AMD Radeon Pro 5500M vs NVIDIA P106-090 Comparison

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA P106-090 holds a higher average benchmark score of 13,470 compared to the AMD Radeon Pro 5500M’s 11,528. The NVIDIA part also sits at the 54th percentile among all GPUs, while the AMD part sits at the 51st percentile.

Q: How do the two GPUs compare in Geekbench OpenCL performance?

A: The AMD Radeon Pro 5500M scores 31,088 in Geekbench OpenCL, which is 31.5% higher than the NVIDIA P106-090’s 21,304. This is a significant lead for the AMD part in compute workloads using OpenCL.

Q: What about Vulkan performance?

A: The AMD Radeon Pro 5500M wins Geekbench Vulkan with a score of 35,134 versus 18,596 for the NVIDIA P106-090, a delta of 47.1% in AMD’s favor. This is the largest head-to-head margin recorded between the two.

Q: Which GPU has more memory and what type?

A: The AMD Radeon Pro 5500M has 8 GB of GDDR6 memory on a 128-bit bus, while the NVIDIA P106-090 has 3 GB of GDDR5 memory on a 192-bit bus. Despite the narrower bus, the AMD card’s bandwidth is 192.0 GB/s, nearly identical to the NVIDIA card’s 192.2 GB/s.

Q: Do both GPUs support the same DirectX and Vulkan versions?

A: Yes, both support DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The core API feature sets are identical, so differences in performance come from architecture and hardware resources rather than API support.

Q: Which GPU is newer and on what process node?

A: The AMD Radeon Pro 5500M was released on 2019-11-12 and uses TSMC’s 7 nm process, while the NVIDIA P106-090 was released on 2017-07-30 and uses TSMC’s 16 nm process. The AMD part also packs more transistors at 6,400 million versus 4,400 million.

Architecture Differences

The two GPUs come from fundamentally different architectural generations. The NVIDIA P106-090 is built on the Pascal architecture with the GP106 chip, while the AMD Radeon Pro 5500M uses the RDNA 1.0 architecture with the Navi 14 chip. This is not a minor refresh; it is a complete design divergence in how each GPU schedules work and processes graphics.

The manufacturing process separates them further. NVIDIA uses TSMC’s 16 nm node, while AMD uses TSMC’s 7 nm node. The die sizes tell a revealing story: the NVIDIA chip is larger at 200 mm² but holds fewer transistors at 4,400 million, giving a transistor density of 22.0M per mm². The AMD chip is smaller at 158 mm² but packs 6,400 million transistors, resulting in a density of 40.5M per mm². In short, the AMD chip is nearly twice as dense and more modern.

The compute resources also differ sharply. The NVIDIA P106-090 has 768 shading units, 48 TMUs, and 48 ROPs. The AMD Radeon Pro 5500M has 1,536 shading units, 96 TMUs, and 32 ROPs. AMD packs double the shading units and TMUs, but fewer ROPs. This explains the divergent pixel and texture rates: NVIDIA achieves 73.49 GPixel/s and 73.49 GTexel/s, while AMD achieves 46.40 GPixel/s and 139.2 GTexel/s. AMD’s texture throughput is nearly double, but its pixel fillrate is lower.

Clock behavior differs too. The NVIDIA card runs a base clock of 1354 MHz and a boost of 1531 MHz. The AMD card runs a lower base of 1000 MHz but boosts to 1450 MHz. Memory speeds also differ: NVIDIA uses 2002 MHz memory with 8 Gbps effective, while AMD uses 1500 MHz with 12 Gbps effective. Both GPUs have no ray tracing cores and no tensor cores, so neither is designed for those workloads.

The FP32 and FP16 throughput shows a major architectural gap. The NVIDIA P106-090 delivers 2.352 TFLOPS FP32 and only 36.74 GFLOPS FP16, a 1:64 ratio, which is extremely weak for half-precision work. The AMD Radeon Pro 5500M delivers 4.454 TFLOPS FP32 and 8.909 TFLOPS FP16, a 2:1 ratio, making it far more capable in mixed-precision tasks.

