AMD Radeon Pro 555X vs NVIDIA P106-090 Comparison

NVIDIA P106-090 and AMD Radeon Pro 555X are both end-of-life GPUs that land in the same performance tier, with average benchmark scores of 13470 and 13321 respectively. The P106-090 edges ahead by roughly 1.1% on average, but the two cards are built for entirely different environments: one is a mining-only board with no display outputs, the other is a mobile integrated GPU for Mac laptops. The data shows that raw compute potential and real-world usability diverge sharply between these two.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA P106-090 scores 13470 on average, while the AMD Radeon Pro 555X scores 13321. This puts the P106-090 ahead by approximately 1.1%, though both cards sit at the 54th percentile among all GPUs.

Q: How do they compare in OpenCL performance?

A: The P106-090 delivers 21304 in Geekbench OpenCL, which is 68.7% higher than the Pro 555X's 12628. This is the largest performance gap between the two in any shared benchmark.

Q: What about Vulkan performance?

A: The P106-090 scores 18596 in Geekbench Vulkan, beating the Pro 555X's 12855 by 44.7%. The NVIDIA card wins this test decisively, though the margin is smaller than in OpenCL.

Q: Do both GPUs support the same API levels?

A: No. The P106-090 supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The Pro 555X supports DirectX 12 (12_0), OpenGL 4.6, and Vulkan 1.3. The NVIDIA card has a higher DirectX feature level and a newer Vulkan version.

Q: Which GPU has more memory bandwidth?

A: The P106-090 has 192.2 GB/s of bandwidth from its 192-bit bus, while the Pro 555X has 94.08 GB/s from a 128-bit bus. The NVIDIA card offers roughly double the memory bandwidth.

Q: Are both cards the same physical size?

A: No. The P106-090 is a dual-slot, 250 mm (9.8 inches) long card that requires a 6-pin power connector. The Pro 555X is an integrated GPU (IGP) with no dimensions listed and no power connectors.

Architecture Differences

The P106-090 is built on NVIDIA's Pascal architecture using the GP106 chip, fabricated on TSMC's 16 nm process. The Pro 555X uses AMD's GCN 4.0 architecture with the Polaris 21 chip, manufactured by GlobalFoundries on a 14 nm process. These are fundamentally different designs: Pascal is a streamlined gaming-oriented architecture, while GCN 4.0 is a compute-focused design with different scheduling and wavefront characteristics.

The transistor counts reflect this divergence. The P106-090 packs 4,400 million transistors on a 200 mm² die, giving a density of 22.0M per mm². The Pro 555X has 3,000 million transistors on a smaller 123 mm² die, achieving a higher density of 24.4M per mm². Despite having fewer transistors, the AMD chip crams them more tightly, which is typical of its smaller process node.

Both GPUs feature 768 shading units and 48 texture mapping units (TMUs), but their render output units (ROPs) differ dramatically: the P106-090 has 48 ROPs, while the Pro 555X has only 16. This 3x difference in ROP count directly impacts pixel throughput, with the P106-090 achieving 73.49 GPixel/s versus the Pro 555X's 14.51 GPixel/s.

Memory subsystems also differ. The P106-090 uses 3 GB of GDDR5 on a 192-bit bus, while the Pro 555X has 4 GB of GDDR5 on a 128-bit bus. The NVIDIA card's memory runs at 2002 MHz (8 Gbps effective), whereas the AMD card's memory runs at 1470 MHz (5.9 Gbps effective). These specs produce bandwidth figures of 192.2 GB/s and 94.08 GB/s respectively.

The power delivery and physical design are polar opposites. The P106-090 is a dual-slot card drawing 75 W TDP with a 250 W suggested PSU and a 6-pin connector, designed for desktop mining rigs. The Pro 555X is an IGP with 75 W TDP and no power connectors, integrated into laptops and relying on the host system's power delivery.

Head-to-Head Benchmarks

Only two benchmarks are shared between these GPUs, and the P106-090 wins both. In Geekbench OpenCL, the NVIDIA card scores 21304 against the AMD card's 12628, a delta of 68.7%. This is a crushing victory that highlights the P106-090's raw compute throughput, which is rated at 2.352 TFLOPS FP32 versus the Pro 555X's 1,393.2 GFLOPS.

The Geekbench Vulkan test shows a closer but still decisive margin. The P106-090 scores 18596, while the Pro 555X scores 12855, giving the NVIDIA card a 44.7% advantage. Vulkan results are less lopsided than OpenCL, but the P106-090 still outpaces the AMD chip by a wide margin.

