Intel Arc B580 vs NVIDIA P106-100 Comparison

The NVIDIA P106-100 and Intel Arc B580 occupy very different positions in the GPU landscape, separated by seven years of architecture evolution and a fundamental shift in design philosophy. The P106-100 is a mining-era product stripped of display outputs, while the Arc B580 is a full-featured consumer graphics card built for modern gaming and compute. Benchmark data shows a decisive overall victory for the Intel part, which wins all three head-to-head tests by substantial margins, yet the P106-100 remains competitive in aggregate scoring due to its specific workload characteristics. The data reveals that the Arc B580 delivers roughly triple the raw performance in modern API tests, but the comparison is not merely about speed — it is about capability, efficiency, and the architectural gulf between 2017 and 2024 designs.

Where Each One Wins

The Intel Arc B580 wins every benchmark category where both products have recorded scores. In the 3DMark Steel Nomad DX12 test, the Arc B580 scores 3068 against the P106-100's 899, a delta of -70.7 percent in favor of Intel. This is a modern, DirectX 12 workload that stresses contemporary GPU features, and the gap is enormous. The Arc B580 also dominates compute-oriented tests: Geekbench OpenCL shows a score of 92821 versus 35951 (a -61.3 percent delta for the P106-100), and Geekbench Vulkan shows 109672 versus 32897 (a -70 percent delta). These are not marginal improvements; they represent a generational leap in throughput.

The P106-100's only area of relative strength is its aggregate benchmark average, which stands at 23249, marginally higher than the Arc B580's 23021. This is a statistical artifact of the different benchmark suites available for each card — the P106-100 has fewer recorded tests, and its scores in Geekbench OpenCL and Vulkan, while lower in absolute terms, are not as far behind as the 3DMark result suggests. In terms of direct head-to-head competition, the Arc B580 is the clear winner in every measurable category, but the P106-100's legacy drivers and specific compute behavior keep its average competitive. The Arc B580 also holds a percentile rank of 68, identical to the P106-100's percentile of 68, indicating that both cards sit in a similar overall performance tier when compared against all GPUs, despite the newer card's dominance in these specific tests.

Architecture Differences

The architectural chasm between these two GPUs is vast, beginning with the manufacturing process. The P106-100 uses a 16 nm TSMC node with 4,400 million transistors on a 200 mm² die, yielding a transistor density of 22.0M per mm². The Arc B580 uses a 5 nm TSMC node with 19,600 million transistors on a 272 mm² die, achieving a density of 72.1M per mm² — more than three times the density. This alone explains much of the performance difference, as the Intel chip packs over four times the transistor count into a die that is only 36 percent larger.

The P106-100 is built on the Pascal architecture and uses the GP106 chip, designed for mining operations with no display outputs. It features 1280 shading units, 80 texture mapping units, and 48 ROPs, with no dedicated ray tracing or tensor cores. The Arc B580, by contrast, uses the Xe2-HPG architecture on the BMG-G21 chip, part of the Battlemage generation. It has 2560 shading units, 160 TMUs, and 80 ROPs — exactly double the P106-100 in each category — plus 20 dedicated ray tracing cores. The Intel card also supports DirectX 12 Ultimate (12_2), while the P106-100 is limited to DirectX 12 (12_1), a key differentiator for modern game features.

Memory architecture follows the same pattern of generational advancement. The P106-100 uses 6 GB of GDDR5 on a 192-bit bus, delivering 192.2 GB/s of bandwidth. The Arc B580 doubles the capacity to 12 GB of GDDR6 on the same 192-bit bus, but memory clocks of 2375 MHz (19 Gbps effective) push bandwidth to 456.0 GB/s — more than double. The Intel card also supports PCIe 4.0 x8, whereas the P106-100 is limited to PCIe 1.0 x16, a legacy interface that could bottleneck data transfer in certain workloads.

Head-to-Head Benchmarks

The 3DMark Steel Nomad DX12 test illustrates the most dramatic performance gap. The Arc B580 scores 3068, while the P106-100 manages only 899, a 70.7 percent deficit for the older card. This test uses modern DirectX 12 features that the Pascal architecture cannot efficiently accelerate, and the result is a crushing loss for the NVIDIA part. The Arc B580's 20 ray tracing cores and DirectX 12 Ultimate feature set enable workloads that the P106-100 simply cannot process at comparable speed.

Geekbench OpenCL results show a similar but slightly narrower gap. The Arc B580 scores 92821, versus 35951 for the P106-100, a 61.3 percent difference. OpenCL is a compute-heavy benchmark that scales with raw shader throughput, and the Arc B580's 2560 shading units at 2670 MHz provide 13.67 TFLOPS of FP32 performance, compared to the P106-100's 1280 shading units at 1709 MHz boost, which yield 4.375 TFLOPS. This is a 3.1x advantage in raw compute, and the benchmark reflects it.

