AMD Radeon Pro W5500 vs NVIDIA P106-090 Comparison

FAQ

Q: What is the core performance difference between the AMD Radeon Pro W5500 and the NVIDIA P106-090?

A: The Radeon Pro W5500 delivers 5.224 TFLOPS of FP32 compute throughput, while the P106-090 manages 2.352 TFLOPS. In Geekbench OpenCL, the AMD card scores 45615 versus 21304 for the NVIDIA card, a 114.1% advantage.

Q: Which card has more memory and bandwidth?

A: The Radeon Pro W5500 comes with 8 GB of GDDR6 on a 128-bit bus, providing 224.0 GB/s. The P106-090 has 3 GB of GDDR5 on a 192-bit bus, giving 192.2 GB/s. The AMD card has over double the memory capacity and roughly 16% more bandwidth.

Q: Can the NVIDIA P106-090 output video to a display?

A: No. The P106-090 has no display outputs. It was designed for mining workloads. The Radeon Pro W5500, by contrast, has 4x DisplayPort 1.4a outputs.

Q: How do their power requirements compare?

A: The Radeon Pro W5500 has a TDP of 125 W with a suggested PSU of 300 W. The P106-090 has a TDP of 75 W with a suggested PSU of 250 W. Both use a single 6-pin power connector.

Q: Which card has a higher pixel fillrate?

A: The P106-090 has the higher pixel rate at 73.49 GPixel/s, despite its lower overall compute performance. The Radeon Pro W5500 is at 59.36 GPixel/s.

Q: How do the cards compare in overall benchmark standing?

A: The Radeon Pro W5500 sits at the 58th percentile of all GPUs with an average benchmark score of 15679. The P106-090 sits at the 54th percentile with an average score of 13470. The AMD card's nearest rival by average score is the GTX 1080 Ti at 15548, only 0.8% different.

The Verdict

The data makes a clear distinction for most buyers. The AMD Radeon Pro W5500 is a workstation-class professional card: it has display outputs, 8 GB of GDDR6 memory, PCIe 4.0 interface support, and twice the FP32 compute throughput. Its average benchmark score of 15679 places it 16% above the P106-090's 13470, and it outperforms the NVIDIA card by triple-digit margins in both shared benchmark tests.

The NVIDIA P106-090 is a mining-oriented product with no display outputs, a 1.0 x1 PCIe interface limited to PCIe 1.0, and a 16 nm process node. It has a lower 75 W TDP and smaller 3 GB memory pool. Its only wins are in pixel rate and its lower power figure, plus its 192-bit memory bus width advantage. None of those make up for a 2.352 TFLOPS ceiling that is less than half the AMD card's 5.224 TFLOPS.

For a user building a system that needs a working display or professional workload acceleration, the Radeon Pro W5500 is the practical choice. The P106-090 is a niche part for compute-only scenarios where the 75 W TDP matters more than memory capacity, CUDA-style compute ecosystem compatibility, or any display capability. The P106-090 was released in 2017 and is end-of-life, and the Radeon Pro W5500 was released in 2020 and is also end-of-life. Neither has a successor listed in the database.

Head-to-Head Benchmarks

The shared benchmark overlap between these two cards is limited to two Geekbench tests. In Geekbench OpenCL, the Radeon Pro W5500 scores 45615 against the P106-090's 21304, a 114.1% advantage. The AMD card nearly doubles the NVIDIA card's OpenCL compute result. In Geekbench Vulkan, the Radeon Pro W5500 scores 126% higher at 42021 versus 18596. These are the only tests where both cards were measured, and the AMD card wins both.

The Radeon Pro W5500 also has a 58th percentile overall standing compared to the P106-090's 54th percentile. Its average benchmark score of 15679 is 16.4% ahead of the NVIDIA card's 13470 average. That puts the AMD card in the same competitive vicinity as the GTX 1080 Ti, whose average score of 15548 is only 0.8% away, and the RX 7700, which sits 1.1% higher. The P106-090's nearest rivals are the GTX 570 at 13515 and the Radeon Pro 555 at 13407, both within 0.5% of its score.

