AMD Radeon RX 6650M XT vs NVIDIA GeForce RTX 5090 D Comparison

The GeForce RTX 5090 D and Radeon RX 6650M XT occupy opposite ends of the desktop-to-mobile spectrum, but the benchmark data places them in the same performance conversation. Their average scores sit within 1.1% of each other, making this a matchup defined by a single shared metric and a massive architectural gulf.

Head-to-Head Benchmarks

The only direct head-to-head result available is the Geekbench OpenCL test, and the margin is decisive. The RTX 5090 D scores 310,674, while the RX 6650M XT scores 76,904. That is a 304% delta in favor of the NVIDIA card — a fourfold difference in compute throughput. No other shared benchmark exists in the dataset, so this single result carries the entire comparative weight.

Looking at overall averages, the RTX 5090 D posts an average benchmark score of 77,712 across ten tests, while the RX 6650M XT averages 76,904 from a single result. The delta is just 1.1% in favor of NVIDIA. However, this narrow gap is misleading. The RX 6650M XT’s average is derived entirely from that one OpenCL run, whereas the RTX 5090 D’s average pulls from a wider suite including 3DMark Steel Nomad DX12 (14,326), Passmark G3D (44,065), and Passmark GPU Compute (28,396). In pure compute terms, the RTX 5090 D also delivers 104.8 TFLOPS FP32 versus 9.896 TFLOPS for the AMD part — a 10.6x raw throughput advantage that the OpenCL score roughly reflects.

The percentile rankings are nearly identical: the RTX 5090 D sits at the 92nd percentile of all GPUs, the RX 6650M XT at the 91st. Both cards are outliers in their own ways — the NVIDIA part is a top-tier desktop flagship, the AMD part is a mobile chip that punches far above its class. The nearest rival data confirms this: the RX 6650M XT is 1% behind the RTX 5090 D, but also 2.6% behind the RX 6850M XT and 3.1% behind the Tesla P100 12 GB. The RTX 5090 D trails the RX 6850M XT by 1.6% and the Tesla P100 16 GB by 2.4%, despite its OpenCL dominance.

Architecture Differences

The RTX 5090 D is built on NVIDIA’s Blackwell 2.0 architecture, using the GB202 chip fabricated on TSMC’s 5 nm node. It packs 92,200 million transistors across a 750 mm² die, yielding a density of 122.9 million transistors per square millimeter. The RX 6650M XT uses AMD’s RDNA 2.0 architecture, featuring the Navi 23 chip on a 7 nm TSMC process. The AMD die holds 11,060 million transistors in 237 mm², for a density of 46.7M/mm². The NVIDIA chip is roughly 8.3x larger in transistor count and 3.2x larger in die area.

The compute resources diverge sharply. The RTX 5090 D has 21,760 shading units, 680 TMUs, and 176 ROPs. The RX 6650M XT has 2,048 shading units, 128 TMUs, and 64 ROPs. NVIDIA also fields 170 RT cores and 680 tensor cores, while AMD offers 32 RT cores and no tensor core equivalent. The FP32 throughput numbers reflect this: 104.8 TFLOPS for NVIDIA versus 9.896 TFLOPS for AMD. FP16 performance is more interesting — the NVIDIA card does 104.8 TFLOPS at a 1:1 ratio, while the AMD card achieves 19.79 TFLOPS at a 2:1 ratio, meaning it halves FP32 throughput when processing FP16.

Memory is another chasm. The RTX 5090 D comes with 32 GB of GDDR7 on a 512-bit bus, delivering 1.79 TB/s of bandwidth. The RX 6650M XT has 8 GB of GDDR6 on a 128-bit bus, with 256.0 GB/s. That is a 7x difference in bandwidth and a 4x difference in capacity. The NVIDIA card’s memory clock is 1750 MHz (28 Gbps effective), while AMD’s runs at 2000 MHz (16 Gbps effective). Pixel and texture rates follow suit: 423.6 GPixel/s and 1,636.8 GTexel/s for NVIDIA versus 154.6 GPixel/s and 309.2 GTexel/s for AMD.

Power and physical specs tell the form-factor story. The RTX 5090 D draws 575 W TDP, requires a 950 W suggested PSU, uses a dual-slot cooler with a single 16-pin connector, and measures 304 mm long. The RX 6650M XT is an IGP (integrated graphics package) with a 120 W TDP, no power connectors, and no listed dimensions — it is designed to be soldered into laptops. The NVIDIA card uses PCIe 5.0 x16; the AMD part uses PCIe 4.0 x8.

The Verdict

The data supports a clear split. If you need maximum compute performance in a desktop workstation, the RTX 5090 D is the only rational choice. Its OpenCL score is 304% higher, its FP32 throughput is 10.6x greater, and its memory bandwidth is 7x higher. No amount of architectural efficiency can offset those gaps.

