Intel Arc Pro B60 Dual vs NVIDIA RTX 3500 Embedded Ada Generation Comparison

Where Each One Wins

The Intel Arc Pro B60 Dual and the NVIDIA RTX 3500 Embedded Ada Generation serve fundamentally different deployment scenarios, and the recorded data reflects that split. The Intel card is a standalone, dual-slot add-in board with four display outputs, built for desktop professional workstations where rendering and compute tasks run at high power. The NVIDIA part is an embedded, integrated GPU package (IGP) with no display outputs, designed to be soldered into mobile or compact systems where space and thermal envelopes are tightly constrained.

In terms of raw compute throughput, the NVIDIA RTX 3500 Embedded Ada Generation holds a decisive advantage. Its FP32 output of 23.04 TFLOPS is nearly double the 12.29 TFLOPS of the Intel Arc Pro B60 Dual. That advantage extends to FP16 as well: the NVIDIA part sustains 23.04 TFLOPS at a 1:1 ratio, while the Intel card reaches 24.58 TFLOPS only via a 2:1 rate, meaning it halves its throughput when precision is maintained. For workloads that rely on FP32, such as simulation, physics, or general compute, the NVIDIA part is the clear winner.

The Intel Arc Pro B60 Dual, however, wins on memory capacity and display flexibility. It carries 24 GB of GDDR6 memory versus the NVIDIA part's 12 GB, a 2x difference that matters for large datasets, high-resolution textures, or multi-model inference scenarios where memory footprint exceeds 12 GB. The Intel card also provides four mini-DisplayPort 2.1 outputs, enabling multi-monitor professional setups, while the NVIDIA embedded part offers no outputs at all, relying on the host system to handle display signal routing.

The power envelope tells a stark story. The Intel Arc Pro B60 Dual draws up to 400 W and requires an 800 W system power supply, along with a single 16-pin power connector. The NVIDIA RTX 3500 Embedded Ada Generation operates at just 100 W, needs no external power connectors, and suggests a 300 W power supply. For embedded or mobile systems where power delivery and cooling are limited, the NVIDIA part is the only viable option; for a desktop workstation with robust power delivery, the Intel card's higher consumption is acceptable.

The Intel card also leads in memory bandwidth, delivering 456.0 GB/s versus the NVIDIA part's 432.0 GB/s, a modest 5.6% margin. Pixel fill rate favors Intel at 192.0 GPixel/s versus NVIDIA's 144.0 GPixel/s, a 33.3% advantage. Texture fill rate is close, with Intel at 384.0 GTexel/s and NVIDIA at 360.0 GTexel/s, a 6.7% edge for Intel. These results indicate that the Intel card is optimized for bandwidth-heavy and rasterization-heavy tasks, while the NVIDIA part is built for compute throughput per watt.

Architecture Differences

The two GPUs stem from different architectural lineages and are fabricated with different transistor budgets. The Intel Arc Pro B60 Dual uses the BMG-G21 chip based on the Xe2-HPG architecture, part of the Battlemage (Pro Series) generation. It is built on a 5 nm process at TSMC, with 19,600 million transistors on a 272 mm² die, yielding a transistor density of 72.1 million transistors per mm².

The NVIDIA RTX 3500 Embedded Ada Generation uses the AD104 chip based on the Ada Lovelace architecture, from the Ada-MW generation. It is also fabricated on a 5 nm process at TSMC, but packs 35,800 million transistors on a 294 mm² die, achieving a transistor density of 121.8 million transistors per mm². That is a 82.7% higher transistor density than the Intel part, reflecting a more compact logic design and a much larger absolute transistor count (35,800 million versus 19,600 million, a 82.7% difference).

Core configuration differs substantially. The Intel card has 2,560 shading units, 160 texture mapping units, 80 ROPs, and 20 ray tracing cores. The NVIDIA part has 5,120 shading units (exactly double), 160 TMUs, 64 ROPs, 40 RT cores, and 160 tensor cores. The NVIDIA part's shading unit count is double that of Intel, and its RT core count is also double. The Intel card has no tensor cores listed, meaning it lacks the dedicated AI acceleration hardware that NVIDIA provides.

Clock behavior also differs. The Intel card runs a base clock of 2000 MHz and boosts to 2400 MHz. The NVIDIA part has a lower base clock of 1725 MHz and a boost of 2250 MHz. Despite the lower clocks, NVIDIA achieves higher FP32 throughput because of its larger shader count.

Memory architecture is similar in bus width (192 bit for both) but differs in capacity and speed. Intel uses 24 GB of GDDR6 at 19 Gbps effective, while NVIDIA uses 12 GB of GDDR6 at 18 Gbps effective. Both run at similar memory clock rates (2375 MHz versus 2250 MHz), but Intel's larger capacity and slightly higher speed give it the bandwidth edge.

Interconnect and power delivery differ as well. The Intel card uses PCIe 5.0 x8, while the NVIDIA part uses PCIe 4.0 x16. The Intel card is a dual-slot board measuring 300 mm in length, 110 mm in height, and 40 mm in width, requiring a 16-pin power connector. The NVIDIA part is an IGP with no listed dimensions, no power connectors, and no display outputs. The Intel card supports four mini-DisplayPort 2.1 outputs; the NVIDIA part supports none.

