Intel Arc Pro B390 vs NVIDIA GeForce RTX 5090 SE Comparison

FAQ

Q: What are the fundamental differences between the Intel Arc Pro B390 and the NVIDIA GeForce RTX 5090 SE?

A: The Arc Pro B390 is an integrated graphics processor (IGP) built on Intel's 3 nm process with the Xe3-LPG architecture, while the RTX 5090 SE is a dual-slot discrete card on TSMC's 5 nm node using the Blackwell 2.0 architecture. The Intel part is designed for portable devices with no dedicated memory, while the NVIDIA card has 24 GB of GDDR7 on a 384-bit bus.

Q: How do the shading unit counts compare between the two?

A: The RTX 5090 SE has 14,080 shading units, which is over nine times the 1,536 shading units found in the Arc Pro B390. The NVIDIA card also carries 440 texture mapping units and 160 ROPs, versus 48 TMUs and 24 ROPs on the Intel part.

Q: What is the difference in FP32 compute throughput?

A: The RTX 5090 SE delivers 66.94 TFLOPS of FP32 performance, while the Arc Pro B390 provides 7.680 TFLOPS. This represents an approximate 8.7x advantage for the NVIDIA card in raw single-precision floating-point throughput.

Q: How do the power requirements differ?

A: The Arc Pro B390 has a TDP of 80 W and uses no power connectors, as it is integrated into a processor. The RTX 5090 SE has a TDP of 500 W, requires a single 16-pin power connector, and the database lists a suggested PSU of 900 W.

Q: What memory configurations do the two cards use?

A: The Arc Pro B390 uses system-shared memory with bandwidth described as system-dependent, meaning its performance relies on the host platform's RAM. The RTX 5090 SE has 24 GB of dedicated GDDR7 memory with a 384-bit bus width and 1.34 TB/s of bandwidth.

Q: What are the release dates for both products?

A: The RTX 5090 SE was released on December 31, 2025, and the Arc Pro B390 followed on January 26, 2026. Both are currently listed as active in production.

Architecture Differences

The Arc Pro B390 uses Intel's Xe3-LPG architecture on a 3 nm process node, manufactured by Intel itself. It is part of the Arc Graphics-WM generation built around the Panther Lake chip. This is an integrated graphics solution, meaning it shares system memory and its performance scales with the host platform's RAM configuration. The architecture supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. Its predecessor is listed as HD Graphics-WM.

The RTX 5090 SE uses NVIDIA's Blackwell 2.0 architecture on a 5 nm process node from TSMC. The chip is GB202 with 92,200 million transistors on a 750 mm² die, resulting in a transistor density of 122.9 million per square millimeter. This is a discrete dual-slot card with its own 24 GB of GDDR7 memory. It uses a PCIe 5.0 x16 bus interface, whereas the Arc Pro B390 uses an IGP bus interface. The NVIDIA card also supports the same API set: DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

The Arc Pro B390's memory architecture is entirely dependent on the host system. Its memory size, type, bus width, and bandwidth are all listed as "System Shared" or "System Dependent." This stands in sharp contrast to the RTX 5090 SE's dedicated memory subsystem with 384-bit bus width and 1.34 TB/s bandwidth. The clock structure also differs: the Intel part has a base clock of 300 MHz and a boost of 2500 MHz, while the NVIDIA card has a higher base of 1740 MHz and a boost of 2377 MHz. Interestingly, the RTX 5090 SE has a lower boost clock than the Intel part, but its massive shading unit count more than compensates.

The RTX 5090 SE includes 110 ray tracing cores and 440 tensor cores, while the Arc Pro B390 has 12 ray tracing cores and no tensor cores listed. Display outputs also differ: the Intel part uses portable-device-dependent outputs, while the NVIDIA card provides 1x HDMI 2.1b and 3x DisplayPort 2.1b.

Head-to-Head Benchmarks

The database currently lists no direct head-to-head benchmark results between these two products, as both have empty benchmark arrays and zero average scores. However, the recorded specifications provide a clear comparative picture based on computational metrics.

In FP32 throughput, the RTX 5090 SE delivers 66.94 TFLOPS versus 7.680 TFLOPS for the Arc Pro B390. That is a 59.26 TFLOPS gap, meaning the NVIDIA card processes roughly 8.7 times more floating-point operations per second. This translates directly to compute-heavy workloads such as rendering, simulation, and professional 3D applications.

In FP16 performance, the NVIDIA card maintains 66.94 TFLOPS with a 1:1 ratio, while the Intel part achieves 15.36 TFLOPS with a 2:1 ratio. The RTX 5090 SE leads by 51.58 TFLOPS, but the Intel implementation's 2:1 ratio suggests it converts FP32 operations to FP16, which can be efficient for certain workloads.

Pixel throughput shows the RTX 5090 SE at 380.3 GPixel/s versus 60.00 GPixel/s for the Arc Pro B390. The NVIDIA card is approximately 6.3 times faster in pixel fill rate. Texture rate follows a similar pattern: 1,045.9 GTexel/s for the NVIDIA card versus 120.0 GTexel/s for the Intel part, an approximate 8.7x difference.