The AMD card's best individual benchmark results are Geekbench Metal at 54302, Geekbench OpenCL at 45615, and Geekbench Vulkan at 42021. It also has Passmark G3D score of 8978 and Passmark GPU Compute of 4804. The NVIDIA card's best recorded results are Geekbench OpenCL at 21304 and Geekbench Vulkan at 18596, with a 3DMark Steel Nomad DX12 score of 509. The Radeon Pro W5500 wins every benchmark where both were tested, and its compute-oriented scores are more than double the NVIDIA card's.

Specification Differences

The Radeon Pro W5500 uses the Navi 14 chip from the Radeon architecture family, built on a 7 nm TSMC process with 6,400 million transistors on a 158 mm² die, giving a density of 40.5 million transistors per square millimeter. The P106-090 uses the GP106 chip from the Pascal architecture on a 16 nm TSMC process with 4,400 million transistors on a 200 mm² die, at a density of 22.0 million per square millimeter.

Clock speeds differ substantially. The Radeon Pro W5500 runs at 1744 MHz base and 1855 MHz boost, while the P106-090 runs at 1354 MHz base and 1531 MHz boost. Memory clocks are 1750 MHz for the AMD card, translating to 14 Gbps effective, versus 2002 MHz for the NVIDIA card at 8 Gbps effective.

Memory configuration also differs: the AMD card has 8 GB of GDDR6 on a 128-bit interface at 224.0 GB/s. The NVIDIA card has 3 GB of core memory size of GDDR5 on a 192-bit interface at 192.2 GB/s. The AMD card has 1408 shading units, 88 TMUs and 32 ROPs, while the P106-090 has 768 shading units, 48 TMUs and 48 ROPs.

The AMD card has a texture rate of 163.2 GTexel/s, while the P106-090's texture rate is 73.49 GTexel/s. The AMD card's pixel rate is 59.36 GPixel/s, while the NVIDIA card's is 73.49 GPixel/s. The P106-090 has a higher pixel rate despite being slower overall, likely because of its 48 ROPs versus the AMD card's 32.

Power and physical dimensions differ: the Radeon Pro W5500 is rated at 125 W TDP and is a single-slot card, 241 mm long and 111 mm tall. The P106-090 is rated at 75 W and is a dual-slot card, measuring 250 mm long. The AMD card uses a PCIe 4.0 x8 bus interface and has 4x DisplayPort 1.4a outputs. The P106-090 uses a PCIe 1.0 x1 bus interface and has no display outputs.

The AMD card's APIs include DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The NVIDIA card has identical API support: DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The AMD card has a launch MSRP of 399 USD. The P106-090 has no recorded launch MSRP in the database.

The Radeon Pro W5500 FP16 performance is 10.45 TFLOPS at a 2:1 ratio, while P106-090 FP16 is 36.74 GFLOPS at a 1:64 ratio. The AMD card has a far higher FP16 throughput, over 284 times the NVIDIA card's FP16 output.

Architecture Differences

The RDNA 1.0 architecture on Navi 14 uses 7 nm manufacturing at TSMC, enabling the 6,400 million transistors in the Radeon Pro Navi generation line on the smaller 158 mm² die. The Pascal architecture on GP106 uses the older 16 nm process node at TSMC with 4,400 million transistors on a larger 200 mm² die. The AMD Radeon Pro W5500 has a transistor density of 40.5 million transistors per square millimeter, versus the P106-090's 22.0 million per square millimeter, reflecting the newer process node's density advantage.

The Radeon Pro W5500 has FP32 compute of 5.224 TFLOPS, FP16 compute of 10.45 TFLOPS, texture rate of 163.2 GTexel/s, and 59.36 GPixel/s. The P106-090 has FP32 compute of 2.352 TFLOPS, FP16 compute of 36.74 GFLOPS, texture rate of 73.49 GTexel/s, and 73.49 GPixel/s. The AMD card has a 2.2x FP32 advantage, a 284x FP16 advantage, a 222% texture rate advantage, and a lower pixel rate.