Head-to-Head Benchmarks

The recorded data contains no head-to-head benchmark entries, so the comparison rests on the specification-level figures. The largest single-number win for the NVIDIA RTX 3500 Embedded Ada Generation is in FP32 compute: 23.04 TFLOPS versus 12.29 TFLOPS, a 87.4% advantage. In FP16, NVIDIA delivers 23.04 TFLOPS at a 1:1 ratio, while Intel reaches 24.58 TFLOPS only at a 2:1 ratio. When both are measured at full precision, NVIDIA's FP16 output is effectively comparable, but its FP32 lead is decisive for most professional workloads.

The Intel Arc Pro B60 Dual counters with a 100% memory capacity advantage: 24 GB versus 12 GB. It also leads in pixel fill rate by 33.3% (192.0 GPixel/s versus 144.0 GPixel/s) and in texture fill rate by 6.7% (384.0 GTexel/s versus 360.0 GTexel/s). Memory bandwidth favors Intel by 5.6% (456.0 GB/s versus 432.0 GB/s).

The transistor count difference is stark: NVIDIA packs 35,800 million transistors versus Intel's 19,600 million, an 82.7% higher count. This explains NVIDIA's superior shader count (5,120 versus 2,560) and RT core count (40 versus 20). The NVIDIA part also includes 160 tensor cores, which the Intel card lacks entirely.

Power consumption is the most extreme differentiator. The Intel card is rated at 400 W TDP versus NVIDIA's 100 W TDP, a 4x difference. The suggested power supply scales accordingly: 800 W for Intel versus 300 W for NVIDIA. The Intel card's higher power budget enables its higher clock speeds (2000 MHz base, 2400 MHz boost) and its larger memory configuration, but it also restricts deployment to systems with robust power delivery.

Release timing also differs: the Intel Arc Pro B60 Dual launched on September 4, 2025, while the NVIDIA RTX 3500 Embedded Ada Generation launched on March 20, 2023. The NVIDIA part has a predecessor (Ampere-MW) and a successor (Blackwell-MW), while the Intel card has neither listed.

FAQ

Q: Which GPU has higher FP32 compute performance?

A: The NVIDIA RTX 3500 Embedded Ada Generation delivers 23.04 TFLOPS FP32, which is 87.4% higher than the Intel Arc Pro B60 Dual's 12.29 TFLOPS.

Q: How much memory does each card have?

A: The Intel Arc Pro B60 Dual has 24 GB of GDDR6, while the NVIDIA RTX 3500 Embedded Ada Generation has 12 GB of GDDR6. Intel's capacity is double that of NVIDIA's.

Q: Can the NVIDIA RTX 3500 Embedded Ada Generation drive displays?

A: No. The NVIDIA part has no display outputs. The Intel Arc Pro B60 Dual provides four mini-DisplayPort 2.1 outputs.

Q: What is the power draw difference?

A: The Intel Arc Pro B60 Dual has a 400 W TDP and requires an 800 W power supply, while the NVIDIA RTX 3500 Embedded Ada Generation has a 100 W TDP and suggests a 300 W power supply.

Q: Does the NVIDIA part have tensor cores?

A: Yes, the NVIDIA RTX 3500 Embedded Ada Generation includes 160 tensor cores. The Intel Arc Pro B60 Dual has no tensor cores listed.

Q: Which GPU has more ray tracing cores?

A: The NVIDIA RTX 3500 Embedded Ada Generation has 40 RT cores, double the 20 RT cores of the Intel Arc Pro B60 Dual.

Specification Differences

| Specification | Intel Arc Pro B60 Dual | NVIDIA RTX 3500 Embedded Ada Generation |

|---|---|---|

| Chip | BMG-G21 | AD104 |

| Architecture | Xe2-HPG | Ada Lovelace |

| Generation | Battlemage (Pro Series) | Ada-MW |

| Transistors | 19,600 million | 35,800 million |

| Die Size | 272 mm² | 294 mm² |

| Transistor Density | 72.1M / mm² | 121.8M / mm² |

| Base Clock | 2000 MHz | 1725 MHz |

| Boost Clock | 2400 MHz | 2250 MHz |

| Memory Clock | 2375 MHz, 19 Gbps effective | 2250 MHz, 18 Gbps effective |

| Memory Size | 24 GB | 12 GB |

| Memory Type | GDDR6 | GDDR6 |

| Memory Bus Width | 192 bit | 192 bit |

| Memory Bandwidth | 456.0 GB/s | 432.0 GB/s |

| Shading Units | 2560 | 5120 |

| TMUs | 160 | 160 |

| ROPs | 80 | 64 |

| RT Cores | 20 | 40 |

| Tensor Cores | None | 160 |

| Pixel Rate | 192.0 GPixel/s | 144.0 GPixel/s |

| Texture Rate | 384.0 GTexel/s | 360.0 GTexel/s |

| FP32 | 12.29 TFLOPS | 23.04 TFLOPS |

| FP16 | 24.58 TFLOPS (2:1) | 23.04 TFLOPS (1:1) |

| TDP | 400 W | 100 W |

| Slot Width | Dual-slot | IGP |

| Power Connectors | 1x 16-pin | None |

| Suggested PSU | 800 W | 300 W |

| Bus Interface | PCIe 5.0 x8 | PCIe 4.0 x16 |

| Display Outputs | 4x mini-DisplayPort 2.1 | No outputs |

| Release Date | 2025-09-04 | 2023-03-20 |

| Predecessor | None | Ampere-MW |

| Successor | None | Blackwell-MW |

| Launch MSRP | 1,199 USD | None |