AMD Instinct MI350P vs NVIDIA GeForce RTX 5090 D Comparison

Head-to-Head Benchmarks

The recorded data shows no direct head-to-head benchmark results between the AMD Instinct MI350P and the NVIDIA GeForce RTX 5090 D. The database contains no shared test entries for these two accelerators, so a direct score comparison is not possible from the available measurements. The MI350P carries a percentile ranking of 50 among all GPUs in the database, with an average benchmark score of zero, indicating that no benchmark submissions exist for this part. The RTX 5090 D, in contrast, holds a percentile ranking of 92 and an average benchmark score of 77,712 across ten recorded tests.

The RTX 5090 D delivers its strongest measured result in the Geekbench Vulkan test with a score of 376,915, followed by Geekbench OpenCL at 310,674. Its Passmark G3D score of 44,065 and Passmark GPU Compute score of 28,396 further characterize its compute-oriented performance. The 3DMark Steel Nomad DX12 test returns 14,326 points. These figures place the card well above the nearest rivals recorded in the database: the AMD Radeon RX 6650M XT sits 1.1% below the 5090 D with an average score of 76,904, while the AMD Radeon RX 6850M XT trails by 1.6% at 78,940. The NVIDIA Tesla P100 PCIe 12 GB and 16 GB variants sit 2.1% and 2.4% below, respectively.

Because the MI350P has no benchmark entries, the data cannot confirm any performance advantage in either direction. The wins tally stands at zero for both parts. The comparison therefore rests entirely on architectural and specification differences rather than measured performance deltas.

Architecture Differences

The two accelerators represent fundamentally different design philosophies. The AMD Instinct MI350P uses the CDNA 4.0 architecture on a 3 nm TSMC process, while the NVIDIA GeForce RTX 5090 D uses the Blackwell 2.0 architecture on a 5 nm TSMC process. Both chips come from TSMC, but the process node difference is substantial: 3 nm versus 5 nm.

Transistor counts and die sizes diverge sharply. The MI350P packs 73,000 million transistors on a 1,190 mm² die, yielding a transistor density of 61.3 million per square millimeter. The RTX 5090 D carries 92,200 million transistors on a 750 mm² die, achieving a much higher density of 122.9 million per square millimeter. The MI350P die is larger by 440 mm², yet the 5090 D contains roughly 26% more transistors, a direct consequence of the denser design.

The chip configurations differ completely. The MI350P is built around 8,192 shading units, 512 texture mapping units, and no ROPs. It has no dedicated ray tracing cores and no tensor cores listed. The RTX 5090 D uses 21,760 shading units, 680 TMUs, 176 ROPs, 170 ray tracing cores, and 680 tensor cores. The shading unit count alone gives the 5090 D a 2.66x advantage in raw shader hardware.

Clock behavior also separates the two. The MI350P runs at a 1,000 MHz base and 2,200 MHz boost, while the RTX 5090 D operates at 2,017 MHz base and 2,407 MHz boost. The NVIDIA part sustains a higher boost clock by 207 MHz and starts from more than double the base clock. Memory clocks show a similar pattern: the MI350P uses 2,000 MHz memory with 8 Gbps effective data rate, while the 5090 D runs 1,750 MHz memory with 28 Gbps effective data rate. The effective rate discrepancy reflects the different memory technologies.

The MI350P is an accelerator without display outputs, while the RTX 5090 D provides 1x HDMI 2.1b and 3x DisplayPort 2.1b. API support follows the same split: the MI350P lists N/A for DirectX, OpenGL, and Vulkan, whereas the 5090 D supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

FAQ

Q: Which product has more memory bandwidth?

A: The AMD Instinct MI350P provides 8.19 TB/s of bandwidth from 144 GB of HBM3e memory on an 8,192-bit bus. The NVIDIA GeForce RTX 5090 D provides 1.79 TB/s from 32 GB of GDDR7 memory on a 512-bit bus. The MI350P delivers 4.58x the bandwidth.

Q: Why does the RTX 5090 D have a higher benchmark percentile despite lower memory capacity?

A: The RTX 5090 D holds a 92nd percentile ranking with an average benchmark score of 77,712 across ten recorded tests. The MI350P has no recorded benchmarks and sits at the 50th percentile with an average score of zero, so the database contains no measured performance for the AMD part.

Q: What are the power requirements for each card?

A: The MI350P has a 600 W TDP and a suggested 1,000 W power supply, while the RTX 5090 D has a 575 W TDP and a suggested 950 W power supply. Both use a single 16-pin power connector and occupy a dual-slot form factor.

Q: Which card supports DirectX and Vulkan?

A: Only the RTX 5090 D supports these APIs, with DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The MI350P lists N/A for all three, consistent with its lack of display outputs and compute-focused design.

Q: How do the physical dimensions compare?

A: The MI350P measures 267 mm in length, 111 mm in height, and 40 mm in width. The RTX 5090 D measures 304 mm in length, 137 mm in height, and 48 mm in width. The NVIDIA card is longer by 37 mm, taller by 26 mm, and wider by 8 mm.

Q: Which card has the higher boost clock?

