NVIDIA Switch GPU 20nm

The NVIDIA Switch GPU 20nm is a console-oriented graphics processor built on the Maxwell 2.0 architecture and fabricated on TSMC's 20 nm process. It integrates 2,000 million transistors into a 118 mm² die, yielding a transistor density of 16.9 million per square millimeter. The chip runs at a base clock of 384 MHz and a boost clock of 768 MHz, paired with 4 GB of DDR4 memory on a 64-bit bus. Designated as a console GPU for Nintendo, this component powers the Nintendo Switch and is now end-of-life, having been released on 2017-03-16.

Benchmark Performance

The fact pack contains no benchmark scores, no nearest rivals, and an average benchmark score of zero, so direct performance comparisons are impossible. What the data does provide are theoretical throughput figures, which offer a baseline for capability. The FP32 compute rate is 393.2 GFLOPS, while FP16 reaches 786.4 GFLOPS via a 2:1 ratio, meaning half-precision operations execute twice as fast as single-precision. This asymmetric FP16 support is typical of Maxwell 2.0, which does not include dedicated tensor cores but can still accelerate certain compute workloads through shader-based paths. The pixel fill rate is 12.29 GPixel/s, and the texture fill rate is 12.29 GTexel/s, exactly matching each other. That parity arises from the 16 ROPs and 16 TMUs operating at the same clock, indicating a balanced rasterization pipeline.

The GPU's 50th percentile ranking among all GPUs in the database places it at the median of the performance distribution. This is a useful anchor: it is neither a bottom-tier part nor a high-end unit. The low clock speeds—384 MHz base and 768 MHz boost—cap the execution rate of the 256 shading units. At boost, the FP32 throughput is 393.2 GFLOPS, which is modest by desktop standards but appropriate for a 15 W power envelope. The pixel and texture rates of 12.29 GPixel/s and 12.29 GTexel/s suggest that the GPU can handle 720p resolution at 60 frames per second for simple scenes, but 1080p would require reduced settings or lower frame rates. The absence of benchmark data means these are only theoretical limits; real-world performance depends on driver optimization and game engine demands.

The FP16 rate of 786.4 GFLOPS is notable because it doubles the compute throughput for half-precision operations, which some modern game effects use. However, without tensor cores, there is no hardware acceleration for AI-based features such as DLSS. The 50th percentile ranking likely reflects the GPU's position relative to a wide range of desktop and mobile parts, but the lack of rivals in the fact pack prevents a more granular comparison. Overall, the theoretical numbers indicate a GPU that prioritizes power efficiency over raw performance, consistent with its console role.

Ray Tracing and Feature Set

The fact pack lists no RT cores and no tensor cores, confirming that this GPU lacks hardware-accelerated ray tracing and dedicated machine learning units. Instead, it relies on the Maxwell 2.0 architecture, which supports DirectX 12 (feature level 12_1), OpenGL 4.6, and Vulkan 1.4. These APIs allow for modern rendering techniques, including asynchronous compute and explicit multi-GPU, though the hardware does not include the specialized units found in later architectures. DirectX 12_1 adds conservative rasterization and rasterizer-ordered views, which are useful for certain post-processing and transparency effects. Vulkan 1.4 support means the GPU can run current Vulkan titles, but the low compute throughput will limit visual fidelity.

Given the absence of RT cores, any ray-traced effects must be computed via shaders, which is inefficient on this hardware. The GPU's single display output is a USB Type-C port, typical for a portable console rather than a desktop card. The API feature set is modern for a 2017 release, but the lack of ray tracing and tensor hardware narrows its applicability. The Maxwell 2.0 architecture itself is known for efficient geometry processing, but without dedicated RT units, the GPU cannot handle real-time global illumination or reflections at acceptable performance levels. The Vulkan 1.4 support is a positive, as it enables cross-platform compatibility, but the hardware's limited compute power will bottleneck advanced effects.

Memory Subsystem

The memory subsystem comprises 4 GB of DDR4 memory on a 64-bit bus, delivering a bandwidth of 25.60 GB/s. This is a low figure, especially for a GPU that may output to a 1080p display. The memory clock is 1600 MHz, with an effective data rate of 3.2 Gbps. The 64-bit bus width halves the data path compared to typical desktop GPUs, which often use 128-bit or wider buses. Consequently, the GPU is constrained in how quickly it can fetch textures and geometry. For high-resolution textures or large framebuffers, this bandwidth will become a bottleneck.

