[ad_1]
NVIDIA announced its GeForce RTX 3060 12GB graphics card at CES 2021 and is expected to debut later this month. The GeForce RTX 3060 Ti, unlike existing GeForce RTX 30 products, will use a brand new Ampere GPU known as the GA106.
The GPU of the NVIDIA GeForce RTX 3060 12 GB graphics card shown, the Ampere GA106
Leaked by Videocardz, the GA106 GPU will be the latest addition to Ampere’s line of gaming GPUs. We have already seen two other GPUs in action in the gaming segment which include the GA102 and GA104 while the GA100 is positioned exclusively towards the HPC and AI segment.
An interesting thing to note here is that the particular SKU of the Ampere GPU is the GA106-400 and not the GA106-300, which NVIDIA will actually use for its GeForce RTX 3060 12GB graphics card and the reason given by the source is that this A particular chip was featured on one of the first ES cards before NVIDIA returned to the drawing board and decided to use the reduced GA106-300 matrix on the GeForce RTX 3060 graphics card. The GA106-400 GPU SKU includes a total of 3,840 CUDA cores while the GA106-300 GPU SKU includes 3,584 CUDA cores. The package size of the GA106 chip is the same as the GA104 GPU, while the die itself is much smaller.
GA106 GPU is estimated to be around 276mm2 which is even smaller than the TU116 GPU which measured 284mm2. So overall you get a lot more power in a much smaller die area than before. The TU116 GPU also lacked ray tracing hardware and this was only enabled on the TU106 GPU which measured at 445mm2 which is bigger than GA104 GPU.
Apart from this, the PCB can be seen using Samsung K4ZAF3258M-HC16 memory modules. These modules are clocked at 16 Gb / s and since they operate on a 192-bit bus interface, we will be looking at a cumulative bandwidth of 384 Gb / s. Each module also has 2GB of VRAM capacity and since there are 6 modules in total, the RTX 3060 will feature 12GB of GDDR6 memory capacity.
NVIDIA GeForce RTX 30 Series ‘Ampere’ graphics card specifications:
Graphics card name | NVIDIA GeForce RTX 3060 | NVIDIA GeForce RTX 3060 Ti | NVIDIA GeForce RTX 3070 | NVIDIA GeForce RTX 3080 | NVIDIA GeForce RTX 3090 |
---|---|---|---|---|---|
GPU name | Ampere GA106-300 | Amper GA104-200 | Ampere GA104-300 | Ampere GA102-200 | Ampere GA102-300 |
Process node | Samsung 8nm | Samsung 8nm | Samsung 8nm | Samsung 8nm | Samsung 8nm |
Die Size | TBC | 395.2 mm2 | 395.2 mm2 | 628.4 mm2 | 628.4 mm2 |
Transistors | TBC | 17.4 billion | 17.4 billion | 28 billion | 28 billion |
CUDA colors | 3584 | 4864 | 5888 | 8704 | 10496 |
TMU / ROP | 112/64 | 152/80 | 184/96 | 272/96 | 328/112 |
Tensor / RT hearts | 112/28 | 152/38 | 184/46 | 272/68 | 328/82 |
Basic clock | 1320 MHz | 1410 MHz | 1,500 MHz | 1440 MHz | 1,400 MHz |
Boost Clock | 1780 MHz | 1665 MHz | 1730 MHz | 1710 MHz | 1700 MHz |
FP32 calculation | 13 TFLOP | 16 TFLOP | 20 TFLOP | 30 TFLOP | 36 TFLOP |
RT TFLOP | 25 TFLOP | 32 TFLOP | 40 TFLOP | 58 TFLOP | 69 TFLOP |
Tensor-TOP | 101 TOP | 192 TOP | 163 TOP | 238 TOP | 285 TOP |
Memory capacity | 12 GB of GDDR6 | 8 GB GDDR6 | 8 GB GDDR6 | 10 GB GDDR6X | 24 GB GDDR6X |
Memory bus | 192 bits | 256 bits | 256 bits | 320 bits | 384 bits |
Memory speed | 16 Gbit / s | 14 Gbit / s | 14 Gbit / s | 19 Gbit / s | 19.5 Gbit / s |
Bandwidth | 384 Gbit / s | 448 Gbit / s | 448 Gbit / s | 760 Gbit / s | 936 Gbit / s |
TGP | 170 W | 175 W | 220 W | 320 W | 350 W |
Price (MSRP / FE) | $ 329 | US $ 399 | US $ 499 | US $ 699 | $ 1,499 |
Launch (availability) | End of February 2021 | December 2, 2020 | October 29, 2020 | September 17, 2020 | September 24, 2020 |
All in all, it will be interesting to see where the final performance sits with the NVIDIA GeForce RTX 3060 12GB graphics card. It will cost US $ 329 and should be as fast as the GeForce RTX 2080 with a lot more power consumption. weak and better ray tracing abilities.
[ad_2]
Source link