Intel Arc B390 vs NVIDIA Switch 2 GPU Comparison

Where Each One Wins

The Intel Arc B390 and the NVIDIA Switch 2 GPU occupy entirely different performance tiers according to the recorded data. The Intel part is a mobile integrated graphics solution for Panther Lake laptops, while the NVIDIA part is a dedicated console GPU for the Nintendo Switch 2. Their benchmark presence could not be more different: the Arc B390 has a single recorded 3DMark Steel Nomad DX12 score of 1482, placing it in the 9th percentile of all GPUs in the database. The Switch 2 GPU has no recorded benchmark scores at all, with an average benchmark score of 0, yet it sits in the 50th percentile based on its hardware profile.

The use-case split is stark. The Arc B390 targets portable computing devices where an integrated GPU must handle light gaming and general graphics acceleration without dedicated memory. Its system-shared memory configuration and IGP bus interface confirm this role. The Switch 2 GPU, conversely, is a fixed-function console part with 12 GB of dedicated LPDDR5X memory and a 128-bit bus, designed to deliver consistent performance in a locked hardware environment. The Arc B390 wins on raw compute throughput: its FP32 rating of 7.680 TFLOPS more than doubles the Switch 2 GPU's 4.301 TFLOPS. In texture fill rate, the Intel part hits 120.0 GTexel/s versus 67.20 GTexel/s for the NVIDIA console chip. Pixel rate also favors Intel at 60.00 GPixel/s against 22.40 GPixel/s.

However, the Switch 2 GPU counters with memory bandwidth. Its 102.4 GB/s is fixed and dedicated, while the Arc B390's bandwidth is "System Dependent," meaning it borrows from system memory with no guaranteed figure. In a console, that dedicated bandwidth ensures predictable frame pacing. The Intel part's higher clocks, 300 MHz base and 2500 MHz boost versus 561 MHz base and 1400 MHz boost, explain its compute advantage, but the NVIDIA chip's lower 40 W TDP versus 80 W for Intel suggests the console part is tuned for efficiency in a fixed thermal envelope.

Architecture Differences

The two GPUs come from different architectural generations and process nodes. The Intel Arc B390 uses the Xe3-LPG architecture on a 3 nm process fabricated by Intel, part of the "Arc Graphics-M (Panther Lake)" generation. The NVIDIA Switch 2 GPU uses the older Ampere architecture on an 8 nm process fabricated by Samsung, belonging to the "Console GPU (Nintendo)" generation. This process gap is significant: 3 nm versus 8 nm means the Intel part can pack more transistors per area, though exact transistor counts are unknown for both.

Both GPUs share identical shading unit and RT core counts: 1536 shading units and 12 RT cores each. Texture mapping units match at 48 per GPU. The divergence appears in ROPs: Intel has 24, NVIDIA has 16. The Switch 2 GPU adds 48 tensor cores, which the Arc B390 lacks entirely. This is a crucial architectural difference for AI workloads, though neither part's recorded benchmarks test tensor performance. The Switch 2 GPU also has a known die size of 200 mm², while the Intel die size is unknown.

Clock behavior separates them further. The Arc B390 boosts to 2500 MHz from a 300 MHz base, a wide range typical of mobile parts that scale down aggressively at idle. The Switch 2 GPU runs at 561 MHz base and 1400 MHz boost, a narrower range suited to a fixed console power budget. Memory architecture is fundamentally different: Intel uses system-shared memory with no dedicated VRAM, while NVIDIA uses 12 GB of LPDDR5X on a 128-bit bus at 800 MHz (6.4 Gbps effective). The Intel memory bus width is also system-shared, making its bandwidth entirely dependent on the host system's memory configuration.

API support is identical on paper: both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. Physical form factors diverge sharply. The Arc B390 is an IGP with no slot width, no power connectors, and display outputs described as "Portable Device Dependent." The Switch 2 GPU has no display outputs at all, as it drives an integrated console display. The NVIDIA part's dimensions are recorded at 272 mm by 116 mm by 14 mm, reflecting the console body rather than the GPU alone. The Intel part's dimensions are unknown.

Head-to-Head Benchmarks

The head-to-head benchmark table is empty, and the win counts are zero for both GPUs. No direct comparative test exists in the database. The only quantitative comparison available comes from the Intel part's matching against its nearest rivals and the theoretical compute figures.

The Arc B390's single score of 1482 in 3DMark Steel Nomad DX12 places it just ahead of four NVIDIA legacy desktop parts. It leads the GeForce GT 520MX (1463) by 1.3%, the GeForce 800M (1460) by 1.5%, the GeForce GT 625 OEM (1446) by 2.5%, and the GeForce GT 710 (1443) by 2.7%. These are narrow margins, indicating the Arc B390 performs at the level of decade-old entry-level discrete GPUs in this specific DirectX 12 test. The percentile ranking of 9 confirms this: 91% of all GPUs in the database score higher.