The power envelope differs as well. NVIDIA’s TDP is 75 W with a 1x 6-pin power connector and a suggested PSU of 250 W. AMD’s TDP is 85 W, but it requires no power connectors and has no suggested PSU listed. The NVIDIA card is dual-slot and 250 mm long, while the AMD card has no listed dimensions and is described as portable device dependent.

The Verdict

The data points to a clear split: the AMD Radeon Pro 5500M is the stronger performer in modern compute APIs, while the NVIDIA P106-090 holds a marginal edge in average aggregate score. The AMD card wins both head-to-head benchmarks recorded, with a 31.5% lead in Geekbench OpenCL and a 47.1% lead in Geekbench Vulkan. That is not a close contest in those workloads.

However, the NVIDIA card has a higher average benchmark score of 13,470 versus 11,528, a difference of roughly 17%. This is because the benchmark sets differ: the NVIDIA card’s scores come from 3DMark Steel Nomad DX12 and Geekbench OpenCL and Vulkan, while the AMD card’s scores include a broader set including PassMark tests and Geekbench Metal. The AMD card’s PassMark DirectX 10, 11, and 12 scores are very low (36, 42, and 29 respectively), which drags down its average.

For users who prioritize OpenCL and Vulkan compute, the AMD Radeon Pro 5500M is the obvious choice. Its FP32 throughput is nearly double, its FP16 throughput is over 240 times higher, and its texture rate is nearly double. For users who care about the aggregate benchmark average or who target DirectX 12 scenarios like 3DMark Steel Nomad, the NVIDIA P106-090 is the better fit.

The AMD card is also the more modern part, with a 7 nm process, 8 GB of memory, and GDDR6, versus the NVIDIA card’s 16 nm process, 3 GB of memory, and GDDR5. The NVIDIA card has no display outputs, while the AMD card is portable device dependent, meaning it is designed for laptops. Neither GPU is current production; both are end-of-life.

Specification Differences

The two GPUs differ across nearly every major specification category. The NVIDIA P106-090 uses a 16 nm process, while the AMD Radeon Pro 5500M uses a 7 nm process. Transistor counts are 4,400 million versus 6,400 million, and die sizes are 200 mm² versus 158 mm². Transistor density is 22.0M per mm² versus 40.5M per mm².

Memory configurations are very different. The NVIDIA card has 3 GB of GDDR5 on a 192-bit bus with 192.2 GB/s bandwidth. The AMD card has 8 GB of GDDR6 on a 128-bit bus with 192.0 GB/s bandwidth. The memory type and capacity favor AMD, while the bus width favors NVIDIA, but bandwidth is nearly identical.

Compute resources diverge strongly. NVIDIA has 768 shading units, 48 TMUs, and 48 ROPs. AMD has 1,536 shading units, 96 TMUs, and 32 ROPs. Pixel rates are 73.49 GPixel/s versus 46.40 GPixel/s, while texture rates are 73.49 GTexel/s versus 139.2 GTexel/s. FP32 is 2.352 TFLOPS versus 4.454 TFLOPS, and FP16 is 36.74 GFLOPS versus 8.909 TFLOPS.

Clocks differ: NVIDIA runs at 1354 MHz base and 1531 MHz boost, AMD runs at 1000 MHz base and 1450 MHz boost. Memory clocks are 2002 MHz (8 Gbps effective) versus 1500 MHz (12 Gbps effective). The NVIDIA card has a TDP of 75 W, while AMD has 85 W. NVIDIA uses a 1x 6-pin power connector, while AMD uses none. NVIDIA’s bus interface is PCIe 1.0 x1, which is extremely limited, while AMD uses PCIe 4.0 x8.

