AMD Instinct MI350X vs NVIDIA GeForce RTX 5090 Mobile Comparison

The Verdict

The recorded database comparison places two fundamentally different compute devices side by side. The AMD Instinct MI350X is a server-oriented accelerator with a 1000 W TDP, while the NVIDIA GeForce RTX 5090 Mobile is a 95 W laptop GPU. The data shows no direct head-to-head benchmark results, so the comparison rests on architectural specifications and the available benchmark scores for the RTX 5090 Mobile only.

For compute-heavy server workloads, the MI350X is the clear choice from the data. It delivers 72.09 TFLOPS FP32 and 72.09 TFLOPS FP16 (1:1), compared to 31.80 TFLOPS for both precision levels on the RTX 5090 Mobile. The MI350X also carries 288 GB of HBM3e memory across an 8192-bit bus, yielding 8.19 TB/s of bandwidth. The RTX 5090 Mobile has 24 GB of GDDR7 on a 256-bit bus, producing 896.0 GB/s. These are orders of magnitude apart in memory capacity and bandwidth.

For mobile gaming and portable graphics, the RTX 5090 Mobile is the only option with display outputs and API support. The MI350X has no display outputs and lists DirectX, OpenGL, and Vulkan as N/A. The RTX 5090 Mobile supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. Its benchmark scores place it at the 84th percentile among all GPUs, with an average benchmark score of 45152. The MI350X has no recorded benchmarks and sits at the 50th percentile with an average score of 0.

The verdict from the data is straightforward: the MI350X targets massive parallel compute with enormous memory resources, while the RTX 5090 Mobile targets portable graphics with established API support and measurable performance in standard benchmarks. Neither device replaces the other; they serve separate markets.

Where Each One Wins

The MI350X wins in raw compute throughput. Its FP32 performance of 72.09 TFLOPS is more than double the RTX 5090 Mobile's 31.80 TFLOPS. The FP16 figures mirror this exactly, with both devices running FP16 at a 1:1 ratio to FP32. Texture rate also favors the MI350X: 2,252.8 GTexel/s versus 496.9 GTexel/s on the RTX 5090 Mobile. The MI350X has 16,384 shading units, 1,024 TMUs, and no ROPs, while the RTX 5090 Mobile has 10,496 shading units, 328 TMUs, and 112 ROPs.

Memory capacity and bandwidth are also decisive wins for the MI350X. With 288 GB of HBM3e and 8.19 TB/s bandwidth, it dwarfs the RTX 5090 Mobile's 24 GB GDDR7 at 896.0 GB/s. The MI350X uses an 8192-bit bus versus 256-bit on the mobile part. For workloads that scale with memory size, such as large model inference or scientific simulation, the MI350X has a massive advantage.

The RTX 5090 Mobile wins in graphics-oriented features and portability. It has 82 RT cores and 328 tensor cores, while the MI350X lists no RT or tensor core counts. Pixel rate on the RTX 5090 Mobile is 169.7 GPixel/s, while the MI350X shows 0 MPixel/s. The RTX 5090 Mobile supports modern graphics APIs, while the MI350X shows N/A for all three listed APIs. The mobile GPU also has a 95 W TDP against 1000 W for the MI350X, making it feasible for laptops. The MI350X is an OAM module, while the RTX 5090 Mobile is an IGP (integrated graphics processor) designed for portable devices.

Benchmark data only exists for the RTX 5090 Mobile. Its best scores include 201834 in Geekbench OpenCL, 198405 in Geekbench Vulkan, 30034 in Passmark G3D, and 5871 in 3DMark Steel Nomad DX12. These results confirm its capability in real-world graphics workloads. The MI350X has no benchmark entries, so its wins are purely specification-based.

Architecture Differences

The MI350X uses the CDNA 4.0 architecture built on a 3 nm TSMC process. It packages 185,000 million transistors on a 2380 mm² die, giving a transistor density of 77.7M per mm². The chip is designated MI350 256CU. This is a massive die optimized for compute throughput without traditional graphics output.

