‘AMD Navi chips based on RDNA 2 have 40 and 80 compute units’
AMD’s ROMc 3.8 software may contain details about the specifications of the upcoming Radeon RX 6000 cards. The software includes firmware for chips code-named Navy Flounder and Sienna Cichlid. Those would be 40 and 80 compute units to have.
The current Navi 10 GPU that AMD uses for its RX 5700 video cards is equipped with 40 compute units. The Navy Flounder chip with an equal number of CUs could be a successor to this, targeting the middle segment. The variant code-named Sienna Cichlid is said to be the top model, with 80 CUs.
Navi chips based on RDNA 2 are made on 7nm just like the first generation. A doubling of the number of compute units also means a much larger chip surface, which makes the GPU more expensive. Presumably, Sienna Cichlid will be positioned opposite Nvidia’s GeForce RTX 3080.
The code names and details are in firmware for ROCm 3.8, or RadeonOpenCompute. Reddit user stblr posted a table with the specifications. The number of compute units can be obtained from the published information by multiplying the first three rows together.
In the fifth row is the number texture channel caches and this is usually related to the width of the memory bus. If so in this case, Navy Flounder would have a 192bit memory bus and for Sienna Chichlid it would be 256bit.
Earlier this month, a photo of an alleged prototype of a Radeon RX 6000 video card with 16GB GDDR6 appeared. That card would contain eight chips, which also indicates a 256-bit memory bus. The bandwidth would thus amount to 512GB / s, which is lower than with high-end cards from Nvidia. For example, the RTX 3080 achieves 760GB / s with its 320bit memory bus and Gddr6x memory. AMD may use other techniques to compensate for the relatively low bandwidth, such as more cache memory.
AMD will present its RX 6000 video cards on October 28. It is not yet known how many cards the manufacturer will announce. AMD has shown one photo of a large card with three fans. Presumably that will be the top model. The manufacturer has not yet given concrete details about specifications. However, AMD announced earlier this year that RDNA 2 provides a 50 percent improvement in performance per watt, compared to the current RDNA architecture in the RX 5000 video cards.
| RX 5700 (XT) | RX 5300 (XT) RX 5500 (XT) | Pro 5600M | RX 6900? | RX 6800? RX 6700? | |
|---|---|---|---|---|---|
| Parameter | Navi 10 | Navi 14 | Navi 12 | Sienna Cichlid | Navy Flounder |
| gc_num_se | 2 | 1 | 2 | 4 | 2 |
| gc_num_cu_per_sh | 10 | 12 | 10 | 10 | 10 |
| gc_num_sh_per_se | 2 | 2 | 2 | 2 | 2 |
| 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 |