AMD Instinct MI300 vs NVIDIA GeForce RTX 5090 Mobile Comparison

FAQ

Q: What are the release dates of the AMD Instinct MI300 and the NVIDIA GeForce RTX 5090 Mobile?

A: The AMD Instinct MI300 was released on January 3, 2023. The NVIDIA GeForce RTX 5090 Mobile was released on March 26, 2025.

Q: Which GPU has more memory, and what type of memory does each use?

A: The AMD Instinct MI300 has 128 GB of HBM3 memory on an 8192-bit bus. The NVIDIA GeForce RTX 5090 Mobile has 24 GB of GDDR7 memory on a 256-bit bus.

Q: What is the TDP difference between these two GPUs?

A: The AMD Instinct MI300 has a TDP of 600 W and requires a 1000 W suggested power supply with 2x 8-pin connectors. The NVIDIA GeForce RTX 5090 Mobile has a TDP of 95 W, uses no external power connectors, and has no suggested PSU listed.

Q: What is the percentile ranking for each GPU in the database?

A: The NVIDIA GeForce RTX 5090 Mobile sits at the 84th percentile among all GPUs. The AMD Instinct MI300 sits at the 50th percentile.

Q: Does the AMD Instinct MI300 have any benchmark scores recorded?

A: No, the database shows no benchmark entries for the AMD Instinct MI300. Its average benchmark score is 0. The NVIDIA GeForce RTX 5090 Mobile has an average benchmark score of 45,152.

Q: What API support does each GPU offer?

A: The AMD Instinct MI300 lists N/A for DirectX, OpenGL, and Vulkan. The NVIDIA GeForce RTX 5090 Mobile supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Architecture Differences

The AMD Instinct MI300 uses the Aqua Vanjaram chip built on CDNA 3.0 architecture, manufactured by TSMC on a 5 nm process. The NVIDIA GeForce RTX 5090 Mobile uses the GB203 chip built on Blackwell 2.0 architecture, also manufactured by TSMC on a 5 nm process. Both GPUs share the same process node and foundry, but their underlying designs diverge significantly.

Transistor counts differ substantially. The MI300 packs 153,000 million transistors on a die size of 1017 mm², yielding a transistor density of 150.4M per mm². The RTX 5090 Mobile has 45,600 million transistors on a 378 mm² die, giving a density of 120.6M per mm². The MI300 has roughly 3.35 times the transistor count and a die nearly 2.7 times larger.

The MI300 uses 14,080 shading units, 880 texture mapping units, and has 0 ROPs. It has no dedicated ray tracing cores and no tensor cores listed. The RTX 5090 Mobile uses 10,496 shading units, 328 TMUs, and 112 ROPs. It includes 82 ray tracing cores and 328 tensor cores. This structural difference reflects the MI300's compute-oriented design versus the RTX 5090 Mobile's graphics and AI feature set.

Memory architecture also diverges completely. The MI300 uses 128 GB of HBM3 across an 8192-bit bus, delivering 5.32 TB/s of bandwidth. The RTX 5090 Mobile uses 24 GB of GDDR7 across a 256-bit bus, delivering 896.0 GB/s. The MI300's memory bandwidth is roughly 5.9 times higher.

Clock behavior differs as well. The MI300 has a base clock of 1000 MHz and a boost clock of 1700 MHz. The RTX 5090 Mobile has a base clock of 990 MHz and a boost clock of 1515 MHz. Memory clocks are 1300 MHz (5.2 Gbps effective) for the MI300 and 1750 MHz (28 Gbps effective) for the RTX 5090 Mobile.

The MI300 has no display outputs, while the RTX 5090 Mobile lists its outputs as "Portable Device Dependent." The MI300 uses a PCIe 5.0 x16 interface and measures 267 mm by 111 mm. The RTX 5090 Mobile is an IGP (integrated graphics package) with no listed dimensions. Power delivery differs sharply: the MI300 draws 600 W with 2x 8-pin connectors, while the RTX 5090 Mobile draws 95 W with no power connectors.

Head-to-Head Benchmarks

The database contains no head-to-head benchmark entries between these two GPUs. The AMD Instinct MI300 has zero benchmark scores recorded, while the NVIDIA GeForce RTX 5090 Mobile has ten individual benchmark results.

The RTX 5090 Mobile's strongest results come from compute and API-specific tests. In Geekbench OpenCL, it scores 201,834. In Geekbench Vulkan, it scores 198,405. Its Passmark G3D score is 30,034, and its Passmark GPU Compute score is 13,401. In DirectX tests, it scores 324 in Passmark DirectX 9, 269 in DirectX 11, 183 in DirectX 10, and 138 in DirectX 12. Its 3DMark Steel Nomad DX12 score is 5,871, and its Passmark G2D score is 1,057.

Because the MI300 has no recorded benchmark scores, direct performance comparisons cannot be made from the database. The RTX 5090 Mobile's average benchmark score of 45,152 places it at the 84th percentile. Its nearest rivals in the database include the AMD Radeon Pro 5500 XT with an average score of 45,384 and a delta of -0.5%, the NVIDIA GeForce RTX 4070 Ti with an average score of 44,795 and a delta of 0.8%, the Intel Arc A730M with an average score of 45,592 and a delta of -1%, and the NVIDIA RTX 5880 Ada Generation with an average score of 45,972 and a delta of -1.8%. These deltas indicate the RTX 5090 Mobile is within 2% of all four rivals, slightly ahead of the RTX 4070 Ti and slightly behind the other three.

