AMD Steam Deck OLED GPU vs NVIDIA GeForce RTX 5090 SE Comparison

Head-to-Head Benchmarks

The database contains no recorded benchmark scores or head-to-head measurements for either the AMD Steam Deck OLED GPU or the NVIDIA GeForce RTX 5090 SE. The head-to-head benchmark table is empty, with zero wins recorded for either part. This is not a case of one part dominating the other; rather, the comparative metric set has not been populated with any executable workloads, synthetic tests, or frame-rate measurements.

Without measured scores, the only quantitative basis for comparison is the raw specification data and the percentile ranking. Both GPUs sit at the 50th percentile against all GPUs in the database. That shared percentile is a consequence of the database's ranking methodology, not an indication of performance parity. A percentile alone does not encode the spread of scores around it, and with zero benchmark entries for both items, the percentile field is a placeholder rather than an observed result.

The recorded average benchmark score for both parts is 0, which reinforces that no test data has been ingested. The practical takeaway: any claim of a specific performance lead or deficit between these two products cannot be supported by the database at this time. The analysis must rest on architectural and specification differences, which are substantial.

Architecture Differences

The AMD Steam Deck OLED GPU and the NVIDIA GeForce RTX 5090 SE are engineered for entirely different segments, and the architectural gap is visible across every major subsystem.

The AMD part uses the Sephiroth chip, built on RDNA 2.0 architecture, manufactured by TSMC on a 6 nm process. The die measures 131 mm² and contains 2,400 million transistors, giving a transistor density of 18.3 million per mm². The NVIDIA part uses the GB202 chip, built on Blackwell 2.0 architecture, also from TSMC but on a 5 nm process. The die is 750 mm² with 92,200 million transistors, yielding a density of 122.9 million per mm². The transistor density difference is stark: NVIDIA packs roughly seven times more transistors into a die that is about 5.7 times larger. That density gap reflects the newer process and the radically different design goals.

Shader resources tell a similar story. The Steam Deck OLED GPU has 512 shading units, 32 texture mapping units, 16 ROPs, and 8 ray tracing cores, with no tensor cores listed. The RTX 5090 SE has 14,080 shading units, 440 TMUs, 160 ROPs, 110 ray tracing cores, and 440 tensor cores. The NVIDIA part also has a 1:1 FP16 to FP32 ratio, delivering 66.94 TFLOPS in both precisions, while the AMD part delivers 3.277 TFLOPS FP16 (2:1) and 1.638 TFLOPS FP32. The raw compute gap is approximately 41x in FP32, though the two products target different power envelopes and use cases.

Clocks differ as well. The AMD GPU runs at a 1000 MHz base and 1600 MHz boost. The NVIDIA GPU runs at 1740 MHz base and 2377 MHz boost. Memory clocks are 1375 MHz (11 Gbps effective) on the AMD side versus 1750 MHz (28 Gbps effective) on the NVIDIA side. The memory subsystems are not comparable in class: 16 GB of LPDDR5 on a 128 bit bus delivers 176.0 GB/s, while 24 GB of GDDR7 on a 384 bit bus delivers 1.34 TB/s.

API support shows a split. Both support DirectX 12 Ultimate (12_2) and OpenGL 4.6. The AMD part supports Vulkan 1.3; the NVIDIA part supports Vulkan 1.4. Display outputs also differ: the AMD part lists a single USB Type-C output, while the NVIDIA part lists 1x HDMI 2.1b and 3x DisplayPort 2.1b.

Physical specifications diverge sharply. The AMD part draws 15 W TDP, has no listed slot width, no power connector, and no suggested PSU. The NVIDIA part draws 500 W TDP, is dual-slot, uses a single 16-pin power connector, and requires a 900 W suggested PSU. Dimensions: the AMD part is 298 mm long, 117 mm tall, and 49 mm wide; the NVIDIA part is 267 mm long, 111 mm tall, and 40 mm wide.

The AMD part has no listed bus interface; the NVIDIA part uses PCIe 5.0 x16. The AMD part's release date is 2023-11-08; the NVIDIA part's release date is 2025-12-31. The NVIDIA part lists a predecessor (GeForce 40) and successor (GeForce 60); the AMD part lists neither. The NVIDIA launch MSRP is 1,499 USD. The AMD part has no launch MSRP listed.

FAQ

Q: Which GPU has more shading units?

A: The NVIDIA GeForce RTX 5090 SE has 14,080 shading units, while the AMD Steam Deck OLED GPU has 512. That is a 27.5x difference in shader count.

Q: What is the memory bandwidth of each GPU?

A: The AMD Steam Deck OLED GPU delivers 176.0 GB/s from 16 GB of LPDDR5 on a 128 bit bus. The NVIDIA GeForce RTX 5090 SE delivers 1.34 TB/s from 24 GB of GDDR7 on a 384 bit bus.

Q: Do both GPUs support the same DirectX version?

A: Yes. Both are listed as DirectX 12 Ultimate (12_2). The Vulkan support differs: AMD lists Vulkan 1.3, NVIDIA lists Vulkan 1.4.

