Intel Arc B370 vs NVIDIA H100 SXM5 96 GB Comparison

Head-to-Head Benchmarks

The database contains a single recorded benchmark result for the Intel Arc B370: a 3DMark Steel Nomad DX12 score of 1184. The NVIDIA H100 SXM5 96 GB has no recorded benchmark scores in the database, and there are no direct head-to-head benchmark entries comparing the two. This makes a performance comparison challenging, but the available data does provide some context.

The Arc B370's score of 1184 places it in the 5th percentile of all GPUs tracked in the database. Its nearest rivals, based on average scores, are all older discrete graphics solutions. The ATI Mobility Radeon HD 5570 averages 1186, a delta of -0.2% relative to the Arc B370. The ATI Radeon HD 5770 averages 1190, a delta of -0.5%. The AMD Radeon HD 7650M averages 1192, a delta of -0.7%. The AMD FirePro M2000 averages 1168, a delta of +1.4%. These deltas are all within a narrow 2% band, indicating that the Arc B370's raw score sits closely alongside these legacy parts.

The H100 SXM5, by contrast, has an average benchmark score of 0 in the database, with no percentile ranking beyond the neutral 50th percentile default. The database records no benchmark runs for this accelerator. Without a Steel Nomad score or any comparable workload result, the data cannot quantify how the H100 performs relative to the Arc B370 in the same test. The only direct conclusion from the recorded data is that the Arc B370 has a measurable score of 1184, while the H100 has no comparable measurement recorded.

The H100's specifications suggest it targets a different performance class. Its FP32 throughput is listed at 66.91 TFLOPS, compared to the Arc B370's 6.144 TFLOPS. Its FP16 throughput is 267.6 TFLOPS (4:1 ratio), versus 12.29 TFLOPS (2:1) for the Arc. Texture rate is 1,045.4 GTexel/s against 96.00 GTexel/s. Pixel rates are close: 47.52 GPixel/s for the H100 and 48.00 GPixel/s for the Arc B370. These figures indicate a massive compute and texture throughput advantage for the H100, though no benchmark score confirms how that translates in the Steel Nomad workload.

Architecture Differences

The two processors come from fundamentally different design philosophies. The Intel Arc B370 uses the Xe3-LPG architecture, built on a 3 nm process at Intel's own foundry. It is part of the Arc Graphics-M (Panther Lake) generation, using the Panther Lake chip. The H100 SXM5 uses the Hopper architecture, built on a 5 nm process at TSMC, using the GH100 chip. The H100 belongs to the Server Hopper (Hxx) generation.

Transistor counts differ enormously. The H100 packs 80,000 million transistors on a die size of 814 mm², yielding a transistor density of 98.3 million per square millimeter. The Arc B370's transistor count and die size are listed as unknown in the database, so no density figure can be computed.

The memory subsystems are entirely different in kind. The Arc B370 uses system shared memory, with its memory size, type, and bus width all listed as "System Shared." Its memory bandwidth is "System Dependent," meaning it relies on the host system's memory configuration. The H100 uses 96 GB of HBM3 memory on a 5120-bit bus, with a fixed bandwidth of 3.36 TB/s. This is a dedicated, high-bandwidth memory pool designed for data center workloads, not a shared pool.

Clock behavior also diverges. The Arc B370 has a base clock of 300 MHz and a boost clock of 2400 MHz. The H100 has a base clock of 1350 MHz and a boost clock of 1980 MHz. The H100's memory clock is listed as 1313 MHz, or 5.3 Gbps effective. The Arc B370's memory clock is tied to the system shared memory, so no dedicated figure exists.

The compute resources differ in scale. The Arc B370 has 1280 shading units, 40 texture mapping units, 20 raster output units, and 10 ray tracing cores. It has no tensor cores listed. The H100 has 16,896 shading units, 528 texture mapping units, 24 raster output units, and 528 tensor cores. The H100 lists no ray tracing cores. The H100's shading unit count is over 13 times that of the Arc B370, and its tensor core count is substantial, reflecting its AI and HPC focus.

API support also separates them. The Arc B370 supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The H100 lists no DirectX, OpenGL, or Vulkan support in the database. This is consistent with a server accelerator that does not target consumer graphics APIs.

Where Each One Wins

Based on the recorded data, the Intel Arc B370 wins in the only benchmark with a measured score. Its 1184 result in 3DMark Steel Nomad DX12 is the sole benchmark figure available for either product. The H100 has no benchmark scores in the database, so the Arc B370 is ahead by default in that specific test.

The H100 wins decisively in raw compute specifications. Its FP32 throughput of 66.91 TFLOPS is roughly 10.9 times the Arc B370's 6.144 TFLOPS. Its FP16 throughput of 267.6 TFLOPS is roughly 21.8 times the Arc B370's 12.29 TFLOPS. Texture rate favors the H100 by a factor of about 10.9 as well. The H100's 96 GB of HBM3 memory with 3.36 TB/s bandwidth is in a different class from the Arc B370's system shared memory, where bandwidth is system dependent and not quantified in the database.

Pixel rate is nearly identical, with the Arc B370 at 48.00 GPixel/s and the H100 at 47.52 GPixel/s. This is a narrow margin in favor of the Arc B370, but it is a specification difference, not a benchmark result.

The Arc B370 wins on power efficiency. Its TDP is 25 W, while the H100's TDP is 700 W. The Arc B370 requires no power connectors and is an integrated graphics processor (IGP) with a slot width of IGP. The H100 is an SXM module requiring an 8-pin EPS power connector and a suggested PSU of 1100 W. The Arc B370 also has display outputs, listed as "Portable Device Dependent," while the H100 has no display outputs at all.

