AMD Radeon Pro 555 vs NVIDIA P106-090 Comparison

FAQ

Q: How do the NVIDIA P106-090 and AMD Radeon Pro 555 compare in average benchmark score?

A: The NVIDIA P106-090 posts an average benchmark score of 13470, while the AMD Radeon Pro 555 scores 13407. The P106-090 leads by 0.5% according to the deltaPct in the nearestRivals data, which places both cards within a negligible performance margin.

Q: Which card wins in Geekbench OpenCL, and by how much?

A: The NVIDIA P106-090 wins Geekbench OpenCL with a score of 21304 versus 11682 for the AMD Radeon Pro 555. This is an 82.4% advantage for the NVIDIA card, making it the single largest performance gap in the head-to-head results.

Q: Is there any benchmark where the AMD Radeon Pro 555 comes out ahead?

A: No. The head-to-head benchmark data shows the NVIDIA P106-090 winning both tests: Geekbench OpenCL and Geekbench Vulkan. The AMD card's only recorded benchmark win is in Geekbench Metal with a score of 16236, but no head-to-head Metal comparison is available against the P106-090.

Q: What is the memory configuration difference between these two cards?

A: The NVIDIA P106-090 has 3 GB of GDDR5 memory on a 192-bit bus, delivering 192.2 GB/s bandwidth. The AMD Radeon Pro 555 has 2 GB of GDDR5 on a 128-bit bus, yielding 81.60 GB/s bandwidth. The NVIDIA card offers 50% more memory capacity and more than double the memory bandwidth.

Q: How do the two cards compare in terms of physical design?

A: The NVIDIA P106-090 is a dual-slot card measuring 250 mm (9.8 inches) in length, requires a 1x 6-pin power connector, and has no display outputs. The AMD Radeon Pro 555 is an integrated GPU (IGP) with no power connectors and portable-device-dependent display outputs.

Q: Which card has the better Vulkan API support?

A: The NVIDIA P106-090 supports Vulkan 1.4, while the AMD Radeon Pro 555 supports Vulkan 1.3. In the Geekbench Vulkan test, the P106-090 scores 18596 versus 12303 for the Radeon Pro 555, a 51.2% advantage.

Architecture Differences

The NVIDIA P106-090 is built on the GP106 chip using the Pascal architecture, manufactured on a 16 nm process at TSMC. It integrates 4,400 million transistors on a 200 mm² die, yielding a transistor density of 22.0M per mm². The AMD Radeon Pro 555 uses the Polaris 21 chip with GCN 4.0 architecture, produced on a 14 nm process at GlobalFoundries. It packs 3,000 million transistors onto a 123 mm² die, achieving a higher transistor density of 24.4M per mm² despite the older process node.

Both cards feature 768 shading units and 48 texture mapping units, but the similarity ends there. The NVIDIA P106-090 includes 48 ROPs, whereas the AMD Radeon Pro 555 has only 16 ROPs. This difference directly impacts pixel throughput: the P106-090 reaches 73.49 GPixel/s, while the Radeon Pro 555 manages just 13.60 GPixel/s. Texture fill rates also diverge, with the NVIDIA card hitting 73.49 GTexel/s versus 40.80 GTexel/s for the AMD card.

Compute performance tells a similar story. The NVIDIA P106-090 delivers 2.352 TFLOPS of FP32 performance, nearly double the 1,305.6 GFLOPS from the Radeon Pro 555. However, the AMD card offers FP16 performance at a 1:1 ratio with FP32 (1,305.6 GFLOPS), while the NVIDIA card cripples FP16 to just 36.74 GFLOPS (1:64 ratio). This makes the Radeon Pro 555 far more capable in workloads that utilize half-precision arithmetic.