Q: What is the power consumption difference?

A: The AMD part is rated at 15 W TDP. The NVIDIA part is rated at 500 W TDP. The NVIDIA part also lists a suggested PSU of 900 W; the AMD part lists no suggested PSU.

Q: How do the process nodes compare?

A: Both are TSMC parts, but the AMD GPU uses a 6 nm process and the NVIDIA GPU uses a 5 nm process. Transistor density is 18.3 million per mm² for AMD versus 122.9 million per mm² for NVIDIA.

Q: Has either GPU been benchmarked in the database?

A: No. Both have an average benchmark score of 0 and zero entries in the head-to-head benchmark table. The percentile for both is 50, but no measured test data is present.

The Verdict

The database does not contain any benchmark results for either GPU, so a performance verdict cannot be derived from measured data. The specification sheet, however, makes the intended roles clear.

The AMD Steam Deck OLED GPU is a 15 W console-class part. It uses RDNA 2.0 on a 6 nm process, has 512 shading units, 16 GB of LPDDR5 on a 128 bit bus, and a single USB Type-C display output. It is physically large (298 mm long) despite the low power draw, which reflects the handheld device integration rather than a desktop card form factor. The RTX 5090 SE is a 500 W desktop flagship with 14,080 shading units, 24 GB of GDDR7 on a 384 bit bus, 440 tensor cores, 110 ray tracing cores, and a dual-slot cooler with a 16-pin power connector.

A buyer choosing between these two is not comparing two options for the same task. The Steam Deck OLED GPU is a low-power, integrated-class design for a portable console; the RTX 5090 SE is a high-power, discrete desktop GPU for maximum compute throughput. The data supports that distinction through TDP, memory bandwidth, shader count, and physical form factor. The RTX 5090 SE exceeds the AMD part on every raw compute and memory metric in the specification set, but it also draws over 33 times the power and requires a 900 W suggested PSU. The AMD part is the only one of the two that fits a 15 W thermal envelope.

The lack of benchmark scores means neither part can be declared faster in any specific workload. The recorded data confirms the architectural and specification gap, nothing more. For applications that need FP32 throughput, tensor cores, or high-bandwidth GDDR7, the RTX 5090 SE is the only viable option in this comparison. For a sub-20 W embedded or handheld use case, the Steam Deck OLED GPU is the only viable option. The verdict is a function of the target environment, not of measured performance.

Specification Differences

| Field | AMD Steam Deck OLED GPU | NVIDIA GeForce RTX 5090 SE |

| --- | --- | --- |

| Architecture | RDNA 2.0 | Blackwell 2.0 |

| Process Node | 6 nm | 5 nm |

| Foundry | TSMC | TSMC |

| Transistors | 2,400 million | 92,200 million |

| Die Size | 131 mm² | 750 mm² |

| Transistor Density | 18.3M / mm² | 122.9M / mm² |

| Base Clock | 1000 MHz | 1740 MHz |

| Boost Clock | 1600 MHz | 2377 MHz |

| Memory Clock | 1375 MHz, 11 Gbps effective | 1750 MHz, 28 Gbps effective |

| Memory Size | 16 GB | 24 GB |

| Memory Type | LPDDR5 | GDDR7 |

| Memory Bus Width | 128 bit | 384 bit |

| Memory Bandwidth | 176.0 GB/s | 1.34 TB/s |

| Shading Units | 512 | 14,080 |

| TMUs | 32 | 440 |

| ROPs | 16 | 160 |

| RT Cores | 8 | 110 |

| Tensor Cores | None listed | 440 |

| Pixel Rate | 25.60 GPixel/s | 380.3 GPixel/s |

| Texture Rate | 51.20 GTexel/s | 1,045.9 GTexel/s |

| FP32 | 1.638 TFLOPS | 66.94 TFLOPS |

| FP16 | 3.277 TFLOPS (2:1) | 66.94 TFLOPS (1:1) |

| TDP | 15 W | 500 W |

| Slot Width | Not listed | Dual-slot |

| Power Connectors | Not listed | 1x 16-pin |

| Suggested PSU | Not listed | 900 W |

| Bus Interface | Not listed | PCIe 5.0 x16 |

| Display Outputs | 1x USB Type-C | 1x HDMI 2.1b, 3x DisplayPort 2.1b |

| DirectX | 12 Ultimate (12_2) | 12 Ultimate (12_2) |

| OpenGL | 4.6 | 4.6 |

| Vulkan | 1.3 | 1.4 |

| Length | 298 mm (11.7 inches) | 267 mm (10.5 inches) |

| Height | 117 mm (4.6 inches) | 111 mm (4.4 inches) |

| Width | 49 mm (1.9 inches) | 40 mm (1.6 inches) |

| Release Date | 2023-11-08 | 2025-12-31 |

| Production Status | Active | Active |

| Predecessor | None listed | GeForce 40 |

| Successor | None listed | GeForce 60 |

| Launch MSRP | None listed | 1,499 USD |