Intel Arc B390 vs NVIDIA GeForce RTX 5070 Comparison

Head-to-Head Benchmarks

The database records a single shared benchmark between these two parts: 3DMark Steel Nomad DX12. The result is decisive. The NVIDIA GeForce RTX 5070 scores 5077 points, while the Intel Arc B390 manages 1482 points. That puts the NVIDIA part ahead by 70.8 percent, a gap that is not close enough to be considered competitive.

For context, the Intel Arc B390 sits at the 9th percentile of all GPUs in the database. Its nearest rivals are legacy low-end parts: the NVIDIA GeForce GT 520MX at 1463 points (1.3 percent behind), the GeForce 800M at 1460 points (1.5 percent behind), the GeForce GT 625 OEM at 1446 points (2.5 percent behind), and the GeForce GT 710 at 1443 points (2.7 percent behind). The B390 leads that group, but the delta is small, between 1.3 and 2.7 percent. These are not meaningful performance separations; they are effectively the same class of result.

The RTX 5070, by contrast, sits at the 82nd percentile of all GPUs. Its nearest rivals in the database are professional and mobile parts: the AMD Radeon Pro 580 at 40318 points (0.1 percent behind), the AMD Radeon Pro WX 7100 at 40063 points (0.8 percent behind), the AMD Radeon Pro 5300 at 40870 points (1.2 percent ahead of the 5070), and the NVIDIA RTX A500 Mobile at 39568 points (2 percent behind). The RTX 5070 essentially trades blows with those parts, but they are all far above the B390's result.

The single head-to-head metric tells the full story: the RTX 5070 is roughly 3.4 times faster in this DX12 workload. No other common benchmark exists in the database, so the comparison rests on this one measurement. That is sufficient to establish the overall performance hierarchy, but it also means the data does not support detailed claims about specific workloads beyond Steel Nomad.

Architecture Differences

The two GPUs come from different design philosophies and different manufacturing ecosystems. The Intel Arc B390 uses the Panther Lake chip with the Xe3-LPG architecture, built on a 3 nm process at Intel's own foundry. The NVIDIA GeForce RTX 5070 uses the GB205 chip with the Blackwell 2.0 architecture, built on a 5 nm process at TSMC. Intel's process node is denser on paper, but the RTX 5070 compensates with raw silicon scale.

The RTX 5070 has 31,100 million transistors on a 263 mm² die, giving a transistor density of 118.3 million per mm². The B390's transistor count and die size are not recorded in the database, so a direct density comparison is not possible. What is known is that the RTX 5070 is a large discrete GPU, while the B390 is an integrated graphics processor.

Clock behavior differs substantially. The B390 runs a base clock of 300 MHz and boosts to 2500 MHz. The RTX 5070 runs a base clock of 2325 MHz and boosts to 2512 MHz. The boost clocks are close, but the B390 starts far lower, which reflects its integrated design and power envelope.

Memory architecture is a fundamental split. The B390 uses system shared memory with a system dependent bandwidth, meaning its memory performance scales with the host platform. The RTX 5070 has 12 GB of GDDR7 on a 192 bit bus with 672.0 GB/s of dedicated bandwidth. The memory clock is 1750 MHz, 28 Gbps effective. The B390's memory clock is likewise system shared and system dependent.

The compute resources are heavily skewed toward NVIDIA. The RTX 5070 has 6144 shading units, 192 texture mapping units, 80 raster output units, 48 ray tracing cores, and 192 tensor cores. The B390 has 1536 shading units, 48 TMUs, 24 ROPs, and 12 ray tracing cores, with no tensor core count recorded. Pixel rate for the RTX 5070 is 201.0 GPixel/s versus 60.00 GPixel/s for the B390. Texture rate is 482.3 GTexel/s versus 120.0 GTexel/s. FP32 throughput is 30.87 TFLOPS versus 7.680 TFLOPS. FP16 throughput is 30.87 TFLOPS at 1:1 ratio for NVIDIA, versus 15.36 TFLOPS at 2:1 for Intel.

Power and physical design are also distinct. The B390 has an 80 W TDP, uses no power connectors, and occupies an IGP slot. The RTX 5070 has a 250 W TDP, requires a 16-pin connector, and is a dual-slot card measuring 245 mm in length, 115 mm in height, and 40 mm in width. The system power supply recommendation for the RTX 5070 is 600 W; no such recommendation exists for the B390.

API support is identical on paper: both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The bus interface differs: the B390 is IGP, while the RTX 5070 uses PCIe 5.0 x16.

Where Each One Wins

The RTX 5070 wins the only recorded benchmark outright. In the 3DMark Steel Nomad DX12 test, it scores 5077 against the B390's 1482, a 70.8 percent advantage. That is the sole head-to-head data point, so the use-case split must be inferred from the architecture fields.

For gaming and high-resolution rendering, the RTX 5070 is the clear choice based on the benchmark data. Its dedicated 12 GB GDDR7 memory with 672.0 GB/s bandwidth, its 48 ray tracing cores, and its 30.87 TFLOPS of FP32 throughput all point to a part designed for demanding workloads. The 82nd percentile ranking confirms it sits well above the median GPU in the database.

