Intel Arc Pro B390 vs NVIDIA Rubin GPU Comparison

Head-to-Head Benchmarks

The recorded data for the Intel Arc Pro B390 and the NVIDIA Rubin GPU presents a stark contrast in compute capabilities, though direct benchmark scores are not available in the database. The most significant gap lies in raw floating-point performance. The NVIDIA Rubin GPU delivers 130.0 TFLOPS of FP32 compute, while the Intel Arc Pro B390 produces 7.680 TFLOPS. This positions the Rubin GPU roughly 17 times ahead of the Arc Pro B390 in single-precision throughput, a margin that dominates any conceivable workload.

For FP16 calculations, the disparity persists. The Rubin GPU reaches 260.0 TFLOPS (2:1), while the Arc Pro B390 achieves 15.36 TFLOPS (2:1). The Rubin part again holds a commanding lead. Texture processing tells a similar story: the Rubin GPU's texture rate of 2,031.2 GTexel/s dwarfs the Arc Pro B390's 120.0 GTexel/s. Pixel rates are closer, with the Arc Pro B390 posting 60.00 GPixel/s against the Rubin GPU's 54.41 GPixel/s, a rare instance where the Intel part edges ahead, albeit by a narrow margin of roughly 10%.

Memory bandwidth amplifies the NVIDIA advantage. The Rubin GPU provides 22.1 TB/s over a 16384-bit HBM4 interface, while the Arc Pro B390 relies on system-shared memory with bandwidth described as system dependent. The difference is not merely large; it is structurally categorical. The Rubin GPU's 288 GB of dedicated HBM4 memory versus the Arc Pro B390's system-shared allocation means memory-bound tasks will favor NVIDIA decisively.

Both parts match in pixel throughput, suggesting that the Arc Pro B390's lower compute is partly compensated by a high pixel fill rate relative to its size. However, with no benchmark scores recorded in the database, the wins column shows zero for both GPUs. The analysis must rely on architectural specifications and measured feature sets rather than application-level results.

Architecture Differences

The two GPUs target entirely different segments, and their architectures reflect that split. The Intel Arc Pro B390 uses the Panther Lake chip with Xe3-LPG architecture, built on a 3 nm process at Intel's foundry. The NVIDIA Rubin GPU uses the GR100 chip with Rubin architecture, also on a 3 nm process but fabricated by TSMC. Both nodes are identical in process size, yet the scale diverges sharply.

Transistor counts highlight this. The Rubin GPU integrates 336,000 million transistors on a 1456 mm² die, yielding a density of 230.8M per mm². The Arc Pro B390's transistor count and die size are listed as unknown, but its shading unit count tells the scale story. The Rubin GPU packs 28,672 shading units, 896 TMUs, and 896 tensor cores. The Arc Pro B390 has 1,536 shading units, 48 TMUs, and no tensor cores listed. The Rubin GPU's shading unit count is roughly 18.7 times higher.

Ray tracing hardware differs as well. The Arc Pro B390 includes 12 dedicated RT cores, while the Rubin GPU lists no RT core count in the database. The Rubin GPU instead emphasizes tensor cores, with 896 present. Both parts feature 24 ROPs, a rare point of parity.

Memory architecture is fundamentally different. The Arc Pro B390 uses system-shared memory with a system-dependent bus width and bandwidth. The Rubin GPU uses 288 GB of HBM4 on a 16384-bit bus, delivering 22.1 TB/s. Clock speeds also differ: the Arc Pro B390 runs at a base of 300 MHz and boosts to 2500 MHz, while the Rubin GPU has a base of 700 MHz and boosts to 2267 MHz. Despite the lower base clock, the Rubin GPU's massive core count and memory subsystem produce far higher aggregate throughput.

Power envelopes reflect the gulf. The Arc Pro B390 has a TDP of 80 W and requires no power connectors, using an IGP bus interface. The Rubin GPU has a TDP of 2300 W, uses an SXM Module slot width, and suggests a 2700 W PSU. The Arc Pro B390 is an integrated graphics processor, while the Rubin GPU is a server module with PCIe 6.0 x16 connectivity.

API support also separates them. The Arc Pro B390 supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The Rubin GPU lists N/A for DirectX, OpenGL, and Vulkan, indicating it is not designed for conventional graphics APIs. Display outputs follow suit: the Arc Pro B390 provides portable-device-dependent outputs, while the Rubin GPU has no outputs at all.

Where Each One Wins

The Intel Arc Pro B390 wins in scenarios that demand integrated, low-power graphics with modern API support. Its 80 W TDP, IGP form factor, and DirectX 12 Ultimate capability make it suitable for portable devices and compact systems where discrete GPUs are impractical. The 12 RT cores provide hardware ray tracing, and the 60.00 GPixel/s pixel rate is competitive for its class. The 2500 MHz boost clock, higher than the Rubin GPU's 2267 MHz, suggests strong single-core throughput for latency-sensitive tasks. Vulkan 1.4 and OpenGL 4.6 support also cover a broad range of legacy and cross-platform graphics workloads.