The Switch 2 GPU has no recorded benchmark scores, so no direct comparison to the Arc B390 or any rival is possible from the data. Its 50th percentile ranking is based on hardware characteristics, not measured performance. The FP32 compute figures offer a rough proxy: 7.680 TFLOPS for Intel versus 4.301 TFLOPS for NVIDIA, a 78.5% advantage for Intel in raw shader throughput. Texture rate shows a similar gap at 120.0 GTexel/s versus 67.20 GTexel/s, a 78.6% difference. Pixel rate is more lopsided: 60.00 GPixel/s versus 22.40 GPixel/s, a 167.9% advantage for Intel.

These theoretical numbers, however, do not account for the Switch 2 GPU's dedicated memory bandwidth of 102.4 GB/s. The Arc B390's bandwidth is "System Dependent," meaning in a laptop with dual-channel DDR5 memory, it might approach similar figures, but in a single-channel configuration, it could fall well below. The NVIDIA part's fixed bandwidth ensures consistent performance regardless of host system. The 40 W TDP of the Switch 2 GPU versus 80 W for the Arc B390 also indicates that the console chip achieves its results at half the power draw, a significant efficiency metric for a handheld device.

FAQ

Q: Which GPU has higher raw compute performance?

A: The Intel Arc B390 delivers 7.680 TFLOPS FP32 and 15.36 TFLOPS FP16 (2:1), while the NVIDIA Switch 2 GPU delivers 4.301 TFLOPS FP32 and 8.602 TFLOPS FP16 (2:1). The Intel part is roughly 78% higher in both FP32 and FP16 throughput.

Q: How do their memory systems differ?

A: The Arc B390 uses system-shared memory with no dedicated VRAM, and its bandwidth is "System Dependent." The Switch 2 GPU has 12 GB of dedicated LPDDR5X memory on a 128-bit bus at 800 MHz (6.4 Gbps effective), providing a fixed 102.4 GB/s of bandwidth.

Q: Do both GPUs support ray tracing and tensor operations?

A: Both have 12 RT cores for ray tracing. The Switch 2 GPU additionally has 48 tensor cores, while the Arc B390 has no tensor cores listed. Both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Q: What is the performance percentile of each GPU?

A: The Arc B390 sits at the 9th percentile of all GPUs in the database, based on its 3DMark Steel Nomad DX12 score of 1482. The Switch 2 GPU sits at the 50th percentile, though it has no recorded benchmark scores and its percentile is derived from hardware specifications.

Q: How does the Arc B390 compare to its nearest rivals?

A: In 3DMark Steel Nomad DX12, the Arc B390 scores 1482, leading the GeForce GT 520MX by 1.3%, the GeForce 800M by 1.5%, the GeForce GT 625 OEM by 2.5%, and the GeForce GT 710 by 2.7%. All four rivals are NVIDIA desktop or mobile parts from an earlier era.

Q: What are the power requirements for each GPU?

A: The Arc B390 has a TDP of 80 W and uses no power connectors, as it is an integrated GPU. The Switch 2 GPU has a TDP of 40 W, with no power connector data recorded, consistent with a console part.

The Verdict

The data supports a clear split based on intended use. The Intel Arc B390 is the stronger compute part on paper. Its 7.680 TFLOPS FP32, 120.0 GTexel/s texture rate, and 60.00 GPixel/s pixel rate all far exceed the Switch 2 GPU's corresponding figures of 4.301 TFLOPS, 67.20 GTexel/s, and 22.40 GPixel/s. The Arc B390 also benefits from a more advanced 3 nm process node versus 8 nm for NVIDIA, and it has more ROPs (24 versus 16). For any workload that stresses raw shader throughput or texture filtering, the Intel part is the clear choice from the recorded specifications.

The Switch 2 GPU, however, offers advantages in specific areas. Its 12 GB of dedicated LPDDR5X memory with 102.4 GB/s fixed bandwidth removes any dependency on system memory performance. The Arc B390's system-shared memory means its effective bandwidth varies with the host laptop's configuration, an uncontrolled variable. The Switch 2 GPU also includes 48 tensor cores for AI acceleration, which the Arc B390 lacks. At 40 W TDP, the NVIDIA part uses half the power of the 80 W Intel chip, a critical factor for battery-operated handheld consoles.

The benchmark gap is unmeasurable because the Switch 2 GPU has no recorded scores. The Arc B390's 1482 Steel Nomad score and 9th percentile ranking show it performs at entry-level discrete GPU levels from a previous generation. The Switch 2 GPU's 50th percentile hardware ranking suggests it should outperform the Arc B390 in real-world gaming, but no measured data confirms this. The absence of head-to-head benchmarks prevents a definitive performance verdict.

Choosing between them depends entirely on the platform. For a laptop requiring an integrated GPU with high compute throughput, the Arc B390's specifications are compelling. For a fixed console environment where dedicated memory, tensor cores, and low power draw matter more than raw TFLOPS, the Switch 2 GPU is the appropriate part. The database records the Arc B390 with a release date of 2026-01-26, while the Switch 2 GPU launched earlier on 2025-06-04 with a launch MSRP of 449 USD. No launch price exists for the Intel part. The production status for both is listed as Active, so both remain current products in their respective markets.