AMD Radeon HD 8950M vs NVIDIA P106-090 Comparison

The Verdict

The benchmark data positions the NVIDIA P106-090 and AMD Radeon HD 8950M as near-equals in aggregate performance, yet their individual profiles diverge sharply. The NVIDIA P106-090 holds a 54.7% lead in Geekbench OpenCL and a 43.3% lead in Geekbench Vulkan, securing both head-to-head wins. The AMD Radeon HD 8950M, meanwhile, posts no wins in the direct comparison. However, the average benchmark scores tell a closer story: the P106-090 averages 13,470, while the HD 8950M averages 13,376, a gap of only 0.7%. This suggests that the HD 8950M's strengths lie outside the tested workloads, or that its performance is more consistent across a broader suite.

For users prioritizing compute workloads—OpenCL or Vulkan—the P106-090 is the clear choice based on its substantial margins. For those needing a mobile form factor with display outputs, the HD 8950M is the only viable option, as the P106-090 has no display outputs and is a dual-slot card. The P106-090 is a mining-specific product with a PCIe 1.0 x1 interface, which limits its utility in standard desktop configurations. The HD 8950M, by contrast, uses PCIe 3.0 x16 and is designed as an MXM module for portable devices. If raw compute is the sole criterion, the data favors the P106-090 decisively; if system integration and display capability matter, the HD 8950M is the practical pick.

Where Each One Wins

The NVIDIA P106-090 wins decisively in both recorded benchmarks. In Geekbench OpenCL, it scores 21,304 versus 13,772 for the HD 8950M, a delta of 54.7%. In Geekbench Vulkan, it scores 18,596 versus 12,979, a delta of 43.3%. These are not marginal victories; they represent a generational leap in compute throughput. The P106-090's FP32 performance is 2.352 TFLOPS, compared to 1.651 TFLOPS for the HD 8950M, a 42.5% advantage. Its pixel rate is 73.49 GPixel/s versus 17.20 GPixel/s for the HD 8950M, and its texture rate is 73.49 GTexel/s versus 51.60 GTexel/s. Memory bandwidth also favors the P106-090 at 192.2 GB/s versus 88.00 GB/s.

The AMD Radeon HD 8950M does not win any benchmark in the head-to-head data. Its only comparative advantage is in form factor and connectivity. It is an MXM module with display outputs that are portable-device dependent, whereas the P106-090 is a dual-slot card with no outputs and a PCIe 1.0 x1 interface. The HD 8950M also supports a wider bus interface (PCIe 3.0 x16) and has a lower transistor density (13.0M / mm² versus 22.0M / mm²), but this does not translate into performance wins. In terms of API support, the P106-090 offers Vulkan 1.4 and DirectX 12 (12_1), while the HD 8950M offers Vulkan 1.2.170 and DirectX 12 (12_0). The P106-090 also lists FP16 compute at 36.74 GFLOPS (1:64), a feature absent from the HD 8950M's data.

Architecture Differences

The two GPUs come from different architectural eras and process nodes. The NVIDIA P106-090 uses the GP106 chip with the Pascal architecture, built on a 16 nm TSMC process. It integrates 4,400 million transistors on a 200 mm² die, yielding a transistor density of 22.0M / mm². The AMD Radeon HD 8950M uses the Saturn chip with the GCN 2.0 architecture, built on a 28 nm TSMC process. It integrates 2,080 million transistors on a 160 mm² die, yielding a density of 13.0M / mm². The P106-090's newer node allows nearly double the transistor count in a 25% larger die area.

The memory subsystems differ significantly. The P106-090 has 3 GB of GDDR5 on a 192-bit bus, delivering 192.2 GB/s of bandwidth. The HD 8950M has 2 GB of GDDR5 on a 128-bit bus, delivering 88.00 GB/s. Memory clocks are 2002 MHz (8 Gbps effective) for the P106-090 and 1375 MHz (5.5 Gbps effective) for the HD 8950M. Both have 768 shading units and 48 TMUs, but the ROP count differs: 48 for the P106-090 versus 16 for the HD 8950M. This explains the P106-090's massive pixel rate advantage (73.49 GPixel/s versus 17.20 GPixel/s).

The P106-090's clock speeds are fixed at 1354 MHz base and 1531 MHz boost, while the HD 8950M lists no base or boost clocks. The P106-090 has a TDP of 75 W with a suggested PSU of 250 W and a single 6-pin power connector. The HD 8950M has a TDP of 100 W with no power connector or PSU guidance listed. The P106-090's bus interface is PCIe 1.0 x1, a severe limitation for a desktop card, while the HD 8950M uses PCIe 3.0 x16. The P106-090 is 250 mm (9.8 inches) long, while the HD 8950M has no listed dimensions. The P106-090's production status is end-of-life, as is the HD 8950M's.