The core counts reinforce this pattern. The RTX 5090 SE has 14,080 shading units, 440 TMUs, 160 ROPs, 110 RT cores, and 440 tensor cores. The Arc Pro B390 has 1,536 shading units, 48 TMUs, 24 ROPs, and 12 RT cores. Each of these ratios falls between 6x and 9x in favor of the NVIDIA card.

Memory bandwidth is perhaps the most decisive differentiator. The RTX 5090 SE provides 1.34 TB/s of dedicated bandwidth, while the Arc Pro B390's bandwidth is system-dependent, with no fixed figure available. For memory-intensive workloads, the discrete NVIDIA card with GDDR7 memory has an inherent structural advantage that cannot be overcome by the integrated Intel solution.

Specification Differences

| Specification | Intel Arc Pro B390 | NVIDIA GeForce RTX 5090 SE |

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

| Architecture | Xe3-LPG | Blackwell 2.0 |

| Process Node | 3 nm | 5 nm |

| Foundry | Intel | TSMC |

| Transistors | Unknown | 92,200 million |

| Die Size | Unknown | 750 mm² |

| Base Clock | 300 MHz | 1740 MHz |

| Boost Clock | 2500 MHz | 2377 MHz |

| Memory Size | System Shared | 24 GB |

| Memory Type | System Shared | GDDR7 |

| Memory Bus Width | System Shared | 384 bit |

| Memory Bandwidth | System Dependent | 1.34 TB/s |

| Shading Units | 1536 | 14080 |

| TMUs | 48 | 440 |

| ROPs | 24 | 160 |

| RT Cores | 12 | 110 |

| Tensor Cores | None | 440 |

| Pixel Rate | 60.00 GPixel/s | 380.3 GPixel/s |

| Texture Rate | 120.0 GTexel/s | 1,045.9 GTexel/s |

| FP32 | 7.680 TFLOPS | 66.94 TFLOPS |

| FP16 | 15.36 TFLOPS (2:1) | 66.94 TFLOPS (1:1) |

| TDP | 80 W | 500 W |

| Slot Width | IGP | Dual-slot |

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

| Suggested PSU | None | 900 W |

| Bus Interface | IGP | PCIe 5.0 x16 |

| Display Outputs | Portable Device Dependent | 1x HDMI 2.1b, 3x DisplayPort 2.1b |

| Release Date | 2026-01-26 | 2025-12-31 |

| Launch MSRP | None | 1,499 USD |

Where Each One Wins

The Arc Pro B390 wins decisively in power efficiency and physical footprint. Its 80 W TDP is 420 W lower than the RTX 5090 SE's 500 W TDP. It requires no power connectors, no dedicated PCIe slot, and no external PSU recommendation. For ultra-portable or embedded systems where space and thermal constraints dominate, the Intel part is the only viable option between these two.

The Arc Pro B390 also has a higher boost clock at 2500 MHz versus 2377 MHz for the NVIDIA card. While this does not compensate for the massive core count difference, it does indicate the Intel part can scale up its clock frequency more aggressively within its 80 W envelope.

The RTX 5090 SE wins in every raw performance category recorded in the database. Its FP32 throughput is 8.7x higher, its pixel rate is 6.3x higher, and its texture rate is 8.7x higher. It has 110 RT cores versus 12, and 440 tensor cores versus none. Its memory system is entirely self-contained with 24 GB of GDDR7 and 1.34 TB/s of bandwidth, making it suitable for workloads that require large datasets to reside on the GPU.

The NVIDIA card also offers fixed display outputs (1x HDMI 2.1b and 3x DisplayPort 2.1b), whereas the Intel part's outputs depend on the portable device it is integrated into. For multi-monitor workstation setups, the RTX 5090 SE provides a predictable and documented output configuration.

The Verdict

The data presents a clear performance hierarchy. The RTX 5090 SE is engineered for maximum compute throughput, with 66.94 TFLOPS FP32, 1.34 TB/s memory bandwidth, and 24 GB of VRAM. It targets workloads that demand sustained high-end rendering, simulation, or AI processing, and its 500 W TDP reflects that focus.

The Arc Pro B390 is designed for integration into portable devices. Its 80 W TDP, IGP form factor, and system-shared memory indicate a low-power solution for devices where discrete graphics is not feasible. Its 7.680 TFLOPS FP32 and 60.00 GPixel/s pixel rate are modest but appropriate for an integrated part on a 3 nm process.

Users requiring maximum GPU performance should choose the RTX 5090 SE, which leads in every computational metric recorded. Users building or selecting portable devices with constrained power budgets should accept the Arc Pro B390's lower performance as the trade-off for its minimal power draw and integrated form factor. The two products occupy entirely different segments, and the benchmark data confirms they do not compete on the same playing field.