Memory architecture differs substantially. The P106-090 runs 3 GB of GDDR5 across a 192-bit bus with memory clocks of 2002 MHz (8 Gbps effective), producing 192.2 GB/s of bandwidth. The Radeon Pro 555 uses 2 GB of GDDR5 on a 128-bit bus at 1275 MHz (5.1 Gbps effective), delivering only 81.60 GB/s. The NVIDIA card's bandwidth advantage is roughly 2.35x, which heavily influences compute-heavy workloads.

The bus interfaces are fundamentally different. The P106-090 uses PCIe 1.0 x1, which severely limits host communication, while the Radeon Pro 555 uses PCIe 3.0 x8. The AMD card is also an integrated GPU with no power connectors, whereas the NVIDIA card is a dual-slot discrete solution requiring a 6-pin connector. Both cards are end-of-life products, with the P106-090 released in July 2017 and the Radeon Pro 555 in June 2017.

Head-to-Head Benchmarks

The head-to-head results show a clean sweep for the NVIDIA P106-090, but the magnitude of each win tells a nuanced story. In Geekbench OpenCL, the P106-090 scores 21304 against 11682 for the Radeon Pro 555. That is an 82.4% advantage, a massive gap that reflects the NVIDIA card's superior memory bandwidth (192.2 GB/s versus 81.60 GB/s) and nearly double the FP32 throughput (2.352 TFLOPS versus 1,305.6 GFLOPS).

Geekbench Vulkan shows a smaller but still decisive win. The P106-090 scores 18596, while the Radeon Pro 555 scores 12303, a 51.2% difference. This margin aligns with the NVIDIA card's higher texture rate (73.49 GTexel/s versus 40.80 GTexel/s) and better ROP count (48 versus 16). The Vulkan API version difference (1.4 versus 1.3) may also contribute to the NVIDIA card's stronger showing in this test.

The average benchmark scores across all tests paint a different picture. The P106-090 averages 13470, and the Radeon Pro 555 averages 13407. The deltaPct between them is just 0.5%, placing both cards in the 54th percentile among all GPUs. This suggests that while the NVIDIA card dominates in the two head-to-head tests, the AMD card's Geekbench Metal score of 16236 helps it stay competitive in the overall average.

Looking at the nearest rivals for each card provides context. The P106-090 sits between the NVIDIA GeForce GTX 570 (avg score 13515, deltaPct -0.3) and the AMD Radeon HD 7770M (avg score 13536, deltaPct -0.5). The AMD Radeon Pro 555 is bracketed by the AMD Radeon HD 8950M (avg score 13376, deltaPct 0.2) and the AMD Radeon RX 5500M (avg score 13356, deltaPct 0.4). Both cards occupy the same performance tier, despite their architectural differences.

The wins tally is straightforward: the NVIDIA P106-090 takes 2 wins, and the AMD Radeon Pro 555 takes 0 wins in the head-to-head comparison. However, the average score data shows that these two cards are effectively peers in overall performance, separated by less than one percent.

The Verdict

The data indicates a clear but narrow overall picture. The NVIDIA P106-090 wins both head-to-head benchmarks decisively, with an 82.4% lead in OpenCL and a 51.2% lead in Vulkan. If your workload relies on OpenCL or Vulkan compute, the P106-090 is the stronger choice by a significant margin.

The AMD Radeon Pro 555 is not without merit. Its Geekbench Metal score of 16236 demonstrates strong performance in Apple's graphics API, which is absent from the NVIDIA card's benchmark results. The AMD card also offers FP16 performance at 1,305.6 GFLOPS in a 1:1 ratio, making it suitable for half-precision workloads where the P106-090's 36.74 GFLOPS FP16 is a severe limitation.

The average benchmark scores show both cards within 0.5% of each other, placing them in the same performance tier. The P106-090's average of 13470 versus the Radeon Pro 555's 13407 means that in a broad mix of workloads, neither card has a meaningful overall advantage.

The physical form factor is a deciding factor. The Radeon Pro 555 is an integrated GPU with no power connectors and portable-device-dependent outputs, making it suitable for laptop or compact systems. The P106-090 is a dual-slot 250 mm card with a 6-pin connector and no display outputs, designed for mining systems with PCIe 1.0 x1 connectivity.