The NVIDIA card is dual-slot and 250 mm long, with no display outputs. The AMD card has no listed dimensions and its display outputs are described as portable device dependent. Both support DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. Neither has ray tracing cores or tensor cores.

Head-to-Head Benchmarks

The database records two direct head-to-head comparisons between these GPUs. The AMD Radeon Pro 5500M wins both.

In Geekbench OpenCL, the AMD card scores 31,088 against the NVIDIA card’s 21,304. The delta is 31.5% in AMD’s favor. This is a substantial margin and indicates that the AMD card’s larger shader count and higher FP32 throughput translate directly into OpenCL compute performance.

In Geekbench Vulkan, the gap widens. The AMD card scores 35,134 versus 18,596 for the NVIDIA card, a delta of 47.1%. This is the largest recorded win in the head-to-head set. The AMD card’s Vulkan score is nearly double the NVIDIA card’s, which suggests that the RDNA architecture handles Vulkan workloads far more efficiently than Pascal.

The NVIDIA card does not win any recorded head-to-head benchmark. Its only competitive advantage appears in the aggregate average score, where it reaches 13,470 versus AMD’s 11,528. That aggregate advantage is driven by the 3DMark Steel Nomad DX12 score of 509, which is not part of the AMD card’s recorded benchmark set, and by the AMD card’s very low PassMark DirectX scores (36, 42, 29, and 96 across DX10, DX11, DX12, and DX9).

Where Each One Wins

The AMD Radeon Pro 5500M is the clear winner in compute-heavy and modern API workloads. It leads by 31.5% in OpenCL and 47.1% in Vulkan, making it the better choice for applications that leverage those APIs. Its FP32 throughput of 4.454 TFLOPS is nearly double the NVIDIA card’s 2.352 TFLOPS, and its FP16 throughput of 8.909 TFLOPS is over 240 times higher. Its texture rate of 139.2 GTexel/s is also nearly double. For machine learning inference, GPU compute tasks, or Vulkan-based games, the AMD card is the superior option.

The AMD card also wins on memory capacity and modern features. It has 8 GB of GDDR6 versus 3 GB of GDDR5, a 7 nm process versus 16 nm, and PCIe 4.0 x8 versus PCIe 1.0 x1. The PCIe 1.0 x1 interface on the NVIDIA card is a severe bottleneck for data transfer, which is notable given that the card has no display outputs and appears designed for mining. The AMD card’s PCIe 4.0 x8 interface is far more capable.

The NVIDIA P106-090 wins where the aggregate benchmark average matters. Its average score of 13,470 is higher than the AMD card’s 11,528, and its 54th percentile places it above the AMD card’s 51st percentile. It also has a higher pixel rate of 73.49 GPixel/s versus 46.40 GPixel/s, which benefits fillrate-bound workloads. Its nearest rivals include the NVIDIA GeForce GTX 570 at 13,515 and the AMD Radeon RX 9070 XT at 13,543, while the AMD card’s nearest rivals include the NVIDIA GeForce GTX 1660 at 11,680 and the AMD Radeon RX 7800 XT at 11,627.

For DirectX 12 workloads, the NVIDIA card’s 3DMark Steel Nomad score of 509 is the only recorded DX12 data point, and it is not directly comparable to the AMD card’s PassMark DX12 score of 29 due to different test methodologies. The AMD card’s low PassMark DX10, DX11, and DX12 scores suggest it is not optimized for legacy DirectX paths, while the NVIDIA card’s single modern DX12 score indicates better modern DirectX performance.

In summary, pick the AMD Radeon Pro 5500M for OpenCL, Vulkan, FP16 compute, texture-heavy workloads, and modern memory capacity. Pick the NVIDIA P106-090 for higher aggregate benchmark scores, higher pixel fillrate, and DirectX 12 scenarios. The AMD card is the more balanced and modern part; the NVIDIA card is a specialized mining GPU with no display outputs and a constrained PCIe interface.