NVIDIA P106-100

The NVIDIA P106-100 is a Pascal-generation mining GPU built on TSMC's 16 nm process. It integrates 4,400 million transistors on a 200 mm² die, with 1,280 shading units, 80 texture mapping units, and 48 ROPs. The card runs at a base clock of 1506 MHz and a boost clock of 1709 MHz, paired with 6 GB of GDDR5 memory on a 192-bit bus delivering 192.2 GB/s of bandwidth. Notably, it has no display outputs, reflecting its mining-oriented design. In aggregate benchmarks, it achieves an average score of 22,950, placing it in the 66th percentile of all GPUs.

Benchmark Performance

The benchmark suite yields three distinct scores. In 3DMark Steel Nomad DX12, the P106-100 records 899 points, a result that stresses DirectX 12 graphics workloads. Geekbench OpenCL and Vulkan scores of 35,654 and 32,296 respectively indicate compute-heavy performance, with OpenCL edging out Vulkan by roughly 10%. The average of these three results, 22,950, is the figure used for percentile ranking. At the 66th percentile, the card outperforms the majority of GPUs in the database, though it sits close to several modern rivals.

Comparing the average score to the nearest rivals reveals a tight cluster. The P106-100 trails the AMD Radeon 760M by 0.2%, with the latter scoring 22,999. It is 0.3% behind the Intel Arc B580, which scores 23,021. Against the NVIDIA GeForce RTX 4060 Mobile, the P106-100 is 1% faster, as the mobile card averages 22,729. Conversely, the desktop RTX 4060 leads the P106-100 by 1%, scoring 23,181. These deltas are all within a single percentage point, indicating that the P106-100 delivers performance essentially equivalent to a broad set of contemporary integrated and discrete GPUs. The 3DMark Steel Nomad score of 899 is notably lower than the Geekbench results, but that test focuses on specific rendering features rather than raw compute throughput.

How It Compares

AMD Radeon 760M: The Radeon 760M averages 22,999, a score 0.2% higher than the P106-100's 22,950. This puts the two effectively at parity. The P106-100 holds a negligible deficit in aggregate performance, meaning that in compute or mining workloads the two would be nearly interchangeable. The 760M is an integrated solution, while the P106-100 is a discrete card, but their benchmark outcomes align closely.

Intel Arc B580: The Arc B580 scores 23,021, which is 0.3% above the P106-100. This is a marginal difference, well within typical run-to-run variance. The Arc B580 is a modern discrete GPU, yet the P106-100's older Pascal architecture keeps pace in these aggregate tests. The 3DMark Steel Nomad score of 899 for the P106-100 may be lower than the Arc's, but the overall average remains comparable.

NVIDIA GeForce RTX 4060 Mobile: The RTX 4060 Mobile averages 22,729, which is 1% lower than the P106-100's 22,950. This makes the P106-100 slightly faster in the combined benchmark suite, despite the RTX 4060 Mobile featuring newer architecture and additional features like ray tracing. The advantage is small, but it shows that the P106-100's raw compute throughput is competitive with a modern mid-range mobile GPU.

NVIDIA GeForce RTX 4060: The desktop RTX 4060 scores 23,181, 1% higher than the P106-100. This is the largest gap among the nearest rivals, yet still a slim margin. The RTX 4060 benefits from newer architecture and dedicated ray tracing cores, but in pure compute benchmarks the P106-100 is nearly as fast. The 66th percentile ranking reflects this tight grouping.

Memory Subsystem

The P106-100 is equipped with 6 GB of GDDR5 memory on a 192-bit bus, yielding a bandwidth of 192.2 GB/s. This configuration is typical for a mid-range card of its generation. The 6 GB capacity is sufficient for 1080p gaming and many compute tasks, but at higher resolutions—such as 1440p or 4K—the frame buffer may become a limiting factor. The 192.2 GB/s bandwidth determines how quickly textures and geometry data can be transferred to and from the GPU. In memory-intensive workloads, this figure is more critical than raw compute throughput. The card's memory clocks at 2002 MHz, which translates to an effective 8 Gbps per pin, a standard speed for GDDR5 of that era. For mining or compute applications that rely on large datasets, the 6 GB capacity is adequate, but larger memory pools would be preferable for very large models. The 192-bit bus width is a compromise between cost and performance, balancing bandwidth against the number of memory chips.

Who Should Consider It

Given that the P106-100 has no display outputs, it cannot be used as a conventional graphics card for gaming or desktop rendering. Its performance profile, indicated by the average benchmark score of 22,950 and the 66th percentile ranking, suggests it could handle 1080p gaming at medium to high settings if a display were available—but the absence of outputs makes that impossible. Instead, this card is positioned for compute or mining workloads where display connectivity is irrelevant. Users who already have a primary GPU for display output could employ the P106-100 as a secondary compute accelerator, offloading tasks such as OpenCL or Vulkan compute to the card. The 4.375 TFLOPS of FP32 performance and 68.36 GFLOPS of FP16 (at a 1:64 ratio) indicate a strong bias toward single-precision floating-point operations, which is well-suited for many scientific and machine-learning inference tasks. The card's 6 GB memory and 192.2 GB/s bandwidth are sufficient for moderate-sized datasets. For those seeking a dedicated compute card without display functionality, the P106-100 is a viable option, though its end-of-life status means availability may be limited.

Ray Tracing and Feature Set

The P106-100 does not include dedicated ray tracing or tensor cores; the fact pack lists both as null. This means the card relies entirely on its rasterization and compute units. It supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4, allowing it to run modern APIs and take advantage of features like asynchronous compute and explicit multi-adapter, but without hardware-accelerated ray tracing or AI-based tensor operations. The FP32 throughput of 4.375 TFLOPS is the primary compute capability, while FP16 is drastically reduced at 68.36 GFLOPS (1:64), indicating that half-precision workloads are not a focus. The card's pixel rate is 82.03 GPixel/s and texture rate is 136.7 GTexel/s, figures that reflect its 48 ROPs and 80 TMUs. For ray-traced workloads, the absence of RT cores means any ray tracing would have to be performed via compute shaders, which is inefficient. The Vulkan 1.4 support does enable modern graphics features, but the lack of dedicated hardware limits advanced effects.

Power and Cooling

The P106-100 has a TDP of 120 W, requiring a single 6-pin power connector. NVIDIA recommends a 300 W power supply, which is modest for a discrete GPU. The card occupies a dual-slot form factor and measures 250 mm (9.8 inches) in length, making it compatible with most mid-tower cases. The 16 nm manufacturing process and 4,400 million transistors contribute to its power efficiency, though the lack of display outputs means the card is entirely dedicated to compute or mining, where power draw is a primary consideration. The 120 W TDP is lower than many modern GPUs, which often exceed 200 W, but the P106-100's performance is correspondingly modest. For mining operations, the combination of 120 W power draw and 4.375 TFLOPS of FP32 compute provides a reasonable efficiency profile. The single 6-pin connector simplifies installation, and the 300 W PSU recommendation leaves ample headroom for a system with a modest CPU. The dual-slot cooler design is standard, and the 250 mm length fits in most enclosures. The card's end-of-life production status means that replacement coolers or support may be harder to find, but the specifications are well-documented.