have been AMD’s goal with the original Zen architecture, and the company exceeded that number significantly in the final products. However, AMD was comparing to its own previous-generation “Piledriver” CPUs at that time, and if you’re reading this, you probably know how those chips went down. Zen 4 is a much higher mountain to climb if you’re trying to achieve a 40% performance gain. We reckon AMD will confirm or deny these rumors in its presentation at Computex barely more than a month away.