In terms of API support, the P106-090 offers DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The HD 8950M offers DirectX 12 (12_0), OpenGL 4.6, and Vulkan 1.2.170. The P106-090's Vulkan version is newer, which may contribute to its 43.3% lead in the Vulkan benchmark. The HD 8950M's predecessor is London and its successor is Gem System, while the P106-090 has no listed predecessor or successor. The P106-090's generation is "Mining GPUs," while the HD 8950M's generation is "Solar System (HD 8900M)."

FAQ

Q: Which GPU has higher raw compute performance?

A: The NVIDIA P106-090 has higher FP32 throughput at 2.352 TFLOPS, versus 1.651 TFLOPS for the AMD Radeon HD 8950M. This translates to a 54.7% lead in Geekbench OpenCL (21,304 versus 13,772).

Q: Can the NVIDIA P106-090 be used for display output?

A: No. The P106-090 has no display outputs, as it is designed for mining applications. The AMD Radeon HD 8950M, by contrast, has display outputs that are portable-device dependent.

Q: What are the memory bandwidth differences?

A: The P106-090 offers 192.2 GB/s over a 192-bit bus with 3 GB GDDR5, while the HD 8950M offers 88.00 GB/s over a 128-bit bus with 2 GB GDDR5. The P106-090's memory clock is 2002 MHz (8 Gbps effective) versus 1375 MHz (5.5 Gbps effective) for the HD 8950M.

Q: How do their form factors differ?

A: The P106-090 is a dual-slot, 250 mm (9.8 inches) long card requiring a 6-pin power connector, with a PCIe 1.0 x1 interface. The HD 8950M is an MXM module with PCIe 3.0 x16 and no listed power connector or dimensions.

Q: Which GPU has better Vulkan performance?

A: The P106-090 scores 18,596 in Geekbench Vulkan, which is 43.3% higher than the HD 8950M's 12,979. The P106-090 also supports Vulkan 1.4, while the HD 8950M supports Vulkan 1.2.170.

Q: Are these GPUs still in production?

A: Both are end-of-life. The P106-090 was released on 2017-07-30, and the HD 8950M was released on 2013-05-13. Neither has a launch MSRP listed in the data.

Head-to-Head Benchmarks

The head-to-head data includes two benchmarks, and the NVIDIA P106-090 wins both. In Geekbench OpenCL, the P106-090 scores 21,304 against the HD 8950M's 13,772, a delta of 54.7%. This is the larger of the two wins and reflects the P106-090's superior FP32 throughput, texture rate (73.49 GTexel/s versus 51.60 GTexel/s), and memory bandwidth. The Geekbench Vulkan test shows a similar pattern: the P106-090 scores 18,596, while the HD 8950M scores 12,979, a delta of 43.3%. The P106-090's newer Vulkan 1.4 support likely contributes to this margin, as does its higher ROP count (48 versus 16), which accelerates pixel-heavy workloads.

The average benchmark scores from the nearestRivals data provide context. The P106-090's average is 13,470, which places it 0.3% below the NVIDIA GeForce GTX 570 (13,515) and 0.5% above the AMD Radeon Pro 555 (13,407). It is also 0.5% below the AMD Radeon HD 7770M (13,536) and 0.5% below the AMD Radeon RX 9070 XT (13,543). The HD 8950M's average is 13,376, which is 0.2% above the AMD Radeon RX 5500M (13,356) and 0.2% above the AMD FirePro M6100 (13,354). It is 0.2% below the AMD Radeon Pro 555 (13,407) and 0.4% below the AMD Radeon Pro 555X (13,321). Both cards sit at the 54th percentile of all GPUs, indicating they are statistically indistinguishable in aggregate performance despite the large head-to-head deltas.

The P106-090's wins are consistent across both APIs, but the OpenCL margin (54.7%) is larger than the Vulkan margin (43.3%). This suggests that the P106-090's advantage is more pronounced in compute-heavy OpenCL workloads than in graphics-oriented Vulkan tasks. The HD 8950M, with its GCN 2.0 architecture and older 28 nm node, cannot match the Pascal-based P106-090's throughput. The P106-090's 48 ROPs versus 16 for the HD 8950M explain the pixel rate gap (73.49 GPixel/s versus 17.20 GPixel/s), which is a 4.3x difference. The texture rate gap is smaller but still significant: 73.49 GTexel/s versus 51.60 GTexel/s, a 42.4% difference.

Memory bandwidth is another differentiator. The P106-090's 192.2 GB/s is 2.18x the HD 8950M's 88.00 GB/s. This directly impacts OpenCL and Vulkan performance, as both benchmarks rely heavily on memory throughput for large datasets. The P106-090 also has more VRAM (3 GB versus 2 GB), which can reduce memory pressure in larger workloads. The HD 8950M's lower TDP (100 W versus 75 W) is notable, but the P106-090's higher performance per watt is evident from the benchmark results. The HD 8950M's only advantages are its PCIe 3.0 x16 interface and display outputs, neither of which affects compute scores.