Nvidia RTX 3070, RTX 3080 Leaked Specifications Hint Up To 20GB VRAM

Nvidia might announce its next-generation GPUs in the next few months. According to a leak, the new graphics cards will be based on Ampere architecture and are expected to be much faster compared to any of the GPUs that are currently available in the market.

As of now, two upcoming high-end Nvidia GPUs -- the Nvidia RTX 3070 (GA103) and the RTX 3080's (GA104) specifications have been leaked, hinting towards a huge leap in performance when compared to the RTX 20 series of GPUs. These GPUs are expected to offer a whopping 16GB and 20GB of DDR6 based VRAM, respectively.

Bit More About Ampere Architecture

According to leaks, the Ampere architecture used on the upcoming Nvidia GPUs is like to offer improved rasterization and shading performance compared to the previous generation GPUs. Besides, this technology is also expected to improve the ray-tracing performance.

GPUs based on Ampere architecture will use PCIe Gen 4.0 to offer higher bandwidth and they are likely to include support for NVLINK, which enables us to use two GPUs at the same time. The company is likely to use 7nm fabrication that makes these GPUs energy efficient compared to the first-gen ray-tracing GPUs.

Nvidia GeForce RTX 3070 Leaked Specifications

According to leaks, the RTX 3070 will offer 16GB of VRAM with 256bit memory bandwidth. It is speculated that the GPU will offer 3072 CUDA cores (48 streaming multiprocessors units), which is identical to the RTX 2080. Having a 256bit bus allows the brand to offer VRAM of up to 16GB, which was limited to 8GB on the RTX 2080. Having video memory as high as 16GB might also drive up the overall price of the GPU.

Nvidia GeForce RTX 3080 Leaked Specifications

The RTX 3080 is likely to come with a 320bit bus/memory interface that allows the company to offer up to 20GB of DDR6 based VRAM. This GPU is expected to offer 60 streaming multiprocessors units with a total of 3840 CUDA cores.

First Gen And Second-Gen RTX GPUs Comparision

GPU Name	NVIDIA GeForce RTX 3070 (GA104)	NVIDIA GeForce RTX 3080 (GA104)	NVIDIA GeForce RTX 2070 (TU106)	NVIDIA GeForce RTX 2080 (TU104)	NVIDIA GeForce RTX 2080 Ti (TU102)
Architecture	Ampere	Ampere	Turing	Turing	Turing
Process	7nm TSMC (Expected)	7nm TSMC (Expected)	12nm FinFET NVIDIA (TSMC)	12nm FinFET (TSMC)	12nm FinFET (TSMC)
No Of Streaming Multiprocessor Units	48	60	36	48	72
No CUDA Cores	3072 CCs	3840 CCs	2304 CCs	3072 CCs	4608 CCs
Memory Size And Type	8/16 GB GDDR6	10/20 GB GDDR6	8 GB GDDR6	8 GB GDDR6	11 GB GDDR6
Memory Bus	256-bit (Expected)	320-bit (Expected)	256-bit	256-bit	320-bit
Memory Speed	16 Gbps (Expected)	16 Gbps (Expected)	14 Gbps	14 Gbps	14 Gbps
Memory Bandwidth	512 GB/s (Expected)	640 GB/s (Expected)	448 GB/s	448 GB/s	640 GB/s
Launch	2020	2020	2018	2018	2018