Intel Arc Pro B370 vs NVIDIA Switch 2 GPU Comparison

Head-to-Head Benchmarks

The recorded data shows no direct head-to-head benchmark comparisons between the Intel Arc Pro B370 and the NVIDIA Switch 2 GPU. Both products carry an identical percentile versus all GPUs of 50, indicating that within the database’s aggregate ranking, neither part separates itself from the other in overall standing. Each also shows an average benchmark score of zero, which reflects the absence of submitted or verified benchmark runs for both units at the time of data collection.

Without measured wins in either direction, the comparison must rely on the theoretical throughput figures recorded in the specification sheets. The Intel Arc Pro B370 reaches 6.144 TFLOPS of FP32 compute, while the NVIDIA Switch 2 GPU reaches 4.301 TFLOPS. That places the Intel part roughly 43% higher in raw floating-point throughput, a considerable margin for shader-heavy workloads. In FP16, the Intel part delivers 12.29 TFLOPS, versus 8.602 TFLOPS for the NVIDIA part, again a roughly 43% lead for Intel.

Pixel throughput tells a different story. The Intel Arc Pro B370 outputs 48.00 GPixel/s, while the NVIDIA Switch 2 GPU outputs 22.40 GPixel/s. Intel’s advantage here is substantial, more than double the NVIDIA figure. Texture rate follows the same pattern: Intel achieves 96.00 GTexel/s, while NVIDIA achieves 67.20 GTexel/s, a 43% lead for Intel once more. These numbers indicate that the Intel part has a clear edge in fill-rate-bound scenarios, such as high-resolution rasterization or heavy post-processing effects.

However, the NVIDIA Switch 2 GPU counters with a wider memory configuration. The NVIDIA part carries 12 GB of LPDDR5X memory on a 128-bit bus, delivering 102.4 GB/s of bandwidth. The Intel Arc Pro B370 uses system shared memory with a bus width and bandwidth that are system dependent. In a standalone comparison, the NVIDIA part’s dedicated memory subsystem provides a predictable bandwidth ceiling, whereas the Intel part’s performance depends entirely on the host platform’s memory architecture. This makes direct bandwidth comparison impossible, but the existence of a fixed 102.4 GB/s figure for NVIDIA suggests a stable baseline for memory-bound operations.

Clock behavior also differs. The Intel part runs at a 300 MHz base and a 2400 MHz boost. The NVIDIA part runs at a 561 MHz base and a 1400 MHz boost. Intel’s boost clock is 71% higher than NVIDIA’s, which helps explain the compute lead despite Intel having fewer shading units (1280 versus 1536). NVIDIA compensates with more shading units, more texture mapping units (48 versus 40), and more ray tracing cores (12 versus 10). Intel counters with more raster output units (20 versus 16).

Power consumption favors Intel in absolute terms. The Intel Arc Pro B370 has a TDP of 25 W, while the NVIDIA Switch 2 GPU has a TDP of 40 W. That difference means Intel delivers its higher compute throughput while drawing 37.5% less power, a notable efficiency gap. However, the NVIDIA part’s 40 W TDP is typical for a console-class component designed to run within a portable system, and the Intel part’s 25 W TDP reflects its integrated graphics position within a Panther Lake processor.

FAQ

Q: Which GPU has higher FP32 compute performance?

A: The Intel Arc Pro B370 delivers 6.144 TFLOPS of FP32, while the NVIDIA Switch 2 GPU delivers 4.301 TFLOPS. Intel’s part provides approximately 43% more raw FP32 throughput.

Q: Does the NVIDIA Switch 2 GPU have more memory?

A: Yes. The NVIDIA Switch 2 GPU has 12 GB of LPDDR5X memory with a 128-bit bus and 102.4 GB/s bandwidth. The Intel Arc Pro B370 uses system shared memory, so its memory size, type, and bandwidth are system dependent.

Q: How do the two GPUs compare in terms of power consumption?

A: The Intel Arc Pro B370 has a TDP of 25 W, while the NVIDIA Switch 2 GPU has a TDP of 40 W. The Intel part consumes less power despite having higher compute throughput.

Q: Which GPU has more shading units?

A: The NVIDIA Switch 2 GPU has 1536 shading units, while the Intel Arc Pro B370 has 1280 shading units. NVIDIA also has more texture mapping units (48 versus 40) and more ray tracing cores (12 versus 10).

Q: Are the two GPUs comparable in pixel fill rate?