Both cards lack ray tracing cores and tensor cores in the database. Both share the same DirectX 12 (12_1 feature), OpenGL 4.6, Vulkan 1.4, and both use a single 6-pin power connector, both are end-of-life, both use TSMC as foundry. The AMD card has a 7 nm process node, 128-bit memory bus, 32 ROPs, PCIe 4.0 x8 interface, 4 display outputs, 8 GB GDDR6, 224.0 GB/s bandwidth, 241 mm length, 111 mm height, 125 W TDP, 300 W suggested PSU. The P106-090 has a 16 nm process node, 192-bit memory bus, 48 ROPs, PCIe 1.0 x1 interface, no display outputs, 3 GB GDDR5, 192.2 GB/s bandwidth, 250 mm length, 75 W TDP, 250 W suggested PSU.

The AMD card has 1408 shading units, 88 TMUs, 32 ROPs, while the NVIDIA card has 768 shading units, 48 TMUs, 48 ROPs. The AMD card has 8 GB memory, the NVIDIA card has 4 GB less memory. The AMD card has a 128-bit memory bus width, the NVIDIA card has a 192-bit memory bus, 64 bit wider than the AMD card's 128-bit bus, which is a 50% wider memory bus width advantage for the P106-090, despite its lower overall bandwidth.

The Radeon product line lists a predecessor, the Radeon Pro Vega, while the P106-090 has no predecessor recorded in the database. The AMD card was released in 2020 and the NVIDIA card was released in 2017, with the AMD card being the newer release. The AMD card is single-slot wide, the NVIDIA card is dual-slot wide.

Where Each One Wins

The AMD Radeon Pro W5500 wins in compute-heavy workloads across every overlapping benchmark. Its 114.1% lead in Geekbench OpenCL shows a card that excels in OpenCL compute workloads, and its 126% Vulkan advantage shows the RDNA 1.0 architecture scales FP32 performance at 5.2 TFLOPS. Its 8 GB GDDR6 memory capacity at 224.0 GB/s bandwidth makes it the clear pick for memory-heavy workloads like large dataset operations, when paired with 1408 shading units, 88 TMUs, 4 DisplayPort 1.4a outputs, PCIe 4.0 x8 interface, and a 58th percentile standing. Its texture rate of 163.2 GTexel/s is well above the P106-090's 73.49 GTexel/s, giving a 2.2x advantage. The AMD card also supports Vulkan 1.4 and DirectX 12 (12_1) and OpenGL 4.6 APIs, matching the NVIDIA card's API support. Its 399 USD launch MSRP is recorded in the database.

The NVIDIA P106-090 wins in pixel rate at 73.49 GPixel/s over the AMD card's 59.36 GPixel/s, a notable win for fillrate-oriented scenarios, though pixel rate is higher than texture rate for the NVIDIA card. Its 75 W TDP and 250 W PSU requirement make it a lower-power option for compute-only nodes, and its 192-bit memory bus width of 192-bit at 192.2 GB/s is wider than the AMD card's 128-bit, though the AMD card compensates with more memory capacity. The P106-090's 48 ROPs count matches its pixel rate advantage over the AMD card's 32 ROPs, and its 3 GB GDDR5 memory uses 3 GB GDDR5, which is lower capacity and lower bandwidth than the AMD card. Its 16 nm process node and 4,400 million transistors give a smaller transistor count than the AMD card's 6,400 million, and its 200 mm² die size exceeds the AMD card's 158 mm². The P106-090 has no display outputs, making it unsuitable for any workstation display use, but its 54th percentile and 13470 average score show it is still a functional compute part for mining workloads.

The data indicates the Radeon Pro W5500 is the superior all-around card for any workload requiring display output, memory capacity, compute throughput, or professional application support. The P106-090 is only preferable when a 75 W card with a 192-bit bus and higher pixel rate is needed, in a mining-oriented build with no display requirement, and where the lower power draw and lower 250 W PSU suggestion are more important than the 2.2x compute advantage of the AMD card.