If you are constrained to a mobile platform, the RX 6650M XT is the relevant part. It is end-of-life, but it delivers 91st-percentile performance within a 120 W envelope. The RTX 5090 D cannot exist in that form factor — it requires a 950 W PSU and a 304 mm dual-slot cooler. The RX 6650M XT is an IGP with no power connector.

The average score delta of 1.1% is essentially noise given the test disparity. The RTX 5090 D’s 77,712 average spans ten benchmarks, while the RX 6650M XT’s 76,904 comes from one. Treat the averages as indicative of similar overall standing, not equivalent capability. The RTX 5090 D’s percentile rank of 92 versus the RX 6650M XT’s 91 is also misleading — the NVIDIA card’s Passmark G2D score of 1,487 and Passmark DX9 score of 434 pull its average down, while the AMD card has no such low outliers.

Specification Differences

| Specification | NVIDIA GeForce RTX 5090 D | AMD Radeon RX 6650M XT |

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

| Architecture | Blackwell 2.0 | RDNA 2.0 |

| Process Node | 5 nm | 7 nm |

| Transistors | 92,200 million | 11,060 million |

| Die Size | 750 mm² | 237 mm² |

| Shading Units | 21,760 | 2,048 |

| TMUs | 680 | 128 |

| ROPs | 176 | 64 |

| RT Cores | 170 | 32 |

| Tensor Cores | 680 | None |

| FP32 | 104.8 TFLOPS | 9.896 TFLOPS |

| FP16 | 104.8 TFLOPS (1:1) | 19.79 TFLOPS (2:1) |

| Memory | 32 GB GDDR7 | 8 GB GDDR6 |

| Memory Bus | 512 bit | 128 bit |

| Bandwidth | 1.79 TB/s | 256.0 GB/s |

| TDP | 575 W | 120 W |

| Slot Width | Dual-slot | IGP |

| Power Connectors | 1x 16-pin | None |

| Bus Interface | PCIe 5.0 x16 | PCIe 4.0 x8 |

| Production Status | Active | End-of-life |

| Release Date | 2025-01-29 | 2022-01-03 |

| Launch MSRP | 2,299 USD | Not available |

FAQ

Q: Which card has a higher average benchmark score?

A: The RTX 5090 D averages 77,712 across ten tests. The RX 6650M XT averages 76,904 from a single test. The delta is 1.1% in favor of NVIDIA.

Q: How much faster is the RTX 5090 D in OpenCL compute?

A: The RTX 5090 D scores 310,674 in Geekbench OpenCL versus 76,904 for the RX 6650M XT — a 304% advantage.

Q: Can the RX 6650M XT be installed in a desktop?

A: No. It is an IGP (integrated graphics package) with no power connectors, no length/height/width dimensions, and a PCIe 4.0 x8 interface. It is designed for portable devices.

Q: What is the memory capacity difference?

A: The RTX 5090 D has 32 GB of GDDR7 on a 512-bit bus (1.79 TB/s). The RX 6650M XT has 8 GB of GDDR6 on a 128-bit bus (256.0 GB/s).

Q: Which card has a higher transistor density?

A: The RTX 5090 D has 122.9 million transistors per mm² on TSMC 5 nm. The RX 6650M XT has 46.7M/mm² on TSMC 7 nm.

Q: Are both cards still in production?

A: No. The RTX 5090 D is marked as Active. The RX 6650M XT is marked as End-of-life.

Where Each One Wins

The RTX 5090 D wins in every scenario where raw compute, memory bandwidth, or capacity matters. The OpenCL result is a 304% blowout. The FP32 throughput advantage (104.8 versus 9.896 TFLOPS) means any GPU-compute workload — rendering, simulation, machine learning inference — will favor NVIDIA overwhelmingly. The 32 GB GDDR7 frame buffer with 1.79 TB/s bandwidth handles datasets that would exhaust the RX 6650M XT’s 8 GB GDDR6 and 256.0 GB/s. The 170 RT cores versus 32 also suggest ray-tracing workloads will scale dramatically better on the desktop part, though no direct ray-tracing benchmark is in the data.

The RX 6650M XT wins on mobility and power efficiency. Its 120 W TDP versus 575 W means it operates in systems the RTX 5090 D physically cannot fit. The IGP form factor, no power connector requirement, and PCIe 4.0 x8 interface make it suitable for thin gaming laptops. Its 91st-percentile ranking shows it competes near the top of the mobile GPU hierarchy, and its FP16 performance at 19.79 TFLOPS (2:1 ratio) is respectable for its class. The RTX 5090 D’s 1,636.8 GTexel/s texture rate and 423.6 GPixel/s pixel rate are irrelevant if the card cannot be installed in your chassis.

For a desktop builder with a 950 W PSU and a full-size case, the RTX 5090 D is the unambiguous pick — the data shows no scenario where the RX 6650M XT outperforms it. For a laptop buyer, the RX 6650M XT is the only option that exists in that form factor, and its performance relative to the desktop flagship is within 1.1% on average score, despite being a 7 nm part from 2022. The choice is not about performance; it is about form factor and power delivery.