Intel Arc B390 vs NVIDIA RTX PRO 6000 Blackwell Server Comparison

The Intel Arc B390 and NVIDIA RTX PRO 6000 Blackwell Server occupy opposite ends of the graphics hardware spectrum. The Arc B390 is an integrated graphics processor designed for mobile platforms, while the RTX PRO 6000 is a dual-slot server add-in card. Their benchmark scores reflect this divide, with the NVIDIA part delivering roughly four times the performance in the single recorded test. This analysis examines where each product fits, their architectural differences, and what the data indicates about their respective capabilities.

Where Each One Wins

The NVIDIA RTX PRO 6000 Blackwell Server wins the only head-to-head benchmark recorded in the database, the 3DMark Steel Nomad DX12 test. It scores 5996 points against the Intel Arc B390’s 1482 points, a delta of -75.3% from the Intel part’s perspective. This is a decisive victory across the board, with no benchmark category where the Arc B390 takes the lead.

The Arc B390’s role is not about winning performance crowns. Its specifications point to a different purpose: an integrated processor with a 3 nm process node, a 300 MHz base clock, and a 2500 MHz boost clock. It shares system memory, has 1536 shading units, 48 texture mapping units, and 24 raster output units. Its 80 W TDP and lack of power connectors confirm it is built for low-power, portable devices. The database places it at the 9th percentile among all GPUs, and its nearest rivals include the NVIDIA GeForce GT 520MX (1.3% ahead), GeForce 800M (1.5% ahead), GT 625 OEM (2.5% ahead), and GT 710 (2.7% ahead). These are all legacy entry-level discrete parts, indicating the Arc B390 competes in the basic graphics tier.

The RTX PRO 6000 Blackwell Server, by contrast, sits at the 34th percentile among all GPUs. Its nearest rivals are the GeForce GTX 770M (0.1% behind), Radeon RX 6400 (0.1% behind), FirePro W4100 (0.2% ahead), and Quadro K4000M (0.2% ahead). These deltas are all within 0.2%, meaning the RTX PRO 6000 performs in a tightly clustered band around this score. The 34th percentile is far from top-tier desktop gaming cards, but the product’s server positioning, with 96 GB of GDDR7 memory and a 512-bit bus, suggests its value lies in capacity and compute features rather than raw 3DMark numbers.

The Arc B390 wins in power efficiency. Its 80 W TDP is a fraction of the RTX PRO 6000’s 600 W. For integrated graphics in thin-and-light laptops, this is the defining metric. The RTX PRO 6000 wins in every performance metric: pixel rate, texture rate, FP32 throughput, memory bandwidth, and shading capability. The data shows two products designed for different workloads, with the NVIDIA part dominating compute-heavy tasks and the Intel part serving basic display and light graphics duties.

Architecture Differences

The Intel Arc B390 uses the Panther Lake chip with the Xe3-LPG architecture, part of the Arc Graphics-M generation. It is fabricated on Intel’s 3 nm process. The transistor count and die size are listed as unknown in the database. Its base clock is 300 MHz with a boost of 2500 MHz. The memory subsystem is entirely system shared, meaning the GPU draws on the host’s RAM with bandwidth described as system dependent. The bus interface is IGP, confirming it is integrated into the processor package. It has 12 ray tracing cores, but no tensor cores are listed. FP16 performance is 15.36 TFLOPS at a 2:1 ratio, indicating it can halve the FP32 throughput for half-precision work.

The NVIDIA RTX PRO 6000 Blackwell Server uses the GB202 chip with Blackwell 2.0 architecture. It is built on TSMC’s 5 nm process with 92,200 million transistors on a 750 mm² die, yielding a transistor density of 122.9 million per mm². The base clock is 1590 MHz with a boost of 2617 MHz. Memory is 96 GB of GDDR7 on a 512-bit bus, delivering 1.79 TB/s bandwidth. The memory clock is 1750 MHz with 28 Gbps effective data rate. It has 24064 shading units, 752 texture mapping units, 192 raster output units, 188 ray tracing cores, and 752 tensor cores. FP16 performance matches FP32 at 126.0 TFLOPS with a 1:1 ratio, meaning it does not halve throughput for half-precision. The card uses a PCIe 5.0 x16 interface, features four DisplayPort 2.1b outputs, and requires a 1000 W suggested power supply. Its predecessor is Server Hopper, and its successor is Server Rubin.

