AMD Radeon R9 M390X vs NVIDIA P106-100 Comparison

Head-to-Head Benchmarks

The recorded data shows a single direct comparison between these two GPUs, and the result is decisive. In the Geekbench Vulkan test, the NVIDIA P106-100 scores 32,897 points against 16,951 for the AMD Radeon R9 M390X. That is a 94.1% advantage for the NVIDIA part, meaning it delivers nearly double the Vulkan performance of the AMD mobile chip. This is not a marginal win; it is a dominant margin that defines the entire head-to-head picture.

Looking at the broader database averages reinforces that gap. The NVIDIA P106-100 holds an average benchmark score of 23,249 across all recorded tests, while the AMD Radeon R9 M390X sits at 20,662. That is a roughly 12.5% overall lead for the NVIDIA part when all workloads are considered. The Vulkan result is much larger than the overall average gap, which suggests the NVIDIA architecture scales particularly well in compute-oriented API workloads compared to the older GCN design.

The percentile rankings tell a similar story. The P106-100 lands in the 68th percentile of all GPUs in the database, while the R9 M390X sits in the 66th percentile. Both are mid-pack performers, but the NVIDIA card is positioned slightly higher in the overall distribution. The nearest rivals for each card provide context for where they sit relative to the broader market. The P106-100's closest competitor is the AMD Radeon Pro Vega 16 with an average score of 23,250, effectively a tie at a 0% delta. The R9 M390X's closest rival is the AMD Radeon RX 5600 XT at 20,713, which is 0.2% behind the M390X. These rival placements show that both cards are bracketed by significantly newer or higher-tier parts, yet the P106-100 holds a clear edge over the M390X in the one test where they directly meet.

The Geekbench OpenCL result for the P106-100 is 35,951, which exceeds its Vulkan score by roughly 9%. The R9 M390X has no OpenCL result in the database, but its Metal score is 24,373. That Metal figure is 43.8% higher than its Vulkan score, indicating the AMD card performs relatively better under Apple's Metal API than under Vulkan. Even so, the NVIDIA card's OpenCL score is 47.5% higher than the AMD card's best recorded result, which is the Metal score. The data consistently points in one direction: the P106-100 is the stronger compute performer across every comparable metric.

Architecture Differences

The two GPUs come from fundamentally different design generations. The NVIDIA P106-100 uses the GP106 chip built on the Pascal architecture, manufactured on a 16 nm process at TSMC. The AMD Radeon R9 M390X uses the Amethyst chip based on GCN 3.0, built on a 28 nm process, also at TSMC. The process node difference is significant: 16 nm versus 28 nm. This explains a great deal of the efficiency and density gap between the two parts.

Transistor counts and die sizes illustrate the design philosophy divergence. The NVIDIA chip packs 4,400 million transistors into a 200 mm² die, yielding a transistor density of 22.0 million per square millimeter. The AMD chip has more transistors overall at 5,000 million, but they are spread across a much larger 366 mm² die, producing a density of only 13.7 million per square millimeter. The Pascal design achieves 60.6% higher transistor density than the GCN 3.0 part, which is a direct consequence of the more advanced manufacturing process.

The compute resources are arranged very differently. The P106-100 has 1,280 shading units, 80 texture mapping units, and 48 render output units. The R9 M390X has 2,048 shading units, 128 TMUs, and only 32 ROPs. AMD's part has 60% more shaders and 60% more TMUs, but NVIDIA's part has 50% more ROPs. Despite having fewer shaders, the P106-100 achieves 4.375 TFLOPS of FP32 performance versus 2.961 TFLOPS for the R9 M390X. That is a 47.8% advantage for the smaller NVIDIA chip, driven by its much higher clock speeds.