For a builder selecting between these two, the choice depends on the target platform. The Radeon Pro 555 fits into existing portable systems without modification, while the P106-090 requires a dedicated slot and power connection. If compute performance in OpenCL or Vulkan is the priority, the NVIDIA card's 82.4% and 51.2% wins make it the obvious pick.

Specification Differences

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

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

| Manufacturer | NVIDIA | AMD |

| Chip | GP106 | Polaris 21 |

| Architecture | Pascal | GCN 4.0 |

| Generation | Mining GPUs | Radeon Pro Mac (500 Series) |

| Process Node | 16 nm | 14 nm |

| Foundry | TSMC | GlobalFoundries |

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

| Die Size | 200 mm² | 123 mm² |

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

| Base Clock | 1354 MHz | Not specified |

| Boost Clock | 1531 MHz | Not specified |

| Memory Clock | 2002 MHz (8 Gbps effective) | 1275 MHz (5.1 Gbps effective) |

| Memory Size | 3 GB | 2 GB |

| Memory Type | GDDR5 | GDDR5 |

| Memory Bus Width | 192 bit | 128 bit |

| Memory Bandwidth | 192.2 GB/s | 81.60 GB/s |

| Shading Units | 768 | 768 |

| TMUs | 48 | 48 |

| ROPs | 48 | 16 |

| Pixel Rate | 73.49 GPixel/s | 13.60 GPixel/s |

| Texture Rate | 73.49 GTexel/s | 40.80 GTexel/s |

| FP32 Performance | 2.352 TFLOPS | 1,305.6 GFLOPS |

| FP16 Performance | 36.74 GFLOPS (1:64) | 1,305.6 GFLOPS (1:1) |

| TDP | 75 W | 75 W |

| Slot Width | Dual-slot | IGP |

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

| Suggested PSU | 250 W | Not specified |

| 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) |

| OpenGL Support | 4.6 | 4.6 |

| Vulkan Support | 1.4 | 1.3 |

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

| Release Date | 2017-07-30 | 2017-06-04 |

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

Where Each One Wins

The NVIDIA P106-090 wins in raw compute performance. Its 2.352 TFLOPS FP32 throughput and 192.2 GB/s memory bandwidth give it an 82.4% advantage in Geekbench OpenCL and a 51.2% edge in Geekbench Vulkan. The 48 ROPs and 73.49 GPixel/s pixel rate make it substantially faster in rasterization-heavy tasks, despite having no display outputs. The 3 GB memory capacity is also 50% larger than the AMD card's 2 GB, which helps in workloads that exceed 2 GB of VRAM.

The AMD Radeon Pro 555 wins in half-precision compute. Its FP16 performance of 1,305.6 GFLOPS matches its FP32 throughput, whereas the NVIDIA card's FP16 is throttled to 36.74 GFLOPS. For machine learning inference or other FP16-heavy workloads, the AMD card is over 35x faster in this specific metric. The Radeon Pro 555 also wins in system integration, being an IGP with no power connectors that works in portable devices.

The average benchmark scores suggest a draw overall. The P106-090's 13470 average and the Radeon Pro 555's 13407 average are within 0.5% of each other, and both sit in the 54th percentile of all GPUs. The Geekbench Metal score of 16236 for the AMD card provides a data point that the NVIDIA card cannot match, as no Metal benchmark exists for the P106-090.

For a mining-focused build with PCIe 1.0 x1 slots and dedicated power, the P106-090 is the clear winner based on its compute dominance. For a portable system requiring integrated graphics with no additional power draw and support for half-precision operations, the Radeon Pro 555 is the only viable option between the two. The end-of-life status of both cards means availability and driver support should be verified before purchase, but the benchmark data speaks clearly: the NVIDIA P106-090 is the faster compute card, while the AMD Radeon Pro 555 offers a different feature set that may suit specific use cases.