The MI300's 50th percentile ranking with an average benchmark score of 0 indicates it has not been evaluated in the same benchmark suite, likely due to its server-oriented role rather than any performance deficiency.

Specification Differences

| Specification | AMD Instinct MI300 | NVIDIA GeForce RTX 5090 Mobile |

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

| Architecture | CDNA 3.0 | Blackwell 2.0 |

| Chip | Aqua Vanjaram | GB203 |

| Process node | 5 nm | 5 nm |

| Transistors | 153,000 million | 45,600 million |

| Die size | 1017 mm² | 378 mm² |

| Transistor density | 150.4M / mm² | 120.6M / mm² |

| Base clock | 1000 MHz | 990 MHz |

| Boost clock | 1700 MHz | 1515 MHz |

| Memory size | 128 GB | 24 GB |

| Memory type | HBM3 | GDDR7 |

| Memory bus | 8192 bit | 256 bit |

| Memory bandwidth | 5.32 TB/s | 896.0 GB/s |

| Memory clock | 1300 MHz 5.2 Gbps effective | 1750 MHz 28 Gbps effective |

| Shading units | 14,080 | 10,496 |

| TMUs | 880 | 328 |

| ROPs | 0 | 112 |

| RT cores | None | 82 |

| Tensor cores | None | 328 |

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

| Texture rate | 1,496.0 GTexel/s | 496.9 GTexel/s |

| FP32 | 47.87 TFLOPS | 31.80 TFLOPS |

| FP16 | 47.87 TFLOPS (1:1) | 31.80 TFLOPS (1:1) |

| TDP | 600 W | 95 W |

| Power connectors | 2x 8-pin | None |

| Suggested PSU | 1000 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 |

| Dimensions | 267 mm x 111 mm | Not listed |

| Release date | 2023-01-03 | 2025-03-26 |

| Production status | Not listed | Active |

The Verdict

The database presents two GPUs with fundamentally different purposes. The AMD Instinct MI300 is a 600 W accelerator with 128 GB of HBM3 memory, 14,080 shading units, and 47.87 TFLOPS of FP32 throughput. It has no display outputs, no graphics API support, and no benchmark scores. Its 50th percentile ranking with zero average score reflects an absence of consumer-facing test data rather than a measured performance level.

The NVIDIA GeForce RTX 5090 Mobile is a 95 W mobile graphics processor with 24 GB of GDDR7 memory, 10,496 shading units, 82 ray tracing cores, and 328 tensor cores. It delivers 31.80 TFLOPS of FP32 and has a full graphics API stack including DirectX 12 Ultimate, OpenGL 4.6, and Vulkan 1.4. Its average benchmark score of 45,152 places it at the 84th percentile, with nearest rivals within roughly 2% in either direction.

For a builder choosing between these two, the decision rests entirely on workload type. The MI300 targets compute-heavy environments that need massive memory capacity and bandwidth, with no rendering or display capability. The RTX 5090 Mobile targets portable systems that need graphics output, ray tracing, and AI acceleration within a 95 W power envelope. The MI300 has 5.9 times the memory bandwidth and 3.1 times the texture fill rate, but the RTX 5090 Mobile has 112 ROPs versus zero for the MI300, enabling pixel output that the MI300 cannot perform at all.

The data indicates the RTX 5090 Mobile is the only one of the two with measured performance in the database. Its 84th percentile ranking suggests strong overall standing among all GPUs. The MI300's absence of benchmark results means its relative position cannot be quantified, only its architectural specifications.

Where Each One Wins

Large memory workloads: The MI300 wins decisively with 128 GB of HBM3 versus 24 GB of GDDR7. Its 5.32 TB/s bandwidth dwarfs the RTX 5090 Mobile's 896.0 GB/s. Any workload that fits in memory capacity or depends on sustained bandwidth favors the MI300.

Pure compute throughput: The MI300 delivers 47.87 TFLOPS in both FP32 and FP16, versus 31.80 TFLOPS for the RTX 5090 Mobile. The MI300 also has 1,496.0 GTexel/s texture rate versus 496.9 GTexel/s. For shader-heavy or texture-heavy compute, the MI300 has the higher theoretical ceiling.

Graphics rendering: The RTX 5090 Mobile wins outright. It has 112 ROPs and a pixel rate of 169.7 GPixel/s, while the MI300 has 0 ROPs and a 0 MPixel/s pixel rate. The MI300 cannot output frames; the RTX 5090 Mobile is designed for exactly that.

Ray tracing and AI acceleration: The RTX 5090 Mobile includes 82 ray tracing cores and 328 tensor cores. The MI300 lists neither. Applications using ray tracing or tensor-accelerated features rely on the NVIDIA GPU.

Power efficiency: The RTX 5090 Mobile draws 95 W versus 600 W for the MI300. The mobile GPU needs no external power connectors and has no suggested PSU requirement, while the MI300 requires 2x 8-pin connectors and a 1000 W power supply.

Software and API compatibility: The RTX 5090 Mobile supports DirectX 12 Ultimate, OpenGL 4.6, and Vulkan 1.4. The MI300 lists N/A for all three. Consumer and professional graphics software expect these APIs, giving the RTX 5090 Mobile a clear compatibility advantage.

Measured performance ranking: The RTX 5090 Mobile has a recorded average benchmark score of 45,152 and an 84th percentile rank. The MI300 has no recorded scores and sits at the 50th percentile by default. The database currently only validates the NVIDIA part's performance through actual benchmarks.