RaiseADog
Cheap Galaxy Buds Live deals: Save up to $35 on Samsung’s true wireless earbuds today only
Technical specs of AMD’s upcoming Navi 21, 22 and 23 GPUs have leaked out by way of reddit user stblr (by way of Videocardz). That is an absolute motherlode of a leak that the consumer has managed to place along with what seems to be painstaking accuracy. The knowledge was reverse-engineered from drivers current within the MacOs and is probably going legit. We is not going to be marking this as a rumor as we have now no cause to imagine this info is inaccurate in any means.
Navi 21: 80 CUs, 2.2 GHz increase clock and 22.5 TFLOPs of compute
From the desk we are able to see that the Navi 21 GPU (doubtless the RX 6900 XT) can have 80 CUs. Assuming the ratio of CU to SP stays the identical we’re 5120 stream processors. Now Navi 21 seems to have two additional variants, 21a and 21b with the previous clocking as much as 2.2 GHz. This is able to put the only precision compute efficiency of the playing cards at a strong 22.5 TFLOPs. For comparability, the NVIDIA RTX 3080 has a single precision compute efficiency of 29.Eight TFLOPs. This card, if priced appropriately by AMD may very well be a an absolute winner. 22 TFLOPs of energy is greater than sufficient for the typical 4k gamer and one thing the 16GB of vRAM would make it tear via AAA video games. AMD does neeed to work on a DLSS various nonetheless.
Navi 22: 40 CUs, 2.5 GHz increase clock and 12.Eight TFLOPs of compute
The youthful brother of flagship, the Navi 22 GPU can have 40 CUs. That is (assuming the identical CU to SP ratio) 2560 shader cores. Impressively nonetheless, it seems to be like we’re lastly going to get a mainstream 2.5 GHz GPU with Navi 22. The excessive clock charge additionally helps the cardboard punch means above its weight class and clock in a large 12.Eight TFLOPs of compute. That is efficiency that’s barely greater than the Xbox Sequence X and greater than sufficient for 4k gaming (when you dont thoughts dialing down the settings just a bit bit).
There’s a Navi 23 current as properly which can have 32 CUs (2048 cores). There may be nonetheless, no extra info and no clock pace info – which doubtless means this GPU remains to be within the pipeline.
| Property | Navi 10 | Navi 14 | Navi 12 | Navi 21 | Navi 22 | Navi 23 | Navi 31 |
|---|---|---|---|---|---|---|---|
| num_se | 2 | 1 | 2 | four | 2 | 2 | four |
| num_cu_per_sh | 10 | 12 | 10 | 10 | 10 | Eight | 10 |
| num_sh_per_se | 2 | 2 | 2 | 2 | 2 | 2 | 2 |
| num_rb_per_se | Eight | Eight | Eight | four | four | four | four |
| num_tccs | 16 | Eight | 16 | 16 | 12 | Eight | 16 |
| num_gprs | 1024 | 1024 | 1024 | 1024 | 1024 | 1024 | 1024 |
| num_max_gs_thds | 32 | 32 | 32 | 32 | 32 | 32 | 32 |
| gs_table_depth | 32 | 32 | 32 | 32 | 32 | 32 | 32 |
| gsprim_buff_depth | 1792 | 1792 | 1792 | 1792 | 1792 | 1792 | 1792 |
| parameter_cache_depth | 1024 | 1024 | 1024 | 1024 | 1024 | 1024 | 1024 |
| double_offchip_lds_buffer | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| wave_size | 32 | 32 | 32 | 32 | 32 | 32 | 32 |
| max_waves_per_simd | 20 | 20 | 20 | 16 | 16 | 16 | 16 |
| max_scratch_slots_per_cu | 32 | 32 | 32 | 32 | 32 | 32 | 32 |
| lds_size | 64 | 64 | 64 | 64 | 64 | 64 | 64 |
| num_sc_per_sh | 1 | 1 | 1 | 1 | 1 | 1 | 1 |
| num_packer_per_sc | 2 | 2 | 2 | four | four | four | four |
| num_gl2a | N/A | N/A | N/A | four | 2 | 2 | four |
| unknown0 | N/A | N/A | N/A | 10 | 10 | Eight | 10 |
| unknown1 | N/A | N/A | N/A | 16 | 12 | Eight | 16 |
| unknown2 | N/A | N/A | N/A | 80 | 40 | 32 | 80 |
| num_cus (computed) | 40 | 24 | 40 | 80 | 40 | 32 | 80 |
| Property | Navi 10 a | Navi 10 b | Navi 14 | Navi 21 a | Navi 21 b | Navi 22 |
|---|---|---|---|---|---|---|
| gfxclk (MHz) | 300 – 1000 | 300 – 1000 | 300 – 1900 | 500 – 2050 | 500 – 2050 | 500 – 2500 |
| uclk (MHz) | 100 – 750 | 100 – 750 | 100 – 875 | 577 – 1089 | 577 – 1089 | 577 – 1089 |
| socket_power_limit_ac[0] (W) | 180 | 180 | 110 | 200 | 238 | 170 |
| freq_table_gfx[0] (MHz) | 300 | 300 | 300 | 500 | 500 | 500 |
| freq_table_gfx[1] (MHz) | 1400 | 1400 | 1900 | 2050 | 2200 | 2500 |
| freq_table_uclk[0] (MHz) | 124 | 100 | 100 | 100 | 100 | 97 |
| freq_table_uclk[1] (MHz) | 500 | 500 | 500 | 500 | 500 | 457 |
| freq_table_uclk[2] (MHz) | 625 | 625 | 625 | 625 | 625 | 674 |
| freq_table_uclk[3] (MHz) | 875 | 750 | 875 | 1000 | 1000 | 1000 |
AMD APUs
The leaker additionally launched a ton of data on AMD’s upcoming APUs:
| Property | Renoir | Cezanne | Van Gogh | Rembrandt |
|---|---|---|---|---|
| num_se | 1 | 1 | 1 | 1 |
| num_cu_per_sh | Eight | Eight | Eight | 6 |
| num_sh_per_se | 1 | 1 | 1 | 2 |
| num_rb_per_se | 2 | 2 | 2 | four |
| num_tccs | four | four | four | four |
| num_gprs | 256 | 256 | 1024 | 1024 |
| num_max_gs_thds | 32 | 32 | 32 | 32 |
| gs_table_depth | 32 | 32 | 32 | 32 |
| gsprim_buff_depth | 1792 | 1792 | 1792 | 1792 |
| parameter_cache_depth | 1024 | 1024 | 512 | 256 |
| double_offchip_lds_buffer | 1 | 1 | 1 | 1 |
| wave_size | 64 | 64 | 32 | 32 |
| max_waves_per_simd | 10 | 10 | 16 | 16 |
| max_scratch_slots_per_cu | 32 | 32 | 32 | 32 |
| lds_size | 64 | 64 | 64 | 64 |
| num_sc_per_sh | N/A | 1 | 1 | 1 |
| num_packer_per_sc | N/A | 2 | 2 | four |
| num_gl2a | N/A | N/A | four | four |
| unknown0 | N/A | N/A | Eight | 6 |
| unknown1 | N/A | N/A | four | four |
| unknown2 | N/A | N/A | Eight | 12 |
| num_cus (computed) | Eight | Eight | Eight | 12 |
What we learn about AMD’s Radeon RX 6000 sequence GPUs up to now
We’ve got had some fairly dependable leaks on the RX 6000 sequence and the specs are apparently the next: The Navi 21 primarily based Radeon RX 6900 XT can have 16GB of vRAM whereas the Navi 22 powered GPU (which could be known as the RX 6800 XT or the RX 6700 XT) can have 12 GB of vRAM. Principally, each tiers of playing cards are going to have extra reminiscence than the RTX 3080. Please needless to say whereas the GPU kind (Navi 21/22 and so forth) and the vRAM configurations have been kind of confirmed, the nomenclature may shift barely relying on what AMD needs.
The RX 6000 sequence flagship: the RX 6900 XT (aka massive Navi) is a dual-slot card whose design has already been teased by the corporate. The reference design has an LED-lit Radeon brand and blows air out the aspect as a substitute of out the again. The cardboard will function a twin Eight-pin connector, USB Kind C, a single HDMI connection and two DisplayPort connections. The fan design is a triple axial-tech fan with a big aluminum heatsink cooling all of the internals. The cooling design is unquestionably a step up from the final era and will permit for a card that runs cooler with out compromising any efficiency. It additionally includes a backplate together with a retention bracket.
The AMD Radeon RX 6800/6700 sequence however will function a white-black aesthetic (on the reference design) and will likely be a slighly compact type issue in comparison with the unique. It includes a twin axial-tech fan design with a big R brand branded within the center and the identical LED-lit Radeon brand on the aspect. Identical to its elder brother, it additionally seems to blow scorching air out the aspect. This card, apparently additionally options twin Eight-pin connectors though AIBs may determined to tone this down to eight+6 as they see match. This GPU will home the smaller Navi 22 card and can have a wealthy IO set that’s the very same as the massive Navi variant, specifically: a twin Eight-pin connector, USB Kind C, a single HDMI connection and two DisplayPort connections. AMD’s RX 6000 sequence will debut on October 28th, 2020 as a tough launch.