A: The RTX 5090 D boosts to 2,407 MHz, which is 207 MHz higher than the MI350P boost of 2,200 MHz. The NVIDIA card also has a much higher base clock at 2,017 MHz versus 1,000 MHz for the AMD part.

Specification Differences

| Specification | AMD Instinct MI350P | NVIDIA GeForce RTX 5090 D |

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

| Architecture | CDNA 4.0 | Blackwell 2.0 |

| Process Node | 3 nm | 5 nm |

| Transistors | 73,000 million | 92,200 million |

| Die Size | 1190 mm² | 750 mm² |

| Transistor Density | 61.3M / mm² | 122.9M / mm² |

| Base Clock | 1000 MHz | 2017 MHz |

| Boost Clock | 2200 MHz | 2407 MHz |

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

| Memory Size | 144 GB | 32 GB |

| Memory Type | HBM3e | GDDR7 |

| Memory Bus Width | 8192 bit | 512 bit |

| Memory Bandwidth | 8.19 TB/s | 1.79 TB/s |

| Shading Units | 8192 | 21760 |

| TMUs | 512 | 680 |

| ROPs | 0 | 176 |

| RT Cores | None listed | 170 |

| Tensor Cores | None listed | 680 |

| Pixel Rate | 0 MPixel/s | 423.6 GPixel/s |

| Texture Rate | 1,126.4 GTexel/s | 1,636.8 GTexel/s |

| FP32 | 36.04 TFLOPS | 104.8 TFLOPS |

| FP16 | 36.04 TFLOPS (1:1) | 104.8 TFLOPS (1:1) |

| TDP | 600 W | 575 W |

| Suggested PSU | 1000 W | 950 W |

| Display Outputs | No outputs | 1x HDMI 2.1b, 3x DisplayPort 2.1b |

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

| OpenGL | N/A | 4.6 |

| Vulkan | N/A | 1.4 |

| Length | 267 mm 10.5 inches | 304 mm 12 inches |

| Height | 111 mm 4.4 inches | 137 mm 5.4 inches |

| Width | 40 mm 1.6 inches | 48 mm 1.9 inches |

| Release Date | 2026-05-06 | 2025-01-29 |

| Production Status | Not listed | Active |

The FP32 throughput difference is notable: the RTX 5090 D reaches 104.8 TFLOPS versus 36.04 TFLOPS for the MI350P, a 2.91x gap. Texture rate favors the NVIDIA card at 1,636.8 GTexel/s versus 1,126.4 GTexel/s. Pixel rate is effectively absent on the MI350P at 0 MPixel/s, while the 5090 D delivers 423.6 GPixel/s. The MI350P compensates with 4.58x the memory bandwidth and 4.5x the memory capacity.

Where Each One Wins

The AMD Instinct MI350P wins decisively on memory-oriented specifications. Its 144 GB of HBM3e memory dwarfs the 32 GB of GDDR7 on the RTX 5090 D, and its 8.19 TB/s bandwidth is 4.58x higher. The 8,192-bit bus width supports this throughput in a way no consumer-oriented GDDR7 implementation can match. The MI350P also carries a larger die at 1,190 mm² and uses a more advanced 3 nm process. These characteristics position it for workloads where memory capacity and bandwidth dominate, such as large model inference or data-intensive compute tasks. The absence of display outputs and API support reinforces its role as a headless accelerator.

The NVIDIA GeForce RTX 5090 D wins on every measured performance metric in the database. It holds the 92nd percentile ranking with an average benchmark score of 77,712, while the MI350P has no recorded scores. The 5090 D delivers 2.91x the FP32 throughput at 104.8 TFLOPS, 2.66x the shading units at 21,760, and higher clocks across the board. Its 176 ROPs, 170 ray tracing cores, and 680 tensor cores provide functionality the MI350P does not list at all. The 5090 D also supports DirectX 12 Ultimate, OpenGL 4.6, and Vulkan 1.4, making it usable in graphics and gaming environments. Its 423.6 GPixel/s pixel rate and 1,636.8 GTexel/s texture rate confirm strong rasterization capabilities.

The transistor density comparison favors the NVIDIA design: 122.9 million transistors per square millimeter versus 61.3 million for AMD. The 5090 D achieves this with a smaller 750 mm² die while packing 92,200 million transistors, compared to 73,000 million on the larger AMD die. The MI350P shows lower power draw per unit of memory bandwidth, but the 5090 D draws less total power at 575 W versus 600 W.

The release timeline also separates the two. The RTX 5090 D launched on 2025-01-29 and is marked as Active production. The MI350P carries a release date of 2026-05-06 with no production status listed. The 5090 D has a launch MSRP of 2,299 USD. The MI350P has no launch MSRP recorded.

The database records show the RTX 5090 D as the only one of the two with measured performance data. Its nearest rivals in the database are mobile and datacenter parts: the AMD Radeon RX 6650M XT (1.1% behind), the AMD Radeon RX 6850M XT (1.6% behind), and two NVIDIA Tesla P100 variants (2.1% and 2.4% behind). These narrow deltas indicate the 5090 D sits in a competitive performance band among recorded GPUs, but the absence of any MI350P benchmark data leaves the AMD part uncharacterized in the database.