The RTX 5090 Mobile uses the Blackwell 2.0 architecture on a 5 nm TSMC process. It contains 45,600 million transistors on a 378 mm² die, yielding a transistor density of 120.6M per mm². The chip is GB203. The higher transistor density on the smaller process reflects a different design philosophy: more logic per area for graphics and tensor workloads.

Clock speeds differ significantly. The MI350X has a 1000 MHz base clock and 2200 MHz boost clock. The RTX 5090 Mobile has a 990 MHz base and 1515 MHz boost. Despite the lower boost clock, the mobile GPU achieves substantial performance per watt due to its 95 W TDP. Memory clocks also differ: the MI350X runs at 2000 MHz with 8 Gbps effective, while the RTX 5090 Mobile runs at 1750 MHz with 28 Gbps effective. The higher effective data rate on GDDR7 compensates somewhat for the narrower 256-bit bus.

The MI350X has no ROPs and no pixel rate, confirming it is not designed for rasterization. The RTX 5090 Mobile has 112 ROPs and a pixel rate of 169.7 GPixel/s. The MI350X also lacks RT cores and tensor cores in the recorded data, while the RTX 5090 Mobile has 82 RT cores and 328 tensor cores. This structural difference defines their roles: the MI350X for pure compute, the RTX 5090 Mobile for graphics and AI acceleration in a portable form.

Power delivery also diverges. The MI350X has a 1000 W TDP and requires a 1400 W suggested PSU, with no power connectors listed (OAM modules use board-level power). The RTX 5090 Mobile has a 95 W TDP and no suggested PSU, as it draws power from the host laptop. Both use PCIe 5.0 x16 interfaces, though the mobile part operates within a laptop's constrained power envelope.

FAQ

Q: Which GPU has higher FP32 compute?

A: The AMD Instinct MI350X delivers 72.09 TFLOPS FP32, more than double the NVIDIA GeForce RTX 5090 Mobile's 31.80 TFLOPS.

Q: How much memory does each GPU have?

A: The MI350X has 288 GB of HBM3e with 8.19 TB/s bandwidth. The RTX 5090 Mobile has 24 GB of GDDR7 with 896.0 GB/s bandwidth.

Q: Does the MI350X support DirectX or Vulkan?

A: No. The MI350X lists DirectX, OpenGL, and Vulkan as N/A. The RTX 5090 Mobile supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Q: What is the power consumption difference?

A: The MI350X has a 1000 W TDP with a 1400 W suggested PSU. The RTX 5090 Mobile has a 95 W TDP and no suggested PSU.

Q: Which GPU has RT cores?

A: The RTX 5090 Mobile has 82 RT cores and 328 tensor cores. The MI350X lists no RT core or tensor core counts.

Q: What benchmark scores exist for the RTX 5090 Mobile?

A: The database records scores including 201834 in Geekbench OpenCL, 198405 in Geekbench Vulkan, 30034 in Passmark G3D, and 5871 in 3DMark Steel Nomad DX12. Its average benchmark score is 45152.

Head-to-Head Benchmarks

No direct head-to-head benchmark results exist in the database for these two devices. The winsA and winsB fields are both 0. This absence reflects their different market positions: the MI350X has no recorded benchmarks, while the RTX 5090 Mobile has ten benchmark entries.

The RTX 5090 Mobile's nearest rivals provide context for its performance. The AMD Radeon Pro 5500 XT scores 45384 on average, 0.5% higher than the RTX 5090 Mobile's 45152. The NVIDIA GeForce RTX 4070 Ti scores 44795, 0.8% lower. The Intel Arc A730M scores 45592, 1.0% higher. The NVIDIA RTX 5880 Ada Generation scores 45972, 1.8% higher. These deltas show the RTX 5090 Mobile sits in a tight competitive band despite being a laptop part.