The NVIDIA Rubin GPU wins in compute-heavy, memory-intensive, and server-oriented tasks. Its 130.0 TFLOPS FP32 and 260.0 TFLOPS FP16 performance, combined with 22.1 TB/s bandwidth, positions it for AI training, scientific simulation, and large-scale data processing. The 896 tensor cores accelerate matrix operations, and the 288 GB HBM4 capacity handles datasets far beyond what system-shared memory can manage. The 2,031.2 GTexel/s texture rate also serves texture-heavy rendering pipelines, though the lack of display outputs and standard graphics APIs restricts it to headless compute environments.

The pixel rate advantage for the Arc Pro B390, 60.00 versus 54.41 GPixel/s, indicates that for pure rasterization fill-rate-bound tasks, the Intel part can outperform in relative terms. However, the Rubin GPU's superior texture rate and compute resources dwarf this edge in any realistic workload. The Arc Pro B390's system-shared memory means bandwidth scales with the host platform, which can be a bottleneck or a benefit depending on the system, but it cannot match dedicated HBM4.

The production status for both is Active, and release dates are close: the Rubin GPU on 2025-12-31 and the Arc Pro B390 on 2026-01-26. The Rubin GPU's predecessor is Server Blackwell, while the Arc Pro B390's predecessor is HD Graphics-WM. Neither has a successor listed.

FAQ

Q: Which GPU has higher FP32 compute performance?

A: The NVIDIA Rubin GPU delivers 130.0 TFLOPS, while the Intel Arc Pro B390 provides 7.680 TFLOPS. The Rubin GPU is approximately 17 times faster in single-precision throughput.

Q: Do both GPUs use the same manufacturing process?

A: Yes, both use a 3 nm process. Intel fabricates the Arc Pro B390 at its own foundry, while TSMC fabricates the Rubin GPU.

Q: What memory configurations do they use?

A: The Rubin GPU has 288 GB of HBM4 with a 16384-bit bus and 22.1 TB/s bandwidth. The Arc Pro B390 uses system-shared memory with system-dependent bandwidth and bus width.

Q: Which GPU supports DirectX 12 Ultimate?

A: Only the Intel Arc Pro B390 supports DirectX 12 Ultimate (12_2). The NVIDIA Rubin GPU lists N/A for DirectX, OpenGL, and Vulkan.

Q: How do their power requirements compare?

A: The Arc Pro B390 has an 80 W TDP with no power connectors. The Rubin GPU has a 2300 W TDP and suggests a 2700 W PSU.

Q: Does either GPU have tensor cores?

A: The NVIDIA Rubin GPU has 896 tensor cores. The Intel Arc Pro B390 lists no tensor core count.

The Verdict

The data points to two devices with no meaningful overlap in purpose. The Intel Arc Pro B390 is an integrated graphics processor with a 50th percentile ranking among all GPUs, built for portable systems that need modern API support, ray tracing, and modest power consumption. Its 80 W TDP, IGP slot width, and system-shared memory make it a fit for lightweight laptops and compact devices. The 12 RT cores and DirectX 12 Ultimate support provide contemporary features, while the 300 MHz base and 2500 MHz boost clocks allow for power scaling.

The NVIDIA Rubin GPU is a server-class compute module. Its 2300 W TDP, SXM Module form factor, and 2700 W suggested PSU place it in data center racks, not consumer desktops. The 130.0 TFLOPS FP32, 260.0 TFLOPS FP16, and 896 tensor cores target AI and high-performance computing workloads. The 288 GB HBM4 with 22.1 TB/s bandwidth supports massive parallel data access, and the absence of display outputs or standard graphics APIs confirms its compute-only role.

The pixel rate comparison, 60.00 GPixel/s for Intel versus 54.41 GPixel/s for NVIDIA, is the only specification where the Arc Pro B390 leads. This does not offset the Rubin GPU's 18.7 times more shading units, 18.7 times higher texture rate, or its overwhelming memory bandwidth advantage. The Rubin GPU's transistor count of 336,000 million on a 1456 mm² die versus the unknown but clearly smaller Intel part further underscores the scale difference.

Users requiring integrated graphics with API compatibility should select the Intel Arc Pro B390. Users needing maximum compute throughput for server-side tasks should select the NVIDIA Rubin GPU. The database records no benchmark scores for either, so the verdict rests on architectural specifications alone. The Rubin GPU's raw numbers place it in a performance class the Arc Pro B390 cannot approach, while the Arc Pro B390's integrated nature and software support make it the only viable option for its target devices. There is no overlap in use cases, and the data supports that separation.