Worth separating the two things people keep mashing together: core count per CCD, and core-to-core latency.
Today, a Ryzen 9 has two CCDs, each with up to 8 cores and its own slice of L3. A game thread that bounces between CCDs pays a real penalty because the two clusters talk over the fabric, not over the on-die L3. That's the classic reason people say "disable the second CCD" or "games only use one CCD."
A 12-core CCD changes the math in a good way for a 12-core part: instead of 6+6 or 8+4 across two clusters, you get one cluster. Fewer cross-CCD trips, not more. For a 12-core chip specifically, that's a genuine latency win, not a marketing one.
What it doesn't magically do:
- A 24-core part is still two CCDs. Games on that are still going to sit on one cluster.
- L3 per CCD matters more than core count for games. If they keep the same total L3 and spread it across 12 cores instead of 8, that's less cache per core, and some titles are cache-hungry.
- Clocks and memory controller behaviour still dominate at 1080p low settings.
So my read: good for mid-tier 12-core parts, basically irrelevant if you're buying 6 or 8 cores, and the interesting question isn't the core count, it's whether L3 grows with it. Leaks rarely say that part clearly.