Hardware leaker kopite7kimi, who has a solid history of Ampere leaks, has revealed the specifications for Nvidia's pending GeForce RTX 3050. The entry-level graphics card, which is based on the Ampere architecture, is primed to contend for a spot on the list of best graphics cards.
Rumor has it that Nvidia originally planned to slip GA107 silicon inside the GeForce RTX 3050. However, kopite7kimi's latest update claims that the chipmaker has backpedaled on the decision and will reportedly use GA106 silicon instead. It's a sound choice since the GA106 is already used for the GeForce RTX 3060, so Nvidia could potentially salvage the garbage dies that don't meet the standard for the GeForce RTX 3060 and stick them in the GeForce RTX 3050.
The GeForce RTX 3050 will come in two versions with different CUDA core counts and memory capacity according to the leaker. The GA106-150 die is specifically tailored to the GeForce RTX 3050 8GB, whereas the GA106-140 die will go into the GeForce RTX 3050 4GB. Logically, the 8GB variant has more firepower out of the two.
The GA106 silicon measures 276 mm² and houses up to 30 streaming multiprocessors (SMs) that come down to 3,840 CUDA cores. The GeForce RTX 3060 has 28 enabled SMs, just two shy of the full die. To ensure a fair margin of separation and headroom for a potential GeForce RTX 3050 Ti in the future, the GeForce RTX 3050 8GB and 4GB will likely arrive with 20 and 18 SMs, respectively. That would be equivalent to 2,560 CUDA cores on the 8GB variant and 2,304 CUDA cores on the 4GB variant.
GeForce RTX 3050 Specifications
| Header Cell - Column 0 | GeForce RTX 3060 | GeForce RTX 3050 8GB* | GeForce RTX 3050 4GB* |
|---|---|---|---|
| Architecture | GA106 | GA106 | GA106 |
| Process Technology | Samsung 8N | Samsung 8N | Samsung 8N |
| Transistors (Billion) | 12 | 12 | 12 |
| Die size (mm^2) | 276 | 276 | 276 |
| SMs / CUs | 28 | 20 | 18 |
| GPU Cores | 3,584 | 2,560 | 2,304 |
| Tensor Cores | 112 | 80 | 72 |
| RT Cores | 28 | 20 | 18 |
| Base Clock (MHz) | 1,320 | ? | ? |
| Boost Clock (MHz) | 1,777 | ? | ? |
| VRAM Speed (Gbps) | 15 | 14 | 14 |
| VRAM (GB) | 12 | 8 | 4 |
| VRAM Bus Width | 192 | 128 | 128 |
| ROPs | 48 | ? | ? |
| TMUs | 112 | 80 | 72 |
| TFLOPS FP32 (Boost) | 12.7 | ? | ? |
| TFLOPS FP16 (Tensor) | 51 (102) | ? | ? |
| RT TFLOPS | 24.9 | ? | ? |
| Bandwidth (GBps) | 360 | 224 | 224 |
| TDP (Watts) | 170 | 90 | 90 |
| Launch Date | Feb 21 | Jan 22 | Jan 22 |
| Launch Price | $329 | ? | ? |
*Specifications are unconfirmed.
Like the GeForce RTX 3060, the GeForce RTX 3050 will presumably utilize GDDR6 memory. However, the latter could use lower clocked memory chips, maybe 14 Gbps instead of the 15 Gbps that Nvidia uses on the former. We should also expect a more narrow memory interface. The odds are that the GeForce RTX 3050 will sport a 128-bit memory bus, which amounts to a maximum memory bandwidth of 224 GBps.
We haven't seen any benchmark leaks for the GeForce RTX 3050 yet. However, being a competitor in the entry-level segment, the GeForce RTX 3050 will take on AMD's also unreleased Radeon RX 6500 XT and Intel's upcoming Arc graphics cards.
The rumor going around town is that Nvidia may launch the GeForce RTX 3050 around CES 2022. However, the Ampere-based graphics card won't hit the shelves until January 27. In addition, while the GeForce RTX 3060 debuted at $329, it's uncertain if Nvidia will keep the GeForce RTX 3050 under $250.
