Intel Arc A350M vs NVIDIA P106-100 Comparison

Intel Arc A350M and NVIDIA P106-100 represent two very different approaches to graphics hardware, separated by five years of architectural evolution and targeting distinct use cases. The benchmark data shows a clear overall winner in raw compute, but the story is more nuanced when considering the feature sets and target platforms of each card. The NVIDIA P106-100, a mining-oriented Pascal part from 2017, leads in both available benchmark tests, while the Intel Arc A350M, a 2022 Alchemist mobile chip, counters with modern API support and dramatically lower power requirements.

Head-to-Head Benchmarks

The head-to-head results are unambiguous in favor of the NVIDIA P106-100 across both synthetic workloads. In Geekbench OpenCL, the P106-100 scores 35,951 against the Arc A350M’s 24,546, a substantial 31.7% advantage for the NVIDIA part. This is a significant margin, especially considering the P106-100’s older Pascal architecture and its origin as a mining-specific product with no display outputs. The OpenCL test heavily favors the P106-100’s raw compute throughput, which is backed by its higher FP32 rating of 4.375 TFLOPS compared to the Arc A350M’s 3.379 TFLOPS.

The Vulkan results tell a similar story, though with a slightly narrower gap. The P106-100 posts 32,897 points in Geekbench Vulkan, while the Arc A350M manages 24,747, giving NVIDIA a 24.8% lead. This margin suggests that while both architectures handle modern graphics APIs, the Pascal design’s higher shading unit count (1280 vs 768) and wider memory bus (192-bit vs 64-bit) provide a tangible performance advantage in compute-heavy workloads. The Arc A350M’s Xe-HPG architecture brings newer features like ray tracing and DirectX 12 Ultimate support, but these do not translate into higher scores in these particular benchmarks.

The aggregate benchmark data reinforces this hierarchy. The Arc A350M’s average benchmark score sits at 24,647, placing it in the 70th percentile of all GPUs. Its nearest rivals include the AMD Radeon RX 590 (24,744, 0.4% higher), the NVIDIA RTX A5000 Mobile (24,763, 0.5% higher), and the AMD Radeon RX 6600 XT (24,442, 0.8% lower). These are tight margins, indicating the Arc A350M performs at a level consistent with mid-range desktop and mobile parts from the previous generation. The P106-100, by contrast, averages 23,249 across its benchmarks (including a 3DMark Steel Nomad DX12 score of 899), placing it in the 68th percentile. Its nearest rivals are the AMD Radeon Pro Vega 16 (23,250, 0% delta), the AMD Radeon RX 6600M (23,273, 0.1% lower), and the AMD Radeon R9 M290X (23,276, 0.1% lower).

Interestingly, the P106-100’s average score is actually lower than the Arc A350M’s average, despite winning both head-to-head tests. This discrepancy arises because the P106-100’s benchmark set includes the 3DMark Steel Nomad DX12 test, which scores only 899, dragging down its average. When isolating the shared Geekbench tests, the P106-100 is clearly ahead. The delta between the two cards in OpenCL is 31.7%, and in Vulkan it is 24.8% — both decisive victories for NVIDIA.

Where Each One Wins

The NVIDIA P106-100 wins decisively in raw compute performance, which is its primary strength. Its 1,280 shading units, 80 texture mapping units, and 48 ROPs, combined with a 192-bit memory bus delivering 192.2 GB/s of bandwidth, make it a capable workhorse for parallel processing tasks. The Geekbench OpenCL score of 35,951 and Vulkan score of 32,897 both exceed the Arc A350M’s results by significant margins. This makes the P106-100 the better choice for compute-heavy applications like rendering, data processing, or any workload that leverages general-purpose GPU computing. Its 6 GB of GDDR5 memory also doubles the Arc A350M’s 4 GB capacity, which could be advantageous for holding larger datasets in VRAM.

The Intel Arc A350M wins on efficiency and modern feature support. Its 25 W TDP is a fraction of the P106-100’s 120 W, making it suitable for thin-and-light laptops where power and thermal constraints are paramount. The Arc A350M also supports DirectX 12 Ultimate (12_2), whereas the P106-100 is limited to DirectX 12 (12_1). This means the Intel part can execute ray tracing workloads via its 6 dedicated RT cores, a feature entirely absent from the Pascal-based NVIDIA card. The Arc A350M also uses PCIe 4.0 x8, doubling the bandwidth per lane compared to the P106-100’s PCIe 1.0 x16 interface, though the latter’s full x16 width partially compensates. For users prioritizing portability, modern API compatibility, and ray tracing capability, the Arc A350M is the clear winner despite its lower benchmark scores.

The P106-100 also has a physical size advantage in terms of board complexity, but this is irrelevant for most users since it is a mining card with no display outputs. The Arc A350M is an IGP (integrated graphics processor) with portable-device-dependent outputs, meaning it is designed to be embedded in a laptop motherboard. The P106-100 is a dual-slot, 250 mm (9.8 inches) card with a 1x 6-pin power connector and a suggested PSU of 300 W. These are fundamentally different form factors serving different markets; the P106-100 targets desktop mining rigs, while the Arc A350M targets mobile computing.

FAQ

Q: Which GPU has a higher average benchmark score?

