AMD Radeon R9 M395X vs NVIDIA P106-100 Comparison

Head-to-Head Benchmarks

The recorded data contains a single shared benchmark result between these two GPUs, and it is a decisive victory for the NVIDIA P106-100. In the Geekbench OpenCL test, the P106-100 scored 35,951 points, while the AMD Radeon R9 M395X managed 17,829 points. The delta is a massive 50.4% in favor of the NVIDIA part, meaning the R9 M395X delivers less than half the compute throughput in this particular workload.

This result is striking because the context of each card's other benchmark appearances tells a more nuanced story. The R9 M395X shows its strength in the Geekbench Metal test, where it scores 33,953 points, a result that does not have a direct counterpart for the P106-100. Conversely, the P106-100 also posts a Geekbench Vulkan score of 32,897, a test the AMD card does not appear in. The only apples-to-apples comparison available is OpenCL, and there, the NVIDIA architecture simply runs away with it.

Looking at average benchmark scores across all recorded tests, the gap narrows considerably. The R9 M395X holds an average score of 25,891, placing it in the 71st percentile of all GPUs. The P106-100, despite winning the head-to-head test, has a lower average of 23,249, sitting in the 68th percentile. This suggests the P106-100's OpenCL advantage is not universal across every workload, and the R9 M395X benefits from a more consistent profile across the tests where it does appear.

The nearest rivals in the database reinforce this split. The R9 M395X sits within 0.6% of the NVIDIA GeForce RTX 3080 Ti Mobile (25,740) and is nearly identical to the AMD FirePro W7100 (25,856, delta 0.1%). The P106-100, on the other hand, is statistically tied with the AMD Radeon Pro Vega 16 (23,250, delta 0%) and sits just 0.3% behind the AMD Radeon AI PRO R9700 (23,315). These figures indicate that while the P106-100 wins the direct confrontation, both cards are firmly in the same performance tier, with the R9 M395X having a slight edge in overall standing.

FAQ

Q: Which GPU is faster in the only benchmark where both were tested?

A: The NVIDIA P106-100 wins decisively in Geekbench OpenCL, scoring 35,951 against the AMD Radeon R9 M395X's 17,829, a 50.4% advantage.

Q: Does the AMD card have any benchmark where it is competitive?

A: Yes. The R9 M395X scores 33,953 in Geekbench Metal, which is higher than the P106-100's Geekbench Vulkan score of 32,897, though these are different APIs and not directly comparable.

Q: How do their overall average scores compare?

A: The AMD Radeon R9 M395X has a higher average benchmark score of 25,891 versus 23,249 for the NVIDIA P106-100, a difference of 2,642 points.

Q: What do the percentile rankings say about each card?

A: The R9 M395X ranks in the 71st percentile of all GPUs, while the P106-100 ranks in the 68th percentile, indicating the AMD card is slightly better positioned overall.

Q: Are these cards closely matched to any specific rivals?

A: The R9 M395X is within 0.1% of the AMD FirePro W7100 (25,856) and within 0.6% of the NVIDIA GeForce RTX 3080 Ti Mobile (25,740). The P106-100 is essentially tied with the AMD Radeon Pro Vega 16 (23,250, delta 0%) and within 0.3% of the AMD Radeon AI PRO R9700 (23,315).

Q: Which card has more benchmark wins in the head-to-head comparison?

A: The NVIDIA P106-100 holds the sole win, with 1 win versus 0 for the AMD Radeon R9 M395X in the recorded head-to-head data.

Architecture Differences

The two GPUs come from fundamentally different design philosophies. The AMD Radeon R9 M395X uses the Amethyst chip built on GCN 3.0 architecture, fabricated on a 28 nm process at TSMC. This older, larger node results in a die size of 366 mm² and a transistor count of 5,000 million, yielding a transistor density of 13.7 million per square millimeter. The NVIDIA P106-100, in contrast, uses the GP106 chip with the Pascal architecture, manufactured on a much more advanced 16 nm process, also at TSMC. This allows a smaller die of 200 mm² with 4,400 million transistors, achieving a higher density of 22.0 million per square millimeter.

The compute layouts diverge sharply. The R9 M395X packs 2,048 shading units, 128 texture mapping units, and 32 raster operation units. The P106-100 has fewer shading units at 1,280, and fewer texture units at 80, but it compensates with a higher ROP count of 48. The AMD card's raw throughput is rated at 2.961 TFLOPS for both FP32 and FP16, indicating a 1:1 ratio. The NVIDIA card is rated higher for FP32 at 4.375 TFLOPS, but its FP16 output is drastically reduced to 68.36 GFLOPS, a 1:64 ratio, showing that the Pascal architecture is heavily optimized for single-precision work at the expense of half-precision.