A: No. The Intel Arc Pro B370 achieves 48.00 GPixel/s, while the NVIDIA Switch 2 GPU achieves 22.40 GPixel/s. Intel’s pixel throughput is more than double NVIDIA’s.

Q: What process nodes do the two GPUs use?

A: The Intel Arc Pro B370 uses a 3 nm process manufactured by Intel. The NVIDIA Switch 2 GPU uses an 8 nm process manufactured by Samsung.

Architecture Differences

The Intel Arc Pro B370 uses the Xe3-LPG architecture, built on Intel’s Panther Lake chip, and belongs to the Arc Graphics-WM generation. The process node is 3 nm, fabricated by Intel, with the transistor count and die size listed as unknown. The NVIDIA Switch 2 GPU uses the Ampere architecture, built on the GA10B chip, and belongs to the Console GPU generation for Nintendo. The process node is 8 nm, fabricated by Samsung, with a die size of 200 mm² and transistor count listed as unknown.

The architectural lineage differs sharply. Intel’s Xe3-LPG is a low-power graphics variant designed for integrated use, which matches the IGP slot width and the absence of power connectors. NVIDIA’s Ampere is a widely deployed architecture across both discrete and console products, and the Switch 2 GPU operates as a separate console component with its own memory subsystem. The Intel part has no display outputs listed beyond “portable device dependent,” while the NVIDIA part explicitly lists “no outputs,” indicating that both are not intended for standalone display connection.

Ray tracing support exists on both, but the counts differ. Intel has 10 ray tracing cores, NVIDIA has 12. Tensor cores are absent from the Intel specification, while NVIDIA includes 48 tensor cores. This difference points toward divergent feature sets: NVIDIA’s tensor cores enable AI-accelerated features, while Intel’s part does not list any equivalent hardware.

Memory architecture represents a fundamental split. Intel uses system shared memory, meaning the GPU borrows from the host processor’s memory pool, with bandwidth described as “system dependent.” NVIDIA uses dedicated 12 GB of LPDDR5X memory on a 128-bit bus with a fixed 102.4 GB/s bandwidth. This means the NVIDIA part has a predictable memory performance profile, while the Intel part’s memory behavior varies with the host platform.

Clock speeds also reflect different design strategies. Intel’s base clock is 300 MHz with a boost clock of 2400 MHz, a very wide range that suggests aggressive dynamic scaling. NVIDIA’s base clock is 561 MHz with a boost clock of 1400 MHz, a narrower range that may indicate a more consistent operating point for console workloads. The Intel part’s higher boost clock contributes to its higher compute and fill rates.

Both GPUs support the same API feature set: DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. This means software written for either API should run on both, though the underlying hardware capabilities differ.

The Verdict

The data indicates a clear split in intended use cases. The Intel Arc Pro B370, as an integrated graphics solution within a Panther Lake processor, targets low-power portable or compact systems where the CPU and GPU share memory. Its 25 W TDP and system shared memory point toward a design that prioritizes efficiency and platform integration over dedicated memory bandwidth.

The NVIDIA Switch 2 GPU, as a console component, targets a fixed hardware environment with dedicated memory. Its 40 W TDP, 12 GB of LPDDR5X memory, and 102.4 GB/s bandwidth provide a stable foundation for game developers who need predictable performance across a uniform platform.

From a pure compute standpoint, the Intel part wins decisively. Its FP32 throughput of 6.144 TFLOPS exceeds NVIDIA’s 4.301 TFLOPS by a substantial margin, and its pixel rate of 48.00 GPixel/s more than doubles NVIDIA’s 22.40 GPixel/s. For workloads that rely heavily on shader math or fill rate, the Intel part appears stronger on paper.

However, the NVIDIA part offers more shading units, more texture units, more ray tracing cores, and tensor cores. These additional resources may translate to better performance in specific scenarios, particularly those that leverage tensor core acceleration or benefit from a higher count of execution units. The Intel part lacks tensor cores entirely, which could limit its ability to run certain AI-enhanced features.

Neither part has a recorded benchmark score, so real-world performance remains unverified. The percentile versus all GPUs is identical at 50 for both, suggesting that the database places them at the same level overall. That equivalence is notable given the compute differences, possibly because the NVIDIA part’s memory advantage and unit counts balance out Intel’s raw throughput lead.

For a user choosing between these two, the decision hinges on the host platform. The Intel Arc Pro B370 is an IGP, meaning it comes with a specific processor and cannot be purchased standalone. The NVIDIA Switch 2 GPU is a console component, likely fixed inside a gaming device. Neither offers upgrade flexibility in the traditional discrete GPU sense.