Geekbench Vulkan shows the largest absolute margin. The Arc B580 scores 109672, while the P106-100 scores 32897 — a 70 percent delta. Vulkan is a low-level API that rewards modern architecture features and efficient driver overhead, and the Intel Xe2-HPG design excels here. The P106-100's Pascal architecture, while capable in its era, cannot match the newer design's scheduling and memory management efficiency. Across all three tests, the Arc B580's smallest win is 61.3 percent, and its largest is 70.7 percent, making the overall verdict unambiguous.

Specification Differences

The two cards differ in nearly every measurable specification. Process node advances from 16 nm to 5 nm, with transistor count jumping from 4,400 million to 19,600 million. Die size increases from 200 mm² to 272 mm², while transistor density triples from 22.0M to 72.1M per mm². Base and boost clocks rise from 1506/1709 MHz to a flat 2670 MHz for both on the Intel card. Memory capacity doubles from 6 GB GDDR5 to 12 GB GDDR6, while bandwidth more than doubles from 192.2 GB/s to 456.0 GB/s, despite both using a 192-bit bus.

Compute resources scale precisely: shading units, TMUs, and ROPs all double from 1280/80/48 to 2560/160/80. The Arc B580 adds 20 ray tracing cores, a feature entirely absent from the P106-100. Pixel rate increases from 82.03 GPixel/s to 213.6 GPixel/s, and texture rate from 136.7 GTexel/s to 427.2 GTexel/s. FP32 performance rises from 4.375 to 13.67 TFLOPS, and FP16 performance jumps from 68.36 GFLOPS (at a 1:64 ratio) to 27.34 TFLOPS (at a 2:1 ratio) — a 400x improvement in half-precision throughput.

Power and interface specifications also diverge significantly. The P106-100 draws 120 W with a 300 W suggested PSU and a single 6-pin connector, while the Arc B580 draws 190 W with a 450 W suggested PSU and a single 8-pin connector. Both are dual-slot designs, but the Intel card is longer at 272 mm (10.7 inches) versus 250 mm (9.8 inches), and adds height and width dimensions of 115 mm and 45 mm, respectively. The P106-100 has no display outputs, while the Arc B580 offers 1x HDMI 2.1a and 3x DisplayPort 2.1. The P106-100 uses PCIe 1.0 x16, while the Arc B580 uses PCIe 4.0 x8. Production status also differs: the P106-100 is end-of-life, while the Arc B580 is active, with a launch MSRP of 249 USD.

FAQ

Q: Which card is faster in DirectX 12 gaming workloads?

A: The Intel Arc B580 wins decisively. In the 3DMark Steel Nomad DX12 test, it scores 3068 against the P106-100's 899, a 70.7 percent advantage. The Arc B580 also supports DirectX 12 Ultimate (12_2), while the P106-100 only reaches DirectX 12 (12_1).

Q: How do the two cards compare in compute performance?

A: The Arc B580 leads substantially. Geekbench OpenCL shows 92821 versus 35951 (61.3 percent faster), and Geekbench Vulkan shows 109672 versus 32897 (70 percent faster). The Arc B580's FP32 throughput of 13.67 TFLOPS is roughly three times the P106-100's 4.375 TFLOPS.

Q: Does the P106-100 have any advantage in benchmark averages?

A: In aggregate, the P106-100 has a slightly higher average benchmark score of 23249 versus the Arc B580's 23021. However, this reflects a limited test suite for the NVIDIA card; in every direct head-to-head comparison, the Arc B580 wins by at least 61 percent.

Q: What are the memory capacity and bandwidth differences?

A: The P106-100 has 6 GB of GDDR5 on a 192-bit bus, providing 192.2 GB/s of bandwidth. The Arc B580 has 12 GB of GDDR6 on the same bus width, but memory clocks of 2375 MHz (19 Gbps effective) yield 456.0 GB/s — more than double the bandwidth.

Q: Are both cards still in production?

A: No. The P106-100 is end-of-life, released in June 2017, while the Arc B580 is active, released in December 2024. The Intel card also has a successor lineage, with the Alchemist architecture as its predecessor.

Q: Which card has better API support for modern features?

A: The Arc B580 supports DirectX 12 Ultimate (12_2), which includes ray tracing via its 20 dedicated RT cores. The P106-100 supports DirectX 12 (12_1) but has no ray tracing hardware. Both cards support OpenGL 4.6 and Vulkan 1.4, though the Arc B580's Vulkan scores are 70 percent higher.