Clock speeds are a major differentiator. The P106-100 runs at a base clock of 1506 MHz and boosts to 1709 MHz. The R9 M390X has no base or boost clock listed in the database, but its memory clock is 1250 MHz with 5 Gbps effective speed. The NVIDIA card's memory runs at 2002 MHz with 8 Gbps effective. The P106-100 memory bandwidth is 192.2 GB/s over a 192-bit bus, while the R9 M390X delivers 160.0 GB/s over a wider 256-bit bus. The NVIDIA part achieves 20.1% higher bandwidth despite a narrower interface, again thanks to faster memory clocks.

The R9 M390X supports FP16 at a 1:1 ratio with FP32, delivering 2.961 TFLOPS. The P106-100 has a 1:64 FP16 ratio, delivering only 68.36 GFLOPS. This is a notable difference for workloads that use half-precision math. In terms of pixel and texture throughput, the P106-100 hits 82.03 GPixel/s and 136.7 GTexel/s, while the R9 M390X manages 23.14 GPixel/s and 92.54 GTexel/s. The NVIDIA part leads by 3.5x in pixel rate and 47.7% in texture rate.

API support differs as well. The P106-100 supports DirectX 12_1, OpenGL 4.6, and Vulkan 1.4. The R9 M390X supports DirectX 12_0, OpenGL 4.6, and Vulkan 1.2.170. The newer Vulkan revision on the NVIDIA side aligns with its dominant Vulkan benchmark result. The AMD card has no display outputs, while the NVIDIA card also has no outputs, though the AMD part is described as "Portable Device Dependent" for display connectivity.

Where Each One Wins

The NVIDIA P106-100 wins decisively in Vulkan compute. The Geekbench Vulkan score of 32,897 versus 16,951 represents a 94.1% lead. This makes the P106-100 the clear choice for any workload that relies on the Vulkan API, whether gaming or general compute. Its FP32 throughput of 4.375 TFLOPS also gives it a 47.8% advantage over the R9 M390X, so numerically intensive single-precision tasks will favor the NVIDIA part.

The P106-100 also wins on memory bandwidth. Its 192.2 GB/s exceeds the R9 M390X's 160.0 GB/s by 20.1%. For texture-heavy workloads, the NVIDIA card's 136.7 GTexel/s beats the AMD card's 92.54 GTexel/s by 47.7%. Pixel throughput is even more lopsided: 82.03 GPixel/s versus 23.14 GPixel/s, a 3.5x advantage. Fill-rate bound scenarios will strongly favor the P106-100.

The R9 M390X wins in shader count. Its 2,048 shading units outnumber the P106-100's 1,280 by 60%. TMU count also favors AMD at 128 versus 80, a 60% advantage. In theory, workloads that scale perfectly with shader count and are not clock-limited, AMD's part has the structural edge. FP16 compute is the only clear numerical win for AMD: 2.961 TFLOPS at 1:1 ratio, massively ahead of the P106-100's 68.36 GFLOPS at 1:64 ratio.

Power consumption favors AMD. The R9 M390X has a TDP of 75 W with no power connectors required, fitting into an MXM Module form factor. The P106-100 has a 120 W TDP and requires a 1x 6-pin power connector with a 300 W suggested PSU. For constrained power environments, the R9 M390X is the lighter load. The AMD card also uses a PCIe 3.0 x16 interface, while the NVIDIA card lists PCIe 1.0 x16, which is an unusual limitation for the P106-100.

The R9 M390X is the only one with a Metal benchmark result in the database. Its Geekbench Metal score of 24,373 is its best recorded performance. This indicates that in Apple ecosystem contexts, the AMD part is the functional option, while the P106-100 has no recorded Metal results. For portable device integration, the R9 M390X's form factor and power profile make it the mobile-friendly pick.

Specification Differences

Memory configuration: The P106-100 has 6 GB of GDDR5 on a 192-bit bus with 2002 MHz memory clock and 8 Gbps effective speed. The R9 M390X has 4 GB of GDDR5 on a 256-bit bus with 1250 MHz memory clock and 5 Gbps effective speed. The P106-100 has 50% more memory capacity and 20.1% more bandwidth.