A: The Intel Arc A350M has a higher average benchmark score of 24,647, compared to the NVIDIA P106-100’s 23,249. However, this average is skewed because the P106-100’s benchmark set includes a 3DMark Steel Nomad DX12 score of 899, which is much lower than its Geekbench results.

Q: How much faster is the NVIDIA P106-100 in OpenCL?

A: The P106-100 scores 35,951 in Geekbench OpenCL, while the Arc A350M scores 24,546, a 31.7% difference. This is the largest performance gap between the two cards in any shared benchmark.

Q: Does the Intel Arc A350M support ray tracing?

A: Yes, the Arc A350M has 6 dedicated RT cores and supports DirectX 12 Ultimate (12_2), which includes ray tracing capabilities. The NVIDIA P106-100 has no RT cores and is limited to DirectX 12 (12_1).

Q: What is the power consumption difference between the two cards?

A: The Arc A350M has a TDP of 25 W, while the P106-100 has a TDP of 120 W. This makes the Intel part significantly more power-efficient, suited for mobile devices, whereas the NVIDIA card requires a desktop power supply rated at 300 W.

Q: Which card has more memory bandwidth?

A: The NVIDIA P106-100 has a 192-bit memory bus providing 192.2 GB/s of bandwidth, while the Arc A350M has a 64-bit bus providing 112.0 GB/s. The P106-100 also has 6 GB of GDDR5 memory compared to 4 GB of GDDR6 on the Arc A350M.

Q: Are both cards still in production?

A: No, both are end-of-life products. The NVIDIA P106-100 was released in June 2017, and the Intel Arc A350M was released in March 2022.

Specification Differences

The two cards differ fundamentally in nearly every specification. The Intel Arc A350M uses a 6 nm process node from TSMC, packing 7,200 million transistors into a 157 mm² die with a density of 45.9M transistors per mm². The NVIDIA P106-100 uses a 16 nm process, also from TSMC, with 4,400 million transistors on a larger 200 mm² die, resulting in a lower density of 22.0M transistors per mm². The Arc A350M has a base clock of 1150 MHz and a boost clock of 2200 MHz, while the P106-100 runs at a higher base clock of 1506 MHz but a lower boost clock of 1709 MHz.

Memory configurations differ sharply: the Arc A350M has 4 GB of GDDR6 on a 64-bit bus with 112.0 GB/s bandwidth and an effective memory speed of 14 Gbps. The P106-100 offers 6 GB of GDDR5 on a 192-bit bus with 192.2 GB/s bandwidth and an effective speed of 8 Gbps. The Arc A350M has 768 shading units, 48 TMUs, and 24 ROPs, while the P106-100 has 1,280 shading units, 80 TMUs, and 48 ROPs. Pixel rates are 52.80 GPixel/s for Intel and 82.03 GPixel/s for NVIDIA; texture rates are 105.6 GTexel/s for Intel and 136.7 GTexel/s for NVIDIA. FP32 compute is 3.379 TFLOPS for Intel versus 4.375 TFLOPS for NVIDIA, while FP16 is 6.758 TFLOPS (2:1) for Intel versus 68.36 GFLOPS (1:64) for NVIDIA — a massive difference in half-precision capability.

Power and physical specs also diverge: the Arc A350M is an IGP with 25 W TDP and no power connectors, while the P106-100 is a dual-slot card with 120 W TDP, a 1x 6-pin power connector, and a 300 W suggested PSU. The Arc A350M uses PCIe 4.0 x8, while the P106-100 uses PCIe 1.0 x16. Display outputs are portable-device-dependent for the Intel card, while the NVIDIA card has no outputs at all. The P106-100 is 250 mm (9.8 inches) long; the Arc A350M has no listed dimensions.

Architecture Differences

The architectural divide is stark, reflecting a five-year generation gap. The Intel Arc A350M is built on the Xe-HPG architecture with the DG2-128 chip, part of the Alchemist generation (Arc 3 Mobile), fabricated on TSMC’s 6 nm node. It includes 6 RT cores for ray tracing and supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, making it fully modern in API coverage. The NVIDIA P106-100 uses the older Pascal architecture with the GP106 chip, from the Mining GPUs generation, on TSMC’s 16 nm node. It has no RT cores and no tensor cores, and its DirectX support tops out at 12 (12_1), though it matches the Intel card on OpenGL 4.6 and Vulkan 1.4.

The transistor count and density differences highlight the process advantage: Intel packs 7,200 million transistors into 157 mm², while NVIDIA fits 4,400 million into 200 mm². This density advantage (45.9M vs 22.0M per mm²) allows the Arc A350M to include modern features like dedicated ray tracing hardware and support for FP16 at a 2:1 ratio, which is 64 times faster than the P106-100’s FP16 rate (6.758 TFLOPS vs 68.36 GFLOPS). The P106-100 compensates with a wider memory bus, more shading units, and higher raw clocks, but it lacks the architectural innovations of the newer Intel design. The P106-100 is also a mining-specific product with no display outputs, whereas the Arc A350M is designed for mobile integration with display outputs dependent on the portable device. These are not merely different tiers of performance — they are different classes of hardware aimed at entirely different workloads and market segments.