The architectural split is stark. Intel uses a leading-edge 3 nm process for integration and power savings, while NVIDIA uses a larger 5 nm process for sheer transistor count and compute density. The RTX PRO 6000 has 15.7 times more shading units, 15.7 times more texture mapping units, 8 times more raster output units, and 15.7 times more ray tracing cores. The NVIDIA part includes 752 tensor cores, which the Intel part lacks entirely. FP32 throughput is 16.4 times higher on the NVIDIA card. Memory capacity and bandwidth are not comparable, as the Intel part uses shared system memory with no dedicated specification.

The process node difference is notable. Intel’s 3 nm is smaller than TSMC’s 5 nm, but the NVIDIA chip uses that larger node to pack 92.2 billion transistors across a massive 750 mm² die. The Intel part’s transistor count is unknown, but its integrated nature and 80 W TDP indicate a much smaller design. The RTX PRO 6000’s 600 W TDP and dual-slot cooler with a 16-pin power connector are required to sustain its clocks and memory interface.

Head-to-Head Benchmarks

The database records one head-to-head benchmark: 3DMark Steel Nomad DX12. The Intel Arc B390 scores 1482, while the NVIDIA RTX PRO 6000 Blackwell Server scores 5996. The NVIDIA part wins with a delta of -75.3% relative to the Intel part. This means the Arc B390 delivers only 24.7% of the RTX PRO 6000’s score. In absolute terms, the NVIDIA card is 4514 points higher.

Interpreting this delta requires context. The RTX PRO 6000’s nearest rivals all score within 0.2% of its 5996, suggesting this benchmark places it in a specific performance band. The Arc B390’s nearest rivals score between 1443 and 1463, all within 2.7% of its 1482. The gap between the two products is roughly 4x, which aligns with the hardware specifications: the NVIDIA card has over 15 times the shading units and 16.4 times the FP32 throughput, but its memory bandwidth advantage and driver overheads compress the real-world delta to about 4x in this test.

The RTX PRO 6000’s pixel rate is 502.5 GPixel/s versus the Arc B390’s 60.00 GPixel/s, an 8.4x advantage. Texture rate is 1,968.0 GTexel/s versus 120.0 GTexel/s, a 16.4x advantage. FP32 is 126.0 TFLOPS versus 7.680 TFLOPS, also 16.4x. These theoretical rates are much larger than the benchmark delta, indicating that the 3DMark Steel Nomad test does not scale perfectly with raw compute. The 34th percentile ranking for the RTX PRO 6000 suggests that in this specific workload, it does not outperform many consumer GPUs, despite its massive memory pool. The Arc B390’s 9th percentile places it among the lowest-performing GPUs in the database.

The single benchmark limits the analysis. There are no results for ray tracing, FP16 compute, or memory-intensive workloads. The data shows a clear performance hierarchy in this one test, but no further conclusions can be drawn about other application types.

The Verdict

Based strictly on the recorded data, the NVIDIA RTX PRO 6000 Blackwell Server is the faster product. Its 5996 score in 3DMark Steel Nomad is 4.0 times the Intel Arc B390’s 1482. It also has vastly superior theoretical specifications, including 126.0 TFLOPS FP32, 96 GB GDDR7 memory, and 1.79 TB/s bandwidth. For any workload that relies on GPU compute, rendering, or large memory buffers, the RTX PRO 6000 is the clear choice.

However, the RTX PRO 6000 is a server card with a 600 W TDP, dual-slot form factor, and a 1000 W suggested power supply. It is not designed for portable devices. The Intel Arc B390 is an integrated processor with an 80 W TDP, no power connectors, and system shared memory. It fits into laptops and compact systems where the RTX PRO 6000 cannot operate at all. The Arc B390’s 9th percentile ranking and nearest rivals like the GeForce GT 710 indicate it handles basic graphics, video playback, and light 3D workloads.

The correct pick depends on the use case. For a server, workstation, or high-end desktop where space and power are available, the RTX PRO 6000’s performance and memory capacity are unmatched in this comparison. For a thin-and-light laptop or low-power embedded system, the Arc B390 is the only viable option between these two, as the RTX PRO 6000’s physical dimensions and power requirements exclude it. The data does not support a single universal winner; it supports two products serving different segments.