Process node: The P106-100 is on 16 nm at TSMC, while the R9 M390X is on 28 nm at TSMC. Transistor density is 22.0M per mm² for NVIDIA versus 13.7M per mm² for AMD. Die size is 200 mm² for the P106-100 and 366 mm² for the R9 M390X. Transistor count is 4,400 million for NVIDIA and 5,000 million for AMD.

Compute units: The P106-100 uses 1,280 shaders, 80 TMUs, and 48 ROPs. The R9 M390X uses 2,048 shaders, 128 TMUs, and 32 ROPs. Clock speeds: the P106-100 has a 1506 MHz base and 1709 MHz boost. The R9 M390X has no base or boost clock listed. The P106-100 FP32 is 4.375 TFLOPS; the R9 M390X FP32 is 2.961 TFLOPS.

Outputs and interface: The P106-100 has no display outputs, a PCIe 1.0 x16 interface, and measures 250 mm or 9.8 inches in length. The R9 M390X also has no direct display outputs, is described as "Portable Device Dependent," uses PCIe 3.0 x16, and has no listed dimensions. The P106-100 is dual-slot with a 1x 6-pin connector and 120 W TDP. The R9 M390X is an MXM Module with no power connectors and 75 W TDP.

Release timing: The P106-100 was released in June 2017. The R9 M390X was released in May 2015. Both are end-of-life. The R9 M390X lists a predecessor called "Solar System" and a successor called "Polaris Mobile," while the P106-100 has no predecessor or successor listed.

FAQ

Q: Which GPU is faster in Vulkan workloads?

A: The NVIDIA P106-100 scores 32,897 in Geekbench Vulkan versus 16,951 for the AMD Radeon R9 M390X, a 94.1% advantage for the NVIDIA card.

Q: How do their average benchmark scores compare?

A: The P106-100 has an average benchmark score of 23,249, while the R9 M390X averages 20,662. The P106-100 sits in the 68th percentile of all GPUs, and the R9 M390X sits in the 66th percentile.

Q: Which card has more memory and bandwidth?

A: The P106-100 has 6 GB of GDDR5 with 192.2 GB/s bandwidth. The R9 M390X has 4 GB of GDDR5 with 160.0 GB/s bandwidth.

Q: What are the FP32 performance figures?

A: The P106-100 delivers 4.375 TFLOPS of FP32 performance, which is 47.8% higher than the R9 M390X's 2.961 TFLOPS.

Q: Which GPU is more power-efficient?

A: The R9 M390X has a 75 W TDP with no power connectors, while the P106-100 has a 120 W TDP and requires a 1x 6-pin connector with a 300 W suggested PSU.

Q: Which card supports half-precision FP16 compute better?

A: The R9 M390X supports FP16 at a 1:1 ratio with FP32, delivering 2.961 TFLOPS. The P106-100 has a 1:64 ratio, delivering only 68.36 GFLOPS.

The Verdict

The data points to the NVIDIA P106-100 as the stronger performer in every directly comparable metric. The Vulkan benchmark shows a 94.1% lead over the R9 M390X. The P106-100 also leads in average benchmark score by roughly 12.5%, in FP32 throughput by 47.8%, in memory bandwidth by 20.1%, in texture rate by 47.7%, and in pixel rate by a 3.5x margin.

The R9 M390X retains advantages in specific areas. It has more shading units and TMUs, supports FP16 at a 1:1 ratio, draws less power at 75 W versus 120 W, uses a PCIe 3.0 x16 interface, and fits an MXM Module form factor. It also has a recorded Metal score of 24,373, making it the only one of the two with Apple ecosystem support in the database.

For most users, the P106-100 is the pick. It delivers higher raw performance, more memory, greater bandwidth, and a dominant Vulkan result. For those constrained by power or form factor, particularly in portable or low-power systems, the R9 M390X offers a lighter footprint and no external power requirement. The decision comes down to whether raw throughput or power efficiency matters more. The benchmark data favors the NVIDIA part in performance, while the AMD part wins on efficiency and form factor flexibility.