AMD Steam Machine GPU vs NVIDIA Jetson Orin Nano Super Comparison

Head-to-Head Benchmarks

The recorded data shows no direct head-to-head benchmark results between the AMD Steam Machine GPU and the NVIDIA Jetson Orin Nano Super. Both products have an average benchmark score of zero in the database, with no individual benchmark entries. Their percentile rankings against all GPUs are identical at 50, indicating both sit at the midpoint of the performance distribution when measured against the full catalog of recorded graphics processors.

The absence of benchmark data does not mean these products are equivalent. The raw specifications reveal a substantial performance gap in several key areas. The AMD Steam Machine GPU delivers 17.56 TFLOPS of FP32 compute, which is 8.4 times the 2.089 TFLOPS offered by the NVIDIA Jetson Orin Nano Super. In FP16 workloads, the AMD part maintains 17.56 TFLOPS with a 1:1 ratio, while the NVIDIA chip reaches 4.178 TFLOPS using a 2:1 ratio. The AMD GPU holds a 4.2 times advantage in FP16 throughput.

Texture and pixel processing show similar disparities. The AMD Steam Machine GPU achieves 274.4 GTexel/s and 156.8 GPixel/s, compared to 32.64 GTexel/s and 16.32 GPixel/s for the NVIDIA Jetson Orin Nano Super. The AMD part is 8.4 times faster in texture fill and 9.6 times faster in pixel fill. These figures indicate that the AMD GPU can drive far higher resolution displays and apply more complex texture filtering without becoming the bottleneck.

The shading architecture reinforces this gap. The AMD chip contains 1792 shading units, 112 texture mapping units, and 64 render output units. The NVIDIA part has 1024 shading units, 32 TMUs, and 16 ROPs. This means the AMD GPU has 75% more shader processors, 3.5 times more TMUs, and 4 times more ROPs. The AMD GPU also includes 28 ray tracing cores, while the NVIDIA product has none listed. Conversely, the NVIDIA Jetson Orin Nano Super includes 32 tensor cores, a feature absent from the AMD Steam Machine GPU.

Memory bandwidth differences further separate the two. Both use 8 GB of memory on a 128 bit bus, but the AMD GPU employs GDDR6 at 18 Gbps effective, producing 288.0 GB/s of bandwidth. The NVIDIA chip uses LPDDR5 at 6.4 Gbps effective, yielding 102.4 GB/s. The AMD GPU provides 2.8 times the memory bandwidth, which directly impacts high-resolution texture streaming and compute workloads that repeatedly access large datasets.

The Verdict

The data indicates two fundamentally different product categories. The AMD Steam Machine GPU is a dedicated console graphics processor designed for rasterization-heavy gaming workloads. The NVIDIA Jetson Orin Nano Super is an integrated graphics solution on a Tegra-class system-on-chip, optimized for edge AI inference with its tensor cores and constrained to a 25 W power envelope.

For pure graphics performance, the AMD Steam Machine GPU is the clear choice. It delivers 17.56 TFLOPS FP32 compute, 288.0 GB/s memory bandwidth, dedicated ray tracing hardware, and a 110 W TDP that allows sustained high clock rates. The 2450 MHz boost clock and 2250 MHz game clock indicate a part engineered for demanding real-time rendering.

The NVIDIA Jetson Orin Nano Super targets a different workload profile. Its 32 tensor cores and 4.178 TFLOPS FP16 performance with a 2:1 ratio suggest a focus on neural network inference, where reduced precision is acceptable and tensor operations dominate. The 25 W TDP makes it suitable for embedded or portable deployments where power draw is the primary constraint.

Users building a gaming system should select the AMD Steam Machine GPU. Users developing AI inference applications for edge devices should select the NVIDIA Jetson Orin Nano Super. The performance gap in traditional graphics is too large to ignore, but the NVIDIA part wins on power efficiency and tensor processing capability, both absent from the AMD specification sheet.

FAQ

Q: Which GPU has higher FP32 compute performance?

A: The AMD Steam Machine GPU delivers 17.56 TFLOPS FP32, which is 8.4 times the 2.089 TFLOPS of the NVIDIA Jetson Orin Nano Super.

Q: Do both GPUs support the same graphics APIs?

A: Yes, both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4 according to the recorded data.

Q: What is the memory bandwidth difference?

A: The AMD Steam Machine GPU provides 288.0 GB/s via GDDR6 memory, while the NVIDIA Jetson Orin Nano Super provides 102.4 GB/s via LPDDR5. The AMD part has 2.8 times more bandwidth.

Q: Which GPU has ray tracing hardware?

A: The AMD Steam Machine GPU includes 28 ray tracing cores. The NVIDIA Jetson Orin Nano Super does not list any ray tracing cores in its specifications.

Q: How does the transistor count compare?

A: The AMD Steam Machine GPU uses 13,300 million transistors on a 204 mm² die. The NVIDIA Jetson Orin Nano Super does not disclose its transistor count, though its die size is 200 mm².

Q: What is the power draw difference?

A: The AMD Steam Machine GPU has a 110 W TDP. The NVIDIA Jetson Orin Nano Super has a 25 W TDP, making it 4.4 times more power-efficient in this metric.

Specification Differences

The AMD Steam Machine GPU and NVIDIA Jetson Orin Nano Super differ across nearly every specification field. The AMD chip uses a 6 nm process from TSMC, while the NVIDIA part uses an 8 nm process from Samsung. The AMD die measures 204 mm² with 13,300 million transistors and a density of 65.2M per mm². The NVIDIA die measures 200 mm² with an undisclosed transistor count and density.