The 4 GB capacity is adequate for 1080p gaming with moderate texture settings, but it is insufficient for 4K or for games that require large asset streaming. The combination of low bandwidth and modest capacity suggests that the GPU is best suited for 720p or 900p rendering, where memory pressure is lower. The DDR4 type is also slower than GDDR5 or GDDR6, further limiting performance. In practice, the memory subsystem will cap the GPU's ability to handle complex scenes, especially those with high-resolution textures or multiple render targets.

The 25.60 GB/s bandwidth is a key differentiator; it is roughly one-tenth of what a typical desktop GPU of the same era offered. This means that even if the compute units were faster, the memory would starve them. For the Nintendo Switch's target use case—portable gaming at lower resolutions—this bandwidth is acceptable, but it is a major limitation for any demanding workload. The 64-bit bus and DDR4 memory are cost-effective choices for a console, but they place a hard ceiling on performance.

FAQ

Q: What is the process node and transistor count?

A: The NVIDIA Switch GPU 20nm is fabricated on TSMC's 20 nm process and contains 2,000 million transistors on a 118 mm² die, giving a density of 16.9 million transistors per square millimeter.

Q: Does this GPU support hardware ray tracing?

A: No. The fact pack lists no RT cores and no tensor cores, so ray tracing and AI acceleration are not available in hardware. The GPU relies on Maxwell 2.0's shader-based rendering.

Q: What is the memory configuration?

A: It has 4 GB of DDR4 memory on a 64-bit bus, with a bandwidth of 25.60 GB/s. The memory clock is 1600 MHz, effective 3.2 Gbps.

Q: What APIs are supported?

A: The GPU supports DirectX 12 (feature level 12_1), OpenGL 4.6, and Vulkan 1.4.

Q: What is the TDP and power connector requirement?

A: The TDP is 15 W. No power connectors are listed, indicating an integrated design that draws power from the console motherboard.

Q: When was it released and what was the launch MSRP?

A: It was released on 2017-03-16 with a launch MSRP of 299 USD. It is now end-of-life.

Who Should Consider It

The NVIDIA Switch GPU 20nm is purpose-built for the Nintendo Switch console. Its 15 W TDP, small die size, and integrated memory make it suitable for portable and handheld gaming. Given the theoretical performance—393.2 GFLOPS FP32 and 25.60 GB/s bandwidth—it is best paired with 720p displays or 1080p at low to medium settings. The 50th percentile ranking among all GPUs in the database suggests it is neither a low-end outlier nor a high-performance part; it sits in the middle of the performance distribution. This means it can handle a wide range of titles, but not at maximum quality or high frame rates.

Gamers who prioritize portability and battery life over raw performance will find this GPU adequate for the Switch's library, which is optimized for its hardware. However, for PC gaming or any task requiring high-resolution textures or ray tracing, this GPU is not suitable. Its end-of-life status means no new drivers or support are forthcoming, so buyers should only consider it if they are acquiring a used Switch console. The lack of RT and tensor cores further limits its relevance for modern game features. In summary, this GPU is a purpose-built component for a closed ecosystem, and its specifications reflect that narrow focus.

Power and Cooling

The TDP is 15 W, which is remarkably low for a GPU with 256 shading units. This power budget allows for passive or fanless cooling in a compact console chassis. No power connectors are specified, consistent with a soldered chip that receives power from the motherboard. The dimensions of the GPU are listed as 239 mm in length, 101 mm in height, and 28 mm in width, but these likely refer to the console shell rather than the chip itself, as the die size is only 118 mm². The low TDP means that a simple heatsink and a small fan are sufficient to dissipate heat. The boost clock of 768 MHz is modest, which helps keep thermals in check. The 20 nm process from TSMC is not the most efficient by modern standards, but the low clock speeds and limited shader count keep power draw low. There is no suggested PSU because the GPU is not a standalone card; it is integrated into a console. For users who might repurpose the chip (unlikely), the 15 W figure indicates that any standard power supply would be overkill, but the lack of a standard form factor means this is not a practical consideration.