Specification Differences

The two parts differ across nearly every major specification category. The Intel Arc Pro B370 uses a 3 nm process from Intel, while the NVIDIA Switch 2 GPU uses an 8 nm process from Samsung. Die size is unknown for Intel, while NVIDIA’s die measures 200 mm². Transistor counts are unknown for both.

Clock speeds differ substantially. Intel has a base clock of 300 MHz and a boost clock of 2400 MHz. NVIDIA has a base clock of 561 MHz and a boost clock of 1400 MHz. Memory differs entirely: Intel uses system shared memory with no fixed size, type, bus width, or bandwidth, while NVIDIA uses 12 GB of LPDDR5X on a 128-bit bus with 102.4 GB/s bandwidth.

Shader resources differ. Intel has 1280 shading units, 40 TMUs, 20 ROPs, and 10 ray tracing cores. NVIDIA has 1536 shading units, 48 TMUs, 16 ROPs, 12 ray tracing cores, and 48 tensor cores. Intel lists no tensor cores.

Pixel rate is 48.00 GPixel/s for Intel versus 22.40 GPixel/s for NVIDIA. Texture rate is 96.00 GTexel/s versus 67.20 GTexel/s. FP32 is 6.144 TFLOPS versus 4.301 TFLOPS. FP16 is 12.29 TFLOPS versus 8.602 TFLOPS, both at 2:1 ratio.

Power consumption differs: Intel has a 25 W TDP, NVIDIA has a 40 W TDP. Slot width is IGP for Intel, unknown for NVIDIA. Power connectors are none for Intel, unknown for NVIDIA. Bus interface is IGP for Intel, unknown for NVIDIA. Display outputs are portable device dependent for Intel, and no outputs for NVIDIA.

Physical dimensions are only specified for NVIDIA: 272 mm in length, 116 mm in height, and 14 mm in width. Intel has no dimensions listed. Release dates differ: Intel released on 2026-01-26, NVIDIA on 2025-06-04. The NVIDIA part carries a launch MSRP of 449 USD; Intel has no launch MSRP. Production status is active for both.

Where Each One Wins

The Intel Arc Pro B370 wins in raw compute throughput. Its FP32 figure of 6.144 TFLOPS leads the NVIDIA Switch 2 GPU’s 4.301 TFLOPS by a wide margin. FP16 follows the same pattern at 12.29 TFLOPS versus 8.602 TFLOPS. Pixel fill rate also favors Intel significantly: 48.00 GPixel/s versus 22.40 GPixel/s. Texture fill rate favors Intel as well, at 96.00 GTexel/s versus 67.20 GTexel/s. Clock speed is another Intel advantage, with a 2400 MHz boost versus NVIDIA’s 1400 MHz boost. Power efficiency favors Intel, as its 25 W TDP delivers higher compute than NVIDIA’s 40 W TDP. The 3 nm process node also represents a manufacturing advantage over NVIDIA’s 8 nm node.

The NVIDIA Switch 2 GPU wins in memory configuration. Its 12 GB of LPDDR5X memory on a 128-bit bus with 102.4 GB/s bandwidth is a fixed, dedicated resource, whereas Intel’s system shared memory depends entirely on the host platform. NVIDIA also wins in execution unit counts: 1536 shading units versus 1280, 48 TMUs versus 40, 12 ray tracing cores versus 10, and the presence of 48 tensor cores versus none on Intel. These additional units may provide advantages in workloads that scale with parallel execution resources rather than raw clock speed or fill rate.

The release schedule differs as well. NVIDIA’s part launched on 2025-06-04, while Intel’s part launched on 2026-01-26. This means NVIDIA’s product has been available earlier, though both remain active in production. The NVIDIA part also has a published launch MSRP of 449 USD, while Intel’s part has no MSRP, consistent with its integrated nature.

Use-case splits follow these specifications. The Intel part suits systems where CPU and GPU share memory and where power draw must stay low. Its high boost clock and fill rates make it appropriate for compute-heavy tasks that do not depend on dedicated memory bandwidth. The NVIDIA part suits fixed console environments where developers can rely on a known memory bandwidth and a consistent hardware configuration. Its tensor cores and higher unit counts may benefit AI-assisted rendering or ray tracing workloads that leverage those resources.

The data does not assign a definitive winner. Both parts sit at the 50th percentile versus all GPUs, and neither has recorded benchmark scores. The choice between them depends on whether the user values raw compute and efficiency (Intel) or dedicated memory and execution unit breadth (NVIDIA).