Intel Arc B390 vs NVIDIA GeForce RTX 5070 SUPER Comparison

Where Each One Wins

The benchmark data splits decisively in favor of the NVIDIA GeForce RTX 5070 SUPER. Out of the single recorded head-to-head test, 3DMark Steel Nomad DX12, the NVIDIA part wins the only contest. The Intel Arc B390 records zero wins in the database, while the RTX 5070 SUPER takes one. The score gap is substantial: the RTX 5070 SUPER posts 2690 points against 1482 for the Arc B390, a delta of 44.9 percent in NVIDIA's favor. That places the NVIDIA card in a different competitive tier entirely, as its nearest rivals (NVIDIA Quadro K1100M at 2664, GeForce GT 1030 at 2662, Intel Arc Pro B50 at 2660) all sit within a 1.7 percent range of its score. The Arc B390, by contrast, clusters with much older parts: its nearest rivals are the GeForce GT 520MX at 1463 (1.3 percent behind), GeForce 800M at 1460 (1.5 percent), GeForce GT 625 OEM at 1446 (2.5 percent), and GeForce GT 710 at 1443 (2.7 percent). The data indicates the Intel part belongs to a performance class roughly 80 percent lower in raw 3DMark output, while the NVIDIA part commands a percentile rank of 18 among all GPUs versus 9 for Intel.

Architecture Differences

The two chips share the same DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4 API support, but the underlying silicon diverges sharply. Intel builds the Arc B390 on a 3 nm process at its own foundry, using the Xe3-LPG architecture with a Panther Lake chip, part of the Arc Graphics-M (Panther Lake) generation. NVIDIA counters with a 5 nm TSMC process, Blackwell 2.0 architecture, and the GB205 chip in the GeForce 50 series. Transistor counts differ dramatically: the RTX 5070 SUPER packs 31,100 million transistors on a 263 mm² die, yielding a density of 118.3M per mm², while the Intel part's transistor count and die size are recorded as unknown. The RTX 5070 SUPER has 6400 shading units, 200 texture mapping units, 80 ROPs, 50 ray tracing cores, and 200 tensor cores. The Arc B390 fields 1536 shading units, 48 TMUs, 24 ROPs, and 12 ray tracing cores, with no tensor core count listed. Clock behavior also separates them: the Intel part runs a 300 MHz base and 2500 MHz boost, while NVIDIA runs 2325 MHz base and 2512 MHz boost. Memory architecture is the largest structural split. The Arc B390 uses system shared memory with system shared type, bus width, and system dependent bandwidth, making it an integrated graphics product. The RTX 5070 SUPER carries 18 GB of GDDR7 on a 192 bit bus with 672.0 GB/s bandwidth and a 1750 MHz memory clock (28 Gbps effective). Power delivery follows the same pattern: the Intel IGP draws 80 W with no power connectors, while the NVIDIA dual-slot card consumes 275 W through a single 16-pin connector. The Intel part is an IGP with portable device dependent display outputs, whereas NVIDIA offers 1x HDMI 2.1b and 3x DisplayPort 2.1b, PCIe 5.0 x16, and physical dimensions of 245 mm length, 115 mm height, and 40 mm width.

FAQ

Q: Which GPU has the higher 3DMark Steel Nomad DX12 score?

A: The NVIDIA GeForce RTX 5070 SUPER scores 2690, which is 44.9 percent higher than the Intel Arc B390's 1482. The NVIDIA part wins the only recorded head-to-head benchmark.

Q: How do the two compare in ray tracing hardware?

A: The RTX 5070 SUPER includes 50 ray tracing cores, while the Arc B390 has 12. The NVIDIA part also carries 200 tensor cores, a feature absent from the Intel specification.

Q: What memory configurations do the two use?

A: The Arc B390 relies on system shared memory with system dependent bandwidth, making it an integrated solution. The RTX 5070 SUPER uses 18 GB of GDDR7 on a 192 bit bus with 672.0 GB/s bandwidth.

Q: Which product has the higher thermal design power?

A: The RTX 5070 SUPER is rated at 275 W, compared to 80 W for the Arc B390. This reflects the NVIDIA card's discrete dual-slot design with a 16-pin power connector versus Intel's connector-free IGP.

Q: How do their manufacturing processes differ?

A: Intel fabricates the Arc B390 on a 3 nm process at its own foundry. NVIDIA uses TSMC's 5 nm process for the GB205 chip. The NVIDIA die measures 263 mm² with 31,100 million transistors.

Q: What is the percentile ranking of each GPU among all GPUs?

A: The RTX 5070 SUPER sits at the 18th percentile, while the Arc B390 ranks at the 9th percentile. The NVIDIA card's nearest rivals score within 1.7 percent of its average, whereas Intel's nearest rivals are within 2.7 percent.

