GeekRadar / The comparison desk
COMPARE.
GPU
NVIDIA

NVIDIA Switch 2 GPU

Graphics processor

Video memory
12 GB*
Memory bus
128 bit*
Rated memory bandwidth
102.40 GB/s*

No matched benchmark results are available for this entry yet.

Processing & platform

Architecture
Ampere*
Process node
8 nm*
GPU cores
1,536*
Reported chip die area
200 mm²*
GPU L2 cache
4 MB*
Graphics processor
GA10B*
Texture mapping units
48*
Raster output units
16*
Streaming multiprocessors
12*
Tensor cores
48*
Ray tracing cores
12*
Chip foundry
Samsung*
Declared GPU TDP
40 W*
DirectX / feature level
12 Ultimate (12_2)*
OpenGL
4.6*
OpenCL
3.0*
Vulkan
1.4*
CUDA compute capability
8.7*
Shader model
6.8*

Memory & storage

Memory options
12*
Video memory
12 GB*
Video memory type
LPDDR5X*
Memory bus
128 bit*
Rated memory bandwidth
102.40 GB/s*

Power specifications

Thermal design power
40 W*

System configuration

Detailed specifications

Specifications 1 · 26 fields
proc die size
200* mm²
cores
1,536*
tensor cores
48*
rt cores
12*
sm count
12*
base clock
561*
boost clock
1,400*
memory clock
800*
ram
12*
name
NVIDIA Switch 2 GPU*
ram type
LPDDR5X*
bus width
128*
ram bandwidth
102.4* GB/s
fp16 performance
8.602*
fp32 performance
4.301*
fp64 performance
2.15*
tdp
40* W
api cuda
8.7*
is retail board
No*
gpi value
3.22*
vendor
nvd*
manufacturer
NVIDIA*
arch name
Ampere*
card release date
2025-06-05*
proc foundry
Samsung*
proc process size
8* nm
Specifications 2 · 35 fields
CUDA compute capability
8.7*
DirectX / feature level
12 Ultimate (12_2)*
OpenCL
3.0*
OpenGL
4.6*
Shader model
6.8*
Vulkan
1.4*
Architecture
Ampere*
Memory bus
128* bit
Reported shader processors
1,536*
Reported theoretical FP16 peak
8.602* TFLOPS
Scope and conditions
Arithmetic density
Not stated*
Reported theoretical FP32 peak
4.301* TFLOPS
Scope and conditions
Arithmetic density
Not stated*
Reported theoretical FP64 peak
2.15* TFLOPS
Scope and conditions
Arithmetic density
Not stated*
Declared L1 cache configuration
128 KB (per SM)*
Declared GPU L2 cache
4 MB*
Reported display outputs
No outputs*
Theoretical pixel fill rate
22.40 GPixel/s*
Reported chip die area
200 mm²*
Chip foundry
Samsung*
Process node declaration
8 nm*
Graphics processor
GA10B*
Reported video memory
12* GB
Rated memory bandwidth
102.4 GB/s*
Video memory type
LPDDR5X*
Raster output units
16*
Ray tracing cores
12*
Reported shading units
1,536*
Streaming multiprocessors
12*
2:4 tensor sparsity
Yes*
Declared GPU TDP
40 W*
Tensor cores
48*
Theoretical texture fill rate
67.20 GTexel/s*
Texture mapping units
48*
GPU cores
1,536*
Video memory
12* GB
Video memory type
LPDDR5X*

* Provisional data