The B390, with its 80 W TDP, no power connectors, and IGP form factor, is not a discrete gaming solution. Its 9th percentile ranking places it among the weakest GPUs in the database, alongside decade-old entry-level parts. Its advantages are practical rather than performance-based: it needs no separate power delivery, no PCIe slot, and no additional cooling beyond what the host system provides. For a portable device or a lightweight system where space and power are constrained, the B390 is usable for basic display output and light acceleration.

The RTX 5070 also holds the advantage in compute throughput. Its FP32 and FP16 figures are identical at 30.87 TFLOPS, indicating full-rate FP16 processing. The B390's FP16 is half-rate at 15.36 TFLOPS relative to its 7.680 FP32. Tensor cores on the RTX 5070 (192 of them) add AI and deep learning capability that the B390 cannot match, since its tensor core count is not recorded.

There is no recorded scenario where the B390 wins over the RTX 5070. The data shows zero wins for the Intel part and one win for the NVIDIA part.

FAQ

Q: How much faster is the RTX 5070 than the Arc B390 in 3DMark Steel Nomad?

A: The RTX 5070 scores 5077 points versus 1482 for the Arc B390, a 70.8 percent advantage.

Q: What is the Arc B390's performance tier in the database?

A: The Arc B390 sits at the 9th percentile of all GPUs, with nearest rivals including the GeForce GT 520MX (1.3 percent behind) and the GeForce 800M (1.5 percent behind).

Q: What is the RTX 5070's performance tier in the database?

A: The RTX 5070 sits at the 82nd percentile of all GPUs, with nearest rivals including the AMD Radeon Pro 580 (0.1 percent behind) and the AMD Radeon Pro WX 7100 (0.8 percent behind).

Q: Does the Arc B390 have dedicated video memory?

A: No, the Arc B390 uses system shared memory with a system dependent bandwidth. The RTX 5070 has 12 GB of GDDR7 with 672.0 GB/s bandwidth.

Q: What power delivery does each GPU require?

A: The Arc B390 has an 80 W TDP and uses no power connectors. The RTX 5070 has a 250 W TDP, uses a single 16-pin connector, and requires a 600 W power supply.

Q: Do both GPUs support the same graphics APIs?

A: Yes, both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, but the RTX 5070 also has 192 tensor cores while the Arc B390 has none recorded.

Specification Differences

| Specification | Intel Arc B390 | NVIDIA GeForce RTX 5070 |

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

| Architecture | Xe3-LPG | Blackwell 2.0 |

| Process Node | 3 nm | 5 nm |

| Foundry | Intel | TSMC |

| Transistors | Unknown | 31,100 million |

| Die Size | Unknown | 263 mm² |

| Transistor Density | Not recorded | 118.3M / mm² |

| Base Clock | 300 MHz | 2325 MHz |

| Boost Clock | 2500 MHz | 2512 MHz |

| Memory Size | System Shared | 12 GB |

| Memory Type | System Shared | GDDR7 |

| Memory Bus Width | System Shared | 192 bit |

| Memory Bandwidth | System Dependent | 672.0 GB/s |

| Shading Units | 1536 | 6144 |

| TMUs | 48 | 192 |

| ROPs | 24 | 80 |

| RT Cores | 12 | 48 |

| Tensor Cores | Not recorded | 192 |

| Pixel Rate | 60.00 GPixel/s | 201.0 GPixel/s |

| Texture Rate | 120.0 GTexel/s | 482.3 GTexel/s |

| FP32 | 7.680 TFLOPS | 30.87 TFLOPS |

| FP16 | 15.36 TFLOPS (2:1) | 30.87 TFLOPS (1:1) |

| TDP | 80 W | 250 W |

| Slot Width | IGP | Dual-slot |

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

| Suggested PSU | Not recorded | 600 W |

| Bus Interface | IGP | PCIe 5.0 x16 |

| Dimensions | Not recorded | 245 mm x 115 mm x 40 mm |

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

| Launch MSRP | Not recorded | 549 USD |

The Verdict

The data supports a clear conclusion. The NVIDIA GeForce RTX 5070 is a discrete, high-performance GPU that ranks at the 82nd percentile of all GPUs and scores 5077 in 3DMark Steel Nomad. The Intel Arc B390 is an integrated graphics processor that ranks at the 9th percentile and scores 1482 in the same test. The 70.8 percent gap is the only head-to-head measurement available, and it is decisive.

For anyone building a system that needs serious gaming, ray tracing, or compute performance, the RTX 5070 is the only viable option between these two. It has dedicated GDDR7 memory, 48 RT cores, 192 tensor cores, and roughly 4 times the shading units and FP32 throughput of the B390.

For a portable device or a system where a discrete GPU cannot be installed, the B390 provides basic graphics capability with no power connectors and an 80 W TDP. Its performance is comparable to decade-old entry-level discrete GPUs like the GeForce GT 710, which is adequate for display output and light workloads but not for modern gaming.

The release dates are also relevant. The RTX 5070 launched on 2025-03-03 with a launch MSRP of 549 USD. The B390 launched on 2026-01-26 with no recorded MSRP. The RTX 5070's predecessor is the GeForce 40 series and its successor is the GeForce 60 series, placing it within an established product line. The B390 has no recorded predecessor or successor.

The verdict is straightforward: the RTX 5070 wins on every measured performance axis, while the B390 wins only on integration efficiency and power draw. Choose the RTX 5070 for performance, choose the B390 only if the system physically cannot accommodate a discrete card.