Intel Arc Graphics 128EU Mobile vs NVIDIA H20 NVL16 Comparison

Where Each One Wins

The recorded data provides no benchmark scores for either the Intel Arc Graphics 128EU Mobile or the NVIDIA H20 NVL16. Both entries show an average benchmark score of zero, and the head-to-head benchmark list is empty. Consequently, the wins count for each product stands at zero. The percentile versus all GPUs is 50 for both, placing them at the midpoint of the database distribution despite their vastly different intended roles.

Without benchmark results, the use-case split must be derived from the architectural and specification records. The Intel Arc Graphics 128EU Mobile is an integrated graphics processor (IGP) with a 28 W TDP, designed for portable devices where display output is device-dependent. It supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4, indicating a focus on general graphics rendering and consumer application compatibility. The NVIDIA H20 NVL16 is a server accelerator with no display outputs, no DirectX, OpenGL, or Vulkan support, and a 400 W TDP. It is built for compute workloads, specifically those leveraging its 312 tensor cores and 96 GB HBM3 memory.

The data suggests the Intel part wins in scenarios requiring integrated graphics, low power envelopes, and standard API support for client devices. The NVIDIA part wins in server-side compute, high-throughput memory access, and tensor-based operations. There is no overlap in their recorded feature sets that would create competitive benchmark scenarios.

FAQ

Q: What is the process node for each GPU?

A: The Intel Arc Graphics 128EU Mobile uses a 10 nm process from Intel. The NVIDIA H20 NVL16 uses a 5 nm process from TSMC.

Q: How much memory does each product have?

A: The Intel part has system-shared memory, with size, type, bus width, and bandwidth all listed as system-dependent. The NVIDIA H20 NVL16 has 96 GB of HBM3 memory on a 6144-bit bus with 4.03 TB/s of bandwidth.

Q: What are the thermal design power ratings?

A: The Intel Arc Graphics 128EU Mobile has a 28 W TDP. The NVIDIA H20 NVL16 has a 400 W TDP, with a suggested PSU of 800 W.

Q: Which APIs are supported by each GPU?

A: The Intel part supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The NVIDIA H20 NVL16 has no API support recorded, with DirectX, OpenGL, and Vulkan all listed as N/A.

Q: What are the boost clock speeds?

A: The Intel Arc Graphics 128EU Mobile boosts to 2250 MHz. The NVIDIA H20 NVL16 boosts to 1980 MHz.

Q: What is the transistor count for each chip?

A: The Intel chip has no transistor count recorded. The NVIDIA GH100 chip has 80,000 million transistors on an 814 mm² die.

Head-to-Head Benchmarks

The head-to-head benchmark table is empty, so no direct performance comparisons can be drawn from measured results. However, the raw specification data offers a basis for theoretical comparison. The NVIDIA H20 NVL16 delivers 39.54 TFLOPS of FP32 compute, which is 8.58 times the 4.608 TFLOPS of the Intel Arc Graphics 128EU Mobile. In FP16, the NVIDIA part achieves 79.07 TFLOPS, versus 9.216 TFLOPS for the Intel part, a ratio of approximately 8.58 as well.

Texture rate follows a similar pattern. The NVIDIA H20 NVL16 reaches 617.8 GTexel/s, while the Intel part reaches 144.0 GTexel/s, meaning the NVIDIA accelerator is 4.29 times faster in texture throughput. Pixel rate is a different story. The Intel Arc Graphics 128EU Mobile records 72.00 GPixel/s, which is 1.52 times the 47.52 GPixel/s of the NVIDIA H20 NVL16. This indicates the Intel integrated GPU has a higher pixel fill rate despite its far lower overall compute and texture capabilities.

Memory bandwidth is heavily skewed toward the NVIDIA part. The H20 NVL16 offers 4.03 TB/s from its HBM3 stack, while the Intel part's bandwidth is system-dependent, meaning it shares main memory with the CPU and has no fixed figure in the database. The NVIDIA part also has 9984 shading units versus 1024 for the Intel part, 312 tensor cores versus none recorded for Intel, and 312 TMUs versus 64. The Intel part has 32 ROPs, while the NVIDIA part has 24, which aligns with the pixel rate finding.

Clock speeds differ as well. The Intel part has a base clock of 300 MHz and a boost of 2250 MHz. The NVIDIA part has a base of 1830 MHz and a boost of 1980 MHz. The Intel part's boost is 270 MHz higher, but its base is 1530 MHz lower, reflecting the integrated design's power-saving idle state.

Specification Differences

The two products differ in nearly every recorded specification. The Intel Arc Graphics 128EU Mobile is built on a 10 nm Intel process, while the NVIDIA H20 NVL16 uses a 5 nm TSMC process. The Intel chip is Meteor Lake with Xe-LPG architecture; the NVIDIA chip is GH100 with Hopper architecture.