Clock speeds show clear separation. The AMD GPU has a 1720 MHz base clock, 2450 MHz boost clock, and 2250 MHz game clock. Its memory runs at 2250 MHz with 18 Gbps effective data rate. The NVIDIA part lists no base, boost, or game clock values, only a memory clock of 800 MHz with 6.4 Gbps effective.

Memory type differs: GDDR6 for AMD versus LPDDR5 for NVIDIA. Both use 8 GB capacity on a 128 bit bus, but the AMD bandwidth of 288.0 GB/s far exceeds the NVIDIA 102.4 GB/s. The AMD GPU has 1792 shading units, 112 TMUs, 64 ROPs, and 28 ray tracing cores. The NVIDIA part has 1024 shading units, 32 TMUs, 16 ROPs, no ray tracing cores, and 32 tensor cores.

The AMD GPU has no tensor cores listed. The NVIDIA GPU has no ray tracing cores listed. Pixel rate is 156.8 GPixel/s for AMD versus 16.32 GPixel/s for NVIDIA. Texture rate is 274.4 GTexel/s versus 32.64 GTexel/s. FP32 is 17.56 TFLOPS versus 2.089 TFLOPS. FP16 is 17.56 TFLOPS (1:1) versus 4.178 TFLOPS (2:1).

Power consumption favors NVIDIA at 25 W TDP versus 110 W TDP for AMD. The NVIDIA part is an integrated graphics processor (IGP) with a PCIe 4.0 x4 bus interface. The AMD GPU lists no slot width, power connectors, or bus interface. Display outputs differ: AMD provides 1x HDMI 2.1a and 1x DisplayPort 2.1, while NVIDIA outputs are described as portable device dependent.

Physical dimensions diverge significantly. The AMD GPU measures 156 mm length, 152 mm height, and 162 mm width. The NVIDIA part measures 70 mm length and 45 mm height. The AMD GPU released on 2026-06-28, while the NVIDIA product released on 2024-12-16. The NVIDIA launch MSRP is 249 USD; no launch MSRP is recorded for the AMD part. Both remain in active production.

Architecture Differences

The AMD Steam Machine GPU uses the RDNA 3.0 architecture on the Navi 33 chip with the codename Hotpink Bonefish. This is a console GPU generation for Valve. The NVIDIA Jetson Orin Nano Super uses the Ampere architecture on the GA10B chip, belonging to the Tegra generation. The AMD architecture is built on a 6 nm process from TSMC, while the NVIDIA architecture uses an 8 nm process from Samsung.

The compute architectures reflect different design priorities. RDNA 3.0 uses a unified shader design with 1792 shading units operating at a 1:1 FP32 to FP16 ratio, meaning both precision levels run at the same throughput. Ampere on the Tegra platform uses a 2:1 FP16 to FP32 ratio, doubling FP16 throughput relative to FP32. This suggests the NVIDIA architecture prioritizes mixed-precision workloads where FP16 is acceptable.

Ray tracing support exists only on the AMD side with 28 dedicated ray tracing cores. The NVIDIA part omits ray tracing hardware entirely, focusing instead on 32 tensor cores for matrix math acceleration. Tensor cores handle neural network operations such as convolution and matrix multiplication, which are the foundation of modern AI inference. The AMD GPU has no tensor cores, limiting its AI acceleration to general FP32 or FP16 compute.

Memory architecture also differs. The AMD GPU uses GDDR6, a dedicated graphics memory standard with high bandwidth per pin. The NVIDIA part uses LPDDR5, a low-power memory standard designed for mobile and embedded systems. Both use a 128 bit bus, but the GDDR6 implementation achieves 288.0 GB/s versus 102.4 GB/s for LPDDR5. The lower bandwidth reflects the NVIDIA part's lower power budget and embedded target.

The AMD GPU is a discrete component with its own power delivery, as indicated by the absence of power connectors and a 110 W TDP. The NVIDIA part is an IGP, meaning it integrates directly into a system-on-chip package. Its 25 W TDP and portable device dependent display outputs confirm an embedded design philosophy.

Where Each One Wins

The AMD Steam Machine GPU wins decisively in every traditional graphics metric. Its 17.56 TFLOPS FP32 compute is 8.4 times higher, its 288.0 GB/s memory bandwidth is 2.8 times higher, and its 156.8 GPixel/s pixel rate is 9.6 times higher. The 28 ray tracing cores provide hardware acceleration for ray-traced lighting and reflections, a feature entirely absent from the NVIDIA part. For gaming at high resolutions, texture-heavy workloads, or any application that demands raw rasterization throughput, the AMD GPU is the only viable option between the two.

The NVIDIA Jetson Orin Nano Super wins on power efficiency and tensor processing. Its 25 W TDP is 4.4 times lower than the AMD 110 W TDP, making it suitable for battery-powered or thermally constrained devices. The 32 tensor cores provide dedicated hardware for AI inference, a capability the AMD GPU lacks. Its 4.178 TFLOPS FP16 performance, while lower than AMD's 17.56 TFLOPS, is achieved at a fraction of the power draw. The PCIe 4.0 x4 interface allows direct integration into embedded systems, and the compact 70 mm by 45 mm footprint fits in space-constrained designs.

The AMD GPU wins for desktop or console gaming, where performance per watt is secondary to absolute frame rates and visual fidelity. The NVIDIA part wins for edge AI deployment, robotics, or portable devices where the total system power budget is limited and the workload centers on neural network inference rather than polygon rendering. The data does not support one product being universally superior; the correct choice depends entirely on whether the priority is graphics compute or efficient AI processing.