Intel Arc B390 vs NVIDIA H200 NVL Comparison

Head-to-Head Benchmarks

The recorded benchmark data for these two accelerators comes from different test suites, making a direct apples-to-apples score comparison impossible. The Intel Arc B390 posts a single 3DMark Steel Nomad DX12 result of 1482 points, while the NVIDIA H200 NVL delivers a Geekbench OpenCL score of 334891 points. These workloads measure different aspects of performance, so the raw numbers should not be read as a head-to-head contest. Instead, the database shows where each device stands relative to its own competitive field.

The Intel Arc B390 sits in the 9th percentile of all GPUs in the database. Its closest rivals are all older NVIDIA consumer parts. It leads the NVIDIA GeForce GT 520MX by 1.3%, the NVIDIA GeForce 800M by 1.5%, the NVIDIA GeForce GT 625 OEM by 2.5%, and the NVIDIA GeForce GT 710 by 2.7%. These are narrow margins, all within a few percentage points. The data indicates the Arc B390 is positioned at the very bottom of the performance spectrum, trading places with decade-old entry-level discrete GPUs. Its 3DMark Steel Nomad score of 1482 is a low-tier result, consistent with an integrated graphics part.

The NVIDIA H200 NVL is at the opposite end of the database. It holds the 100th percentile, meaning it outperforms every other recorded GPU in the database on its benchmark. Its Geekbench OpenCL score of 334891 places it ahead of the AMD Instinct MI300X by 5.3%, which scored 317994. The nearest recorded rival is the NVIDIA B200 at 345482, which beats the H200 NVL by 3.1%. The NVIDIA B300 SXM6 AC leads by 9.4% with a score of 369831, while the NVIDIA L40S trails by 13.2% at 295763. The H200 NVL is clearly a top-tier compute accelerator, but the data also shows it is not the absolute fastest in its own family; the B200 and B300 parts both outpace it.

The performance gap between the two reviewed products is enormous. The H200 NVL's Geekbench OpenCL score is roughly 226 times larger than the Arc B390's 3DMark result, though the differing workloads mean this ratio is not a direct performance multiplier. What the data does confirm is that these two devices occupy entirely different performance classes, with no overlap in their nearest rival groups.

Where Each One Wins

The Intel Arc B390 wins in the integrated graphics space. It is an IGP with no power connectors, no bus interface beyond the integrated path, and a 80 W TDP. Its benchmark score of 1482 puts it just ahead of several legacy NVIDIA mobile and OEM parts, which shows it can handle basic 3D workloads at entry-level settings. The 3DMark Steel Nomad DX12 result indicates it supports modern DirectX 12 Ultimate features, including ray tracing hardware with 12 RT cores. This makes it suitable for lightweight gaming and portable device use, where its display outputs are dependent on the host device.

The NVIDIA H200 NVL wins decisively in compute throughput. Its 60.32 TFLOPS FP32 and 120.6 TFLOPS FP16 (2:1) figures dwarf the Arc B390's 7.680 TFLOPS FP32 and 15.36 TFLOPS FP16 (2:1). The H200 NVL also carries 528 tensor cores, which the Arc B390 lacks entirely. Its 141 GB of HBM3e memory with 4.89 TB/s bandwidth is in a different universe from the Arc B390's system-shared memory, whose bