It looks like AMD is heavily investing in stacked cache and chiplet technologies for their next-gen CPU & GPU designs. Based on a new rumor by Greymon55, it looks like after EPYC & Ryzen, AMD will be bringing stacked 3D Cache design to RDNA in the form of 3D Infinity Cache.
Next-Gen AMD RDNA GPUs Might Feature 3D Infinity Cache Technology - 3D Stacking Cache on MCM GPUs
With RDNA 2, AMD introduced its first-generation Infinity Cache architecture which is a fast and high-bandwidth on-die cache that the GPU can quickly access. The existing cache designs can be scaled up to 128 MB capacities and up to 2 TB/s bandwidth. With next-gen RDNA 3 GPUs, it has been rumored that the GPUs will double the amount of Infinity cache with Navi 33 expected to offer 256 MB and Navi 31 expected to offer up to 512 MB of Infinity Cache.
3D IFC
— Greymon55 (@greymon55) November 8, 2021
N31
N32
— Greymon55 (@greymon55) November 8, 2021
The IFC on Navi 31 GPUs will be split between two chips since it is an MCM design so it's still 256 MB per chip. The latest rumors now suggest that the Infinity Cache is also moving to a 3D stack design. So we aren't just getting MCM GPUs with RDNA 3 but there's also a possibility of getting stacking technology on the next-gen chips. So in total, AMD will have 3D cache technology on its entire lineup including Ryzen, EPYC, and Radeon.

AMD is also going to introduce its server aimed CDNA 2 GPUs today which will be the first to feature an MCM technology. So far, there have been no reports of Infinity Cache on the current CDNA design but it may be possible that future GPUs could include the same cache stacking technology as bandwidth demand increases.
AMD RDNA 3 Navi 3X GPU Configurations (Preliminary)
| GPU Name | Navi 21 | Navi 33 | Navi 32 | Navi 31 | Navi 3X |
|---|---|---|---|---|---|
| Codename | Sienna Cichlid | Hotpink Bonefish | Wheat Nas | Plum Bonito | TBD |
| GPU Process | 7nm | 6nm | 5nm/6nm | 5nm/6nm | 5nm/6nm |
| GPU Package | Monolithic | Monolithic | MCM (1 GCD + 4 MCD) | MCM (1 GCD + 6 MCD) | MCM (TBD) |
| GPU Die Size | 520mm2 | 203mm2 (Only GCD) | 200mm2 (Only GCD) 425mm2 (with MCDs) | 300mm2 (Only GCD) 522mm2 (with MCDs) | TBD |
| Shader Engines | 4 | 2 | 3 | 6 | 8 |
| GPU WGPs | 40 | 16 | 30 | 48 | 64 |
| SPs Per WGP | 128 | 128 | 128 | 128 | 128 |
| Compute Units (Per Die) | 80 | 32 | 60 | 96 | 128 (per GPU) 256 (Total) |
| SPs (Per Die) | 5120 | 2048 | 3840 | 6144 | 4096 |
| ALUs (Per Die) | 5120 | 4096 | 7680 | 12288 | 16384 |
| Memory Bus | 256-bit | 128-bit | 256-bit | 384-bit | 384-bit x2? |
| Memory Type | GDDR6 | GDDR6 | GDDR6 | GDDR6 | GDDR6 |
| Memory Capacity | Up To 16 GB | Up To 8 GB | Up To 16 GB | Up To 24 GB | Up To 32 GB |
| Memory Speed | 16-18 Gbps | TBD | TBD | 20 Gbps | TBD |
| Memory Bandwidth | 512-576 GB/s | TBD | TBD | 960 GB/s | TBD |
| Infinity Cache | 128 MB | 32 MB | 64 MB | 96/192 MB | TBD |
| Flagship SKU | Radeon RX 6900 XTX | Radeon RX 7600 XT? Radeon RX 7700M? | Radeon RX 7800 XT? Radeon RX 7900M? | Radeon RX 7900 XTX | Radeon Pro |
| TBP (Max) | 330W | ~150W | ~250W | 335W | TBD |
| Launch | Q4 2020 | Q4 2022? | Q4 2022? | Q4 2022? | 2023? |









