Thanks for the explanation, I hadn't read about CCD's. And I just noticed I didn't finish my first paragraph hehehe, can't recall what I intended to type.
I suppose AMD are working to reduce the latencies between CCX and CCD to further improve performance. Why is it that games are more sensitive to this than productivity applications (judging from relative performance when compared to Intel in benchmarks)? Sorry for this off topic tangent but I don't want to trawl through some of the massive Ryzen threads for the info.
Zen 3 will have single 8 core 16 thread CCX CCD's, so its like 9900K chiplets, no more splitting the chip dies into CCX's, that on its own will make these chiplet dies behave like an 8 core 3300X
AMD also said they will have 15% higher IPC, so far with Zen AMD have always did what they said or under promised, but you know.... we'll see.
The best example of productivity i can give you is this....
Single threaded:
11, Score 547: AMD Ryzen R9 3950X at 4.7Ghz,
jordand77
12, Score 547: Intel Core i7 8700K at 5.2Ghz,
Radox-0
4.7Ghz vs 5.2Ghz is about 10%.
Multithreaded:
Score 4125: AMD Ryzen R5 3600 at 4.5Ghz,
RavenXXX2
Score 4124: Intel Core i7 8700K at 5.2Ghz,
Chaos666
4.5Ghz vs 5.2Ghz is about 15%
So 10% higher per core IPC, the 15% is probably because AMD have slightly better SMT.
If you look back at this slide, 3300X vs the 7700K is 1% faster here, but the 7700K is 15% higher clocked, so 14% higher IPC, now look at the difference between the single CCX 3300X and the dual CCX 3100. both at 4.4Ghz the 3100 209 FPS, the 3300X 242 FPS, a difference of 16%.
All margin of error with in 1 or 2% similar to the IPC difference in MT Cinebench R20
So again, just having a single CCX vs dual increases the gaming performance 'in this game at least' by 16%, HUB tested one at 7%, so it depends on the game. i guess where the game is really sensitive to intercore latency the more it hits performance.