Memory is a major divergence. The Intel part has system-shared memory with no dedicated size, type, bus width, or bandwidth. The NVIDIA part has 96 GB of HBM3, a 6144-bit bus, and 4.03 TB/s bandwidth. The Intel memory clock is listed as system-shared, while the NVIDIA memory clock is 1313 MHz with 5.3 Gbps effective.

Compute units differ significantly. The Intel part has 1024 shading units, 64 TMUs, and 32 ROPs. The NVIDIA part has 9984 shading units, 312 TMUs, and 24 ROPs. The NVIDIA part also has 312 tensor cores; the Intel part has none recorded. Neither has RT cores listed.

Power and physical design are distinct. The Intel part has a 28 W TDP and is an IGP with a bus interface of Ring Bus. The NVIDIA part has a 400 W TDP, an 800 W suggested PSU, and is an SXM Module with PCIe 5.0 x16 interface. The Intel part's display outputs are portable device dependent; the NVIDIA part has no outputs.

API support is exclusive to the Intel part. DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4 are all listed for Intel, while NVIDIA shows N/A for each. Release dates differ: the Intel part was released on 2023-12-13, and the NVIDIA part on 2025-09-01.

Architecture Differences

The Intel Arc Graphics 128EU Mobile uses the Xe-LPG architecture on the Meteor Lake chip. This is a graphics-focused design with 1024 shading units arranged for consumer rendering tasks. It supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4, and its pixel rate of 72.00 GPixel/s suggests a strong focus on rasterized output for integrated use. The architecture has no tensor cores recorded, indicating no dedicated AI acceleration hardware in the database.

The NVIDIA H20 NVL16 uses the Hopper architecture on the GH100 chip. This is a server-focused compute design with 9984 shading units and 312 tensor cores. The tensor cores are the key differentiator, enabling workloads that require matrix math, such as AI inference and training. The Hopper architecture supports no graphics APIs in the database, confirming its role as an accelerator rather than a display adapter. The 80,000 million transistor count on an 814 mm² die with a transistor density of 98.3M per mm² shows a manufacturing approach aimed at dense compute, not integrated efficiency.

The Intel part's 10 nm process is larger than the NVIDIA part's 5 nm process, but the Intel chip is designed for low power (28 W) and system integration. The NVIDIA part's 400 W TDP and 800 W suggested PSU reflect a design that prioritizes raw throughput over energy efficiency. The Intel part's boost clock of 2250 MHz is higher than the NVIDIA part's 1980 MHz, but the NVIDIA part's base clock of 1830 MHz is far higher than Intel's 300 MHz, indicating the NVIDIA chip runs near its maximum continuously.

The memory architecture also reflects the differing purposes. The Intel part shares system memory, making its bandwidth dependent on the host platform. The NVIDIA part uses dedicated HBM3 with 4.03 TB/s, a fixed and massive bandwidth that suits large data movement in server workloads. The Intel part's texture rate of 144.0 GTexel/s and FP32 of 4.608 TFLOPS are modest, but its pixel rate of 72.00 GPixel/s is higher than the NVIDIA part's 47.52 GPixel/s, a quirk of the ROP count difference (32 for Intel, 24 for NVIDIA).

The Verdict

The data indicates two products with no shared use case. The Intel Arc Graphics 128EU Mobile is an integrated GPU for portable devices, with a 28 W TDP, system-shared memory, and full graphics API support. It offers 4.608 TFLOPS FP32, 144.0 GTexel/s, and 72.00 GPixel/s, and it can output to displays. Its release date of 2023-12-13 places it in the Meteor Lake generation, and it is listed as active in production.

The NVIDIA H20 NVL16 is a server accelerator with a 400 W TDP, 96 GB HBM3, and 4.03 TB/s bandwidth. It delivers 39.54 TFLOPS FP32, 79.07 TFLOPS FP16, and 617.8 GTexel/s, with 312 tensor cores for compute acceleration. It has no display outputs and no graphics API support. Its release date of 2025-09-01 places it in the Server Hopper generation, and it is also active.

The selection between these two is determined by the workload. For any client-side graphics rendering, integrated display, or application requiring DirectX, OpenGL, or Vulkan, the Intel part is the only option with those capabilities in the database. For server-side compute, high-bandwidth memory access, or tensor-based processing, the NVIDIA part is the only option with those features. The Intel part's higher pixel rate (72.00 GPixel/s versus 47.52 GPixel/s) and higher boost clock (2250 MHz versus 1980 MHz) do not compensate for the NVIDIA part's 8.58 times higher FP32 throughput and 4.29 times higher texture rate. The database shows no benchmark overlap, so no direct performance ranking can be assigned beyond the specification deltas.