Please note that this post is tagged as a rumor.
AMD Sienna Cichlid and Navy Flounder specifications
AMD RDNA2-based graphics cards (Rumored Radeon RX 6700/6800 – left, Rumored RX 6900 series – right)
The next-generation desktop graphics processors based on AMD RDNA2 architecture are expected to debut on October 28th. AMD did not confirm which GPUs will be unveiled but the manufacturer did announce it will reveal Radeon RX 6000 series, so there are at least two cards coming.
AMD is allegedly developing three Navi 2X GPUs: Navi 21, Navi 22, and Navi 23. Only Navi 21 (Big Navi) has been confirmed to carry a codename of Sienna Cichlid, while the Navy Flounder codename could refer to either 23 or 22.
It is expected that the higher the Navi suffix the lower the GPU will be positioned in the lineup. Thus, the Navi 21 is going to be the top tier GPU, while 23 is likely to land in an entry-level space.
Only leaks Sienna Cichlid and Navy Flounder leaks have been appearing in open-source software patches, such as Linux kernel or, like in this example, ROCm software.
Reddit stblr discovered that the recent ROCm update (3.8) features firmware for Navy Flounder. This file contains variables defining the configuration of the graphics processor.
The Navi Flounder is believed to offer 40 Compute Units, just as many as Navi 10, but the memory bus width is rumored to be limited to 192-bit. This is not confirmed, of course, but it is a result of the simple multiplication of three variables present in the firmware data. Assuming that each Compute Unit carries exactly 64 Stream Processors, the Navy Flounder would feature 2560 unified cores.
Unless there is something that we do not know yet, the GPU would land in a mid-range space, competing directly with yet unannounced NVIDIA Ampere GA106 GPU (also debuting with the 192-bit bus).
| AMD Navi 2x (RDNA2) vs Navi 1x (RDNA 1) Graphics Processors | |||||
|---|---|---|---|---|---|
| Parameter | Navi 10 | Navi 14 | Navi 12 | Sienna Cichlid | Navy Flounder |
| gc_num_se (1) | 2 | 1 | 2 | 4 | 2 |
| gc_num_cu_per_sh (2) | 10 | 12 | 10 | 10 | 10 |
| gc_num_sh_per_se (3) | 2 | 2 | 2 | 2 | 2 |
| Compute Units (1 x 2 x 3) | 40 | 24 | 40 | 80 | 40 |
| gc_num_rb_per_se | 8 | 8 | 8 | 4 | 4 |
| gc_num_tccs | 16 | 8 | 16 | 16 | 12 |
| gc_num_gprs | 1024 | 1024 | 1024 | 1024 | 1024 |
| gc_num_max_gs_thds | 32 | 32 | 32 | 32 | 32 |
| gc_gs_table_depth | 32 | 32 | 32 | 32 | 32 |
| gc_gsprim_buff_depth | 1792 | 1792 | 1792 | 1792 | 1792 |
| gc_double_offchip_lds_buffer | 1024 | 512 | 1024 | 1024 | 1024 |
| gc_wave_size | 32 | 32 | 32 | 32 | 32 |
| gc_max_waves_per_simd | 20 | 20 | 20 | 16 | 16 |
| gc_lds_size | 64 | 64 | 64 | 64 | 64 |
| num_sc_per_sh | 1 | 1 | 1 | 1 | 1 |
| num_packer_per_sc | 2 | 2 | 2 | 4 | 4 |
Source: Reddit via @TUM_APISAK
