Rumors have been circulating that NVIDIA plans to release two versions of the upcoming GeForce RTX 5060 Ti - one with 8GB of VRAM and one with 16 GB. With the latest leak from industry insider @kopite7kimi on X, we now have specs for the upcoming GPU confirming two models: a GeForce RTX 5060 Ti with 8GB of GDDR7 memory on a 128-bit bus and one with 16GB of GDDR7 memory on the same 128-bit bus.
The GB206-300-A1 RTX Blackwell chip will power both cards and ship with 4608 CUDA Cores. This is a modest 6% increase in hardware compared to the GeForce RTX 4060 Ti, which features 4352 CUDA Cores. The GeForce RTX 5070 shipped with a 4% increase in CUDA Cores compared to the GeForce RTX 4070 and delivered 20% faster performance.
So, with these specs, what can we expect from the GeForce RTX 5060 Ti? Assuming RTX Blackwell performance scaling remains the same, the lower the stack we go, the GeForce RTX 5060 Ti could deliver up to GeForce RTX 4070 performance levels.
Of course, the value proposition will come down to price, and the hope is NVIDIA doesn't follow the GeForce RTX 4060 Ti launch which saw its 16GB variant launch for $499 USD compared to the baseline 8GB model's $399 price point. The shift to faster GDDR7 memory could be the show's star here as having the new cutting-edge memory running on a 128-bit memory interface should lead to a much higher memory bandwidth of up to 448 GB/sec - a massive 55% gen-on-gen increase.
Both GeForce RTX 5060 Ti cards are rated at 180W, 20W higher than the GeForce RTX 40 Series versions. NVIDIA's GeForce RTX 5060 Series is rumored to launch in April and May. According to some, NVIDIA could announce the line-up as early as this week, so stay tuned.
@kopite7kimi also provided specs for the upcoming GeForce RTX 5050, a new budget 130W model from NVIDIA with GDDR6 memory - something we haven't seen since the RTX 30 Series.
Here's a breakdown of the specs.
GPU | GeForce RTX 5050 | GeForce RTX 5060 | GeForce RTX 5060 Ti | GeForce RTX 5070 |
---|---|---|---|---|
Architecture | Blackwell | Blackwell | Blackwell | Blackwell |
Process | TSMC 4N | TSMC 4N | TSMC 4N | TSMC 4N |
Chip | GB207 | GB206 | GB206 | GB205 |
CUDA Cores | 2560 | TBC | 4608 | 6144 |
Tensor Cores (AI) | TBC | TBC | TBC | 192 (5th Gen) |
RT Cores | TBC | TBC | TBC | 48 (4th Gen) |
Boost Clock | TBC | TBC | TBC | 2510 MHz |
Memory | 8GB GDDR6 | 8GB GDDR7 | 8GB/16GB GDDR7 | 12GB GDDR7 |
Memory Interface | 128-bit | 128-bit | 128-bit | 192-bit |
Bandwidth | TBC | 448 GB/s | 448 GB/s | 672 GB/s |
TGP | 130W | 145W | 180W | 250W |
Price | TBC | TBC | TBC | $549 |
Availability | TBC | TBC | TBC | Mar-25 |