Specification Differences

The recorded data shows broad divergence across every major specification category. Process node: Intel uses 3 nm, NVIDIA uses 5 nm. Foundry: Intel versus TSMC. Transistors: unknown for Intel, 31,100 million for NVIDIA. Die size: unknown versus 263 mm². Base clock: 300 MHz versus 2325 MHz. Boost clock: 2500 MHz versus 2512 MHz. Memory size: system shared versus 18 GB. Memory type: system shared versus GDDR7. Bus width: system shared versus 192 bit. Bandwidth: system dependent versus 672.0 GB/s. Shading units: 1536 versus 6400. TMUs: 48 versus 200. ROPs: 24 versus 80. Ray tracing cores: 12 versus 50. Tensor cores: not listed versus 200. Pixel rate: 60.00 GPixel/s versus 201.0 GPixel/s. Texture rate: 120.0 GTexel/s versus 502.4 GTexel/s. FP32: 7.680 TFLOPS versus 32.15 TFLOPS. FP16: 15.36 TFLOPS (2:1) versus 32.15 TFLOPS (1:1). TDP: 80 W versus 275 W. Slot width: IGP versus dual-slot. Power connectors: none versus 1x 16-pin. Bus interface: IGP versus PCIe 5.0 x16. Display outputs: portable device dependent versus 1x HDMI 2.1b and 3x DisplayPort 2.1b. Dimensions are absent for Intel and 245 mm by 115 mm by 40 mm for NVIDIA. Release dates differ: Intel on 2026-01-26, NVIDIA on 2025-12-31. The API sets are identical: DirectX 12 Ultimate (12_2), OpenGL 4.6, Vulkan 1.4.

Head-to-Head Benchmarks

The sole recorded comparison, 3DMark Steel Nomad DX12, delivers a clear verdict. The NVIDIA GeForce RTX 5070 SUPER scores 2690 points, while the Intel Arc B390 scores 1482. The delta of 44.9 percent means the Intel part achieves roughly 55 percent of the NVIDIA score. To put that in context, the Arc B390's result places it among legacy mobile and OEM parts: the GeForce GT 520MX trails by only 1.3 percent, the GeForce 800M by 1.5 percent, the GeForce GT 625 OEM by 2.5 percent, and the GeForce GT 710 by 2.7 percent. These are not modern gaming parts; they represent a much older performance envelope. The RTX 5070 SUPER, meanwhile, sits within a tight cluster of its own: the Quadro K1100M is 1 percent behind, the GeForce GT 1030 is 1.1 percent behind, the Intel Arc Pro B50 is 1.1 percent behind, and the GeForce GT 440 is 1.7 percent behind. The NVIDIA card's 18th percentile rank versus Intel's 9th percentile confirms the gap in absolute terms. In raw throughput metrics, the RTX 5070 SUPER delivers 32.15 TFLOPS FP32 against 7.680 TFLOPS for the Arc B390, a 4.2x advantage in compute. Texture rate favors NVIDIA at 502.4 GTexel/s versus 120.0 GTexel/s, and pixel rate favors NVIDIA at 201.0 GPixel/s versus 60.00 GPixel/s. The memory subsystem amplifies the difference: 672.0 GB/s of dedicated GDDR7 bandwidth versus system dependent shared memory, which is a fundamental constraint on the integrated Intel part. The RTX 5070 SUPER's 200 tensor cores and 50 ray tracing cores give it hardware capabilities the Arc B390 cannot match, as the Intel part lists no tensor cores and only 12 ray tracing cores.

The Verdict

The data directs each product to a distinct use case. The NVIDIA GeForce RTX 5070 SUPER is the discrete performance part: it wins the only benchmark, holds an 18th percentile rank, and delivers 32.15 TFLOPS FP32, 672.0 GB/s memory bandwidth, and 50 ray tracing cores. Its 275 W power draw and dual-slot footprint with a 16-pin connector indicate a desktop add-in card built for sustained rendering workloads. The Intel Arc B390, with its 80 W TDP, IGP slot width, no power connectors, and system shared memory, functions as an integrated graphics solution for portable devices. Its 9th percentile rank and nearest rivals from the GeForce 500M and 800M era show that its 3DMark output sits in a legacy performance class. The 44.9 percent benchmark deficit, combined with the 4.2x FP32 gap and the absence of dedicated memory, makes the RTX 5070 SUPER the only choice for demanding 3D workloads. The Arc B390 offers a low-power integrated alternative whose 3D capability, per the recorded data, does not compete with the NVIDIA part. For users constrained to an IGP, the Arc B390 provides 12 ray tracing cores, DirectX 12 Ultimate support, and Vulkan 1.4 in a 3 nm design, but the benchmark evidence places it firmly below the RTX 5070 SUPER in every measured category. The RTX 5070 SUPER, released earlier on 2025-12-31 versus the Arc B390's 2026-01-26, also carries the advantage of a mature discrete ecosystem in the form of PCIe 5.0 x16 and standard display outputs. The verdict from the database is unambiguous: the RTX 5070 SUPER wins the head-to-head, and the Arc B390 occupies a separate, lower tier of integrated graphics performance.