The RTX PRO 6000’s nearest rivals, all within 0.2% in score, include the GeForce GTX 770M and Radeon RX 6400. This places it in a mid-range performance tier despite its server branding. The Arc B390’s nearest rivals, all within 2.7%, include the GeForce GT 710 and GT 625 OEM, placing it firmly in the entry-level category. The benchmark delta of -75.3% for the Arc B390 is consistent with this class division.

FAQ

Q: Which GPU has a higher benchmark score in 3DMark Steel Nomad DX12?

A: The NVIDIA RTX PRO 6000 Blackwell Server scores 5996, while the Intel Arc B390 scores 1482. The NVIDIA part wins this test by 4514 points.

Q: What is the performance difference between the two in the recorded benchmark?

A: The RTX PRO 6000 is 75.3% ahead of the Arc B390. The Arc B390’s score is 24.7% of the NVIDIA card’s score.

Q: How much memory does each GPU have?

A: The Intel Arc B390 uses system shared memory with no dedicated size. The NVIDIA RTX PRO 6000 has 96 GB of GDDR7 memory on a 512-bit bus.

Q: What are the power requirements for each?

A: The Arc B390 has an 80 W TDP and no power connectors. The RTX PRO 6000 has a 600 W TDP, uses a 16-pin power connector, and has a suggested power supply of 1000 W.

Q: Which GPU has more shading units?

A: The RTX PRO 6000 has 24064 shading units, while the Arc B390 has 1536. The NVIDIA part has 15.7 times more.

Q: Do both GPUs support the same APIs?

A: Yes, both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Q: What process nodes are used?

A: The Intel Arc B390 uses a 3 nm process from Intel. The NVIDIA RTX PRO 6000 uses a 5 nm process from TSMC.

Specification Differences

| Specification | Intel Arc B390 | NVIDIA RTX PRO 6000 Blackwell Server |

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

| Chip | Panther Lake | GB202 |

| 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 |

| Transistor Density | Not listed | 122.9M / mm² |

| Base Clock | 300 MHz | 1590 MHz |

| Boost Clock | 2500 MHz | 2617 MHz |

| Memory Size | System Shared | 96 GB |

| Memory Type | System Shared | GDDR7 |

| Memory Bus Width | System Shared | 512 bit |

| Memory Bandwidth | System Dependent | 1.79 TB/s |

| Shading Units | 1536 | 24064 |

| TMUs | 48 | 752 |

| ROPs | 24 | 192 |

| Ray Tracing Cores | 12 | 188 |

| Tensor Cores | Not listed | 752 |

| Pixel Rate | 60.00 GPixel/s | 502.5 GPixel/s |

| Texture Rate | 120.0 GTexel/s | 1,968.0 GTexel/s |

| FP32 | 7.680 TFLOPS | 126.0 TFLOPS |

| FP16 | 15.36 TFLOPS (2:1) | 126.0 TFLOPS (1:1) |

| TDP | 80 W | 600 W |

| Slot Width | IGP | Dual-slot |

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

| Suggested PSU | Not listed | 1000 W |

| Bus Interface | IGP | PCIe 5.0 x16 |

| Display Outputs | Portable Device Dependent | 4x DisplayPort 2.1b |

| Dimensions | Not listed | 267 mm length, 111 mm height, 40 mm width |

| Release Date | 2026-01-26 | 2025-03-17 |

| Predecessor | Not listed | Server Hopper |

| Successor | Not listed | Server Rubin |

The specification table shows every point of divergence. The RTX PRO 6000 exceeds the Arc B390 in all compute metrics except process node, where Intel’s 3 nm is smaller than TSMC’s 5 nm. The Intel part is an integrated solution with no dedicated memory, while the NVIDIA part is a standalone card with 96 GB of GDDR7. The release dates differ by roughly ten months, with the RTX PRO 6000 launching on 2025-03-17 and the Arc B390 on 2026-01-26. The RTX PRO 6000 has a defined successor in Server Rubin, while the Arc B390 has neither a listed predecessor nor successor. These differences confirm that the two products target entirely different hardware segments.