Examining the nearest rivals provides context for these scores. The P106-090's 13470 average sits just 0.3% below the GeForce GTX 570 (13515) and 0.5% above the Radeon Pro 555 (13407). The Pro 555X's 13321 average is 0.2% above the GeForce GTX 480 (13300) and 0.3% below the Radeon RX 5500M (13356). Both cards are clustered tightly in the same performance band, yet the head-to-head results show the P106-090 is clearly the stronger compute performer.

The Pro 555X does have one unique benchmark not shared with the P106-090: Geekbench Metal, where it scores 14479. This is irrelevant for cross-comparison since the NVIDIA card lacks a Metal score, and the Pro 555X's OpenCL and Vulkan numbers are far lower than its Metal result, suggesting its compute performance varies significantly by API.

Specification Differences

| Specification | NVIDIA P106-090 | AMD Radeon Pro 555X |

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

| Process Node | 16 nm (TSMC) | 14 nm (GlobalFoundries) |

| Transistors | 4,400 million | 3,000 million |

| Die Size | 200 mm² | 123 mm² |

| Transistor Density | 22.0M / mm² | 24.4M / mm² |

| Memory Size | 3 GB GDDR5 | 4 GB GDDR5 |

| Memory Bus | 192 bit | 128 bit |

| Memory Bandwidth | 192.2 GB/s | 94.08 GB/s |

| Memory Clock | 2002 MHz (8 Gbps) | 1470 MHz (5.9 Gbps) |

| ROPs | 48 | 16 |

| Pixel Rate | 73.49 GPixel/s | 14.51 GPixel/s |

| Texture Rate | 73.49 GTexel/s | 43.54 GTexel/s |

| FP32 Performance | 2.352 TFLOPS | 1,393.2 GFLOPS |

| FP16 Performance | 36.74 GFLOPS (1:64) | 1,393.2 GFLOPS (1:1) |

| Slot Width | Dual-slot | IGP |

| Power Connectors | 1x 6-pin | None |

| Bus Interface | PCIe 1.0 x1 | PCIe 3.0 x8 |

| Display Outputs | No outputs | Portable Device Dependent |

| DirectX Support | 12 (12_1) | 12 (12_0) |

| Vulkan Support | 1.4 | 1.3 |

| Dimensions | 250 mm (9.8 inches) | Not listed |

The FP16 disparity is stark: the Pro 555X delivers 1,393.2 GFLOPS at 1:1 ratio, while the P106-090 manages only 36.74 GFLOPS at a 1:64 ratio. This makes the AMD card far more capable for workloads using half-precision arithmetic.

The Verdict

The data points to a clear performance hierarchy. The P106-090 wins both shared benchmarks by substantial margins: 68.7% in OpenCL and 44.7% in Vulkan. Its average score of 13470 also beats the Pro 555X's 13321, and its nearest rivals include the GTX 570 and Radeon Pro 555, both of which are desktop-class parts.

However, the P106-090 is a mining-specific card with no display outputs and a PCIe 1.0 x1 interface, which renders it useless for normal desktop use. The Pro 555X, despite losing all compute comparisons, is an integrated GPU designed for Mac portables with DisplayPort capability and PCIe 3.0 x8. This means the P106-090 only makes sense for computation in systems that don't need video output, while the Pro 555X is the only viable option for driving a screen.

The 16 ROP count on the Pro 555X versus 48 on the P106-090 is the most telling architectural gap, as it explains the 5x difference in pixel rate. The AMD card trades raw throughput for power efficiency and portability, while the NVIDIA card maximizes compute regardless of physical footprint or interface limitations.

Where Each One Wins

NVIDIA P106-090 wins in: OpenCL compute (21304 vs 12628, +68.7%), Vulkan compute (18596 vs 12855, +44.7%), memory bandwidth (192.2 GB/s vs 94.08 GB/s), FP32 throughput (2.352 TFLOPS vs 1,393.2 GFLOPS), pixel rate (73.49 GPixel/s vs 14.51 GPixel/s), and average benchmark score (13470 vs 13321). It also supports newer API versions: DirectX 12_1 versus 12_0 and Vulkan 1.4 versus 1.3.

AMD Radeon Pro 555X wins in: FP16 performance (1,393.2 GFLOPS at 1:1 ratio versus 36.74 GFLOPS at 1:64), memory capacity (4 GB versus 3 GB), transistor density (24.4M / mm² versus 22.0M / mm²), and practical usability with display outputs and a PCIe 3.0 x8 interface. It also has a unique Metal benchmark score of 14479, which the P106-090 cannot contest. The Pro 555X's portable form factor as an IGP with no power connectors makes it suitable for laptops, while the P106-090 requires a dual-slot desktop slot and 6-pin power.

For raw compute numbers, the P106-090 is the clear winner. For any system that needs to output video, the Pro 555X is the only option between these two.