Memory bandwidth tells a different story. The R9 M395X features 8 GB of GDDR5 on a 256-bit bus, delivering 160.0 GB/s. The P106-100 has 6 GB of GDDR5 on a narrower 192-bit bus, yet it achieves higher bandwidth at 192.2 GB/s due to a faster memory clock of 2002 MHz (8 Gbps effective) versus the AMD card's 1250 MHz (5 Gbps effective). The NVIDIA card also leads in pixel and texture rates: 82.03 GPixel/s versus 23.14 GPixel/s, and 136.7 GTexel/s versus 92.54 GTexel/s, respectively.

API support is similar but not identical. Both support DirectX 12, but the P106-100 reaches feature level 12_1 while the R9 M395X stops at 12_0. OpenGL 4.6 is common to both. The Vulkan versions differ: the R9 M395X supports Vulkan 1.2.170, while the P106-100 lists Vulkan 1.4. The NVIDIA card also uses a PCIe 1.0 x16 bus interface, an unusual choice that likely reflects its mining-focused design, while the AMD card uses the more standard PCIe 3.0 x16.

Specification Differences

The most obvious difference is memory capacity: the AMD Radeon R9 M395X offers 8 GB, while the NVIDIA P106-100 offers 6 GB. The bus widths also differ, with the AMD card at 256-bit and the NVIDIA card at 192-bit. Despite the narrower bus, the P106-100 achieves higher memory bandwidth at 192.2 GB/s versus 160.0 GB/s, as noted above.

Clock speeds are only partially specified for the AMD card. The R9 M395X has no listed base or boost clock, only a memory clock of 1250 MHz (5 Gbps effective). The P106-100 has a base clock of 1506 MHz and a boost clock of 1709 MHz, with a memory clock of 2002 MHz (8 Gbps effective). The power profiles are distinct: the R9 M395X is rated at 75 W TDP with no power connectors, while the P106-100 draws 120 W and requires a single 6-pin connector, with a suggested PSU of 300 W.

Form factors differ significantly. The AMD card is an MXM Module with display outputs described as "Portable Device Dependent," reflecting its mobile-oriented design. The NVIDIA card is a Dual-slot card measuring 250 mm (9.8 inches) in length, with no display outputs at all, a clear sign of its mining-specific purpose. The bus interface also differs, with the AMD card on PCIe 3.0 x16 and the NVIDIA card on PCIe 1.0 x16. Release dates are recorded: the R9 M395X launched on 2015-05-04, while the P106-100 launched on 2017-06-18. Both are end-of-life products.

The Verdict

The data presents a curious case. The NVIDIA P106-100 wins the only direct benchmark comparison by a wide margin, 50.4% in Geekbench OpenCL. Its compute throughput of 4.375 TFLOPS FP32 is substantially higher than the AMD card's 2.961 TFLOPS. For any workload that relies heavily on OpenCL compute, the P106-100 is clearly the superior choice based on measurements.

However, the average benchmark scores tell a different story. The AMD Radeon R9 M395X has a higher overall average of 25,891 and a better percentile ranking at 71st versus 68th. This suggests that the P106-100's OpenCL dominance does not translate into universal superiority across all test types. The AMD card also offers more memory (8 GB versus 6 GB), a wider memory bus, and a much lower power draw at 75 W versus 120 W.

The P106-100's lack of display outputs and its dual-slot, 250 mm form factor make it unsuitable for conventional desktop use, while the R9 M395X's MXM form factor and portable-device-dependent outputs target mobile systems. The choice depends entirely on the intended application. For raw compute in a mining or server context where OpenCL performance is paramount, the P106-100 is the data-backed winner. For a more balanced profile with higher average scores, more memory, and lower power consumption, the R9 M395X holds the edge.

Where Each One Wins

AMD Radeon R9 M395X wins on overall standing and capacity. Its average benchmark score of 25,891 exceeds the P106-100's 23,249, and its 71st percentile rank is three points higher. The 8 GB memory capacity is 2 GB larger, and the 256-bit bus provides a wider data path, even if total bandwidth is lower. The 75 W TDP makes it a far more power-efficient option, and its MXM form factor with portable-device-dependent outputs means it can be integrated into systems where the P106-100, with no outputs, cannot function as a display adapter. Its Geekbench Metal score of 33,953 also demonstrates strong performance in Apple-oriented compute workloads.

NVIDIA P106-100 wins on raw compute and modern process technology. The 50.4% lead in Geekbench OpenCL is the single largest recorded difference between the two cards. Its 4.375 TFLOPS FP32 performance is 47.8% higher than the AMD card's 2.961 TFLOPS. The 16 nm process node allows a smaller die (200 mm² versus 366 mm²) with higher transistor density (22.0M / mm² versus 13.7M / mm²). It also delivers higher pixel rate (82.03 GPixel/s versus 23.14 GPixel/s), higher texture rate (136.7 GTexel/s versus 92.54 GTexel/s), and higher memory bandwidth (192.2 GB/s versus 160.0 GB/s). Its DirectX 12_1 support and Vulkan 1.4 API compatibility are more recent than the AMD card's 12_0 and 1.2.170.