For gaming and client workloads, the Arc B370's API support gives it a functional advantage. It supports DirectX 12 Ultimate, OpenGL 4.6, and Vulkan 1.4, making it usable for consumer graphics applications. The H100 has no listed API support, so it cannot run those workloads in a traditional sense.

Specification Differences

The table below summarizes only the fields where the two products differ, based on the database records.

| Specification | Intel Arc B370 | NVIDIA H100 SXM5 96 GB |

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

| Chip | Panther Lake | GH100 |

| Architecture | Xe3-LPG | Hopper |

| Generation | Arc Graphics-M (Panther Lake) | Server Hopper (Hxx) |

| Process Node | 3 nm | 5 nm |

| Foundry | Intel | TSMC |

| Transistors | unknown | 80,000 million |

| Die Size | unknown | 814 mm² |

| Transistor Density | null | 98.3M / mm² |

| Base Clock | 300 MHz | 1350 MHz |

| Boost Clock | 2400 MHz | 1980 MHz |

| Memory Clock | System Shared | 1313 MHz, 5.3 Gbps effective |

| Memory Size | System Shared | 96 GB |

| Memory Type | System Shared | HBM3 |

| Memory Bus Width | System Shared | 5120 bit |

| Memory Bandwidth | System Dependent | 3.36 TB/s |

| Shading Units | 1280 | 16896 |

| TMUs | 40 | 528 |

| ROPs | 20 | 24 |

| RT Cores | 10 | null |

| Tensor Cores | null | 528 |

| Pixel Rate | 48.00 GPixel/s | 47.52 GPixel/s |

| Texture Rate | 96.00 GTexel/s | 1,045.4 GTexel/s |

| FP32 | 6.144 TFLOPS | 66.91 TFLOPS |

| FP16 | 12.29 TFLOPS (2:1) | 267.6 TFLOPS (4:1) |

| TDP | 25 W | 700 W |

| Slot Width | IGP | SXM Module |

| Power Connectors | None | 8-pin EPS |

| Suggested PSU | null | 1100 W |

| Bus Interface | IGP | PCIe 5.0 x16 |

| Display Outputs | Portable Device Dependent | No outputs |

| DirectX | 12 Ultimate (12_2) | null |

| OpenGL | 4.6 | null |

| Vulkan | 1.4 | null |

| Release Date | 2026-01-26 | 2023-03-20 |

| Predecessor | null | Server Ada |

| Successor | null | Server Blackwell |

| Percentile | 5 | 50 |

| Average Benchmark Score | 1184 | 0 |

The two share no identical specifications in the database. Even the pixel rate, which is close, differs: 48.00 versus 47.52 GPixel/s.

FAQ

Q: Which GPU has a higher benchmark score in the database?

A: The Intel Arc B370 has a recorded 3DMark Steel Nomad DX12 score of 1184. The NVIDIA H100 SXM5 96 GB has no recorded benchmark scores, so the Arc B370 is the only one with a measurable result.

Q: What is the memory capacity difference?

A: The Arc B370 uses system shared memory with no dedicated capacity. The H100 has 96 GB of HBM3 memory on a 5120-bit bus.

Q: How do the FP32 performances compare?

A: The H100 delivers 66.91 TFLOPS of FP32 throughput. The Arc B370 delivers 6.144 TFLOPS. The H100 is about 10.9 times higher.

Q: Does the Arc B370 support ray tracing?

A: Yes, it has 10 ray tracing cores. The H100 lists no ray tracing cores in the database.

Q: What are the power requirements?

A: The Arc B370 has a 25 W TDP and requires no power connectors. The H100 has a 700 W TDP, requires an 8-pin EPS power connector, and has a suggested PSU of 1100 W.

Q: Which GPU supports DirectX?

A: The Arc B370 supports DirectX 12 Ultimate (12_2), along with OpenGL 4.6 and Vulkan 1.4. The H100 lists no DirectX, OpenGL, or Vulkan support.

The Verdict

The data shows two products aimed at entirely different use cases. The Intel Arc B370 is an integrated graphics processor with a 25 W TDP, system shared memory, and full support for consumer graphics APIs. Its only benchmark score, 1184 in Steel Nomad DX12, places it in the 5th percentile of all GPUs, with nearest rivals being legacy ATI and AMD parts from over a decade ago. It delivers 6.144 TFLOPS FP32 and includes 10 ray tracing cores. It is designed to fit into a Panther Lake mobile platform, with display outputs dependent on the portable device.

The NVIDIA H100 SXM5 96 GB is a server accelerator with no benchmark scores in the database. Its specifications point to compute density: 16,896 shading units, 528 tensor cores, 96 GB of HBM3 with 3.36 TB/s bandwidth, and 66.91 TFLOPS FP32. It has a 700 W TDP, requires a dedicated power connector and an 1100 W PSU, and has no display outputs. It does not support consumer graphics APIs in the database records.

For a builder selecting a part for client-side graphics, the Arc B370 is the only choice with recorded graphics API support and a measurable benchmark result. For a data center workload requiring massive FP16 throughput, tensor cores, and dedicated high-bandwidth memory, the H100's specifications dominate. The pixel rates are nearly identical, but everything else diverges sharply.

The verdict from the data is straightforward: the Arc B370 is for integrated, low-power client graphics; the H100 is for high-performance server compute. Neither product can substitute for the other in its intended role. The Arc B370's 1184 benchmark score is the only performance measurement available, but it does not reflect the H100's capabilities in its target workloads, for which no benchmark data exists in the database.