The RTX 5090 Mobile's percentile rank of 84 among all GPUs indicates strong overall performance. Its individual scores vary by workload: 3DMark Steel Nomad DX12 at 5871, Geekbench OpenCL at 201834, Geekbench Vulkan at 198405, Passmark DirectX 10 at 183, DirectX 11 at 269, DirectX 12 at 138, DirectX 9 at 324, G2D at 1057, G3D at 30034, and GPU Compute at 13401. The high Geekbench scores suggest strong compute capability, while the Passmark DirectX scores are lower, reflecting the laptop GPU's power constraints under sustained load.

For the MI350X, the specification-based advantages are clear: 2.27 times the FP32 throughput, 9.14 times the memory bandwidth, 12 times the memory capacity, and 4.53 times the texture rate compared to the RTX 5090 Mobile. These ratios come directly from the recorded figures. The MI350X also has 1.56 times the shading units and 3.12 times the TMUs.

The RTX 5090 Mobile counters with 112 ROPs versus 0, a 169.7 GPixel/s pixel rate versus 0, and a 95 W TDP versus 1000 W. Its transistor density of 120.6M per mm² exceeds the MI350X's 77.7M per mm², indicating more efficient use of silicon area. The mobile GPU also has a higher memory data rate at 28 Gbps effective versus 8 Gbps effective.

Specification Differences

| Specification | AMD Instinct MI350X | NVIDIA GeForce RTX 5090 Mobile |

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

| Architecture | CDNA 4.0 | Blackwell 2.0 |

| Process Node | 3 nm | 5 nm |

| Foundry | TSMC | TSMC |

| Transistors | 185,000 million | 45,600 million |

| Die Size | 2380 mm² | 378 mm² |

| Transistor Density | 77.7M / mm² | 120.6M / mm² |

| Base Clock | 1000 MHz | 990 MHz |

| Boost Clock | 2200 MHz | 1515 MHz |

| Memory Clock | 2000 MHz, 8 Gbps effective | 1750 MHz, 28 Gbps effective |

| Memory Size | 288 GB | 24 GB |

| Memory Type | HBM3e | GDDR7 |

| Memory Bus Width | 8192 bit | 256 bit |

| Memory Bandwidth | 8.19 TB/s | 896.0 GB/s |

| Shading Units | 16384 | 10496 |

| TMUs | 1024 | 328 |

| ROPs | 0 | 112 |

| RT Cores | None listed | 82 |

| Tensor Cores | None listed | 328 |

| Pixel Rate | 0 MPixel/s | 169.7 GPixel/s |

| Texture Rate | 2,252.8 GTexel/s | 496.9 GTexel/s |

| FP32 | 72.09 TFLOPS | 31.80 TFLOPS |

| FP16 | 72.09 TFLOPS (1:1) | 31.80 TFLOPS (1:1) |

| TDP | 1000 W | 95 W |

| Slot Width | OAM Module | IGP |

| Suggested PSU | 1400 W | None |

| Bus Interface | PCIe 5.0 x16 | PCIe 5.0 x16 |

| Display Outputs | No outputs | Portable Device Dependent |

| DirectX | N/A | 12 Ultimate (12_2) |

| OpenGL | N/A | 4.6 |

| Vulkan | N/A | 1.4 |

| Release Date | 2025-06-11 | 2025-03-26 |

| Predecessor | Radeon Instinct | GeForce 40 Mobile |

| Percentile vs All GPUs | 50 | 84 |

| Average Benchmark Score | 0 | 45152 |

The release dates show the RTX 5090 Mobile launched earlier on 2025-03-26, while the MI350X followed on 2025-06-11. The RTX 5090 Mobile has a production status of "Active," while the MI350X has no recorded status. Both use TSMC as the foundry and PCIe 5.0 x16 as the bus interface. The MI350X has no power connectors listed, while the RTX 5090 Mobile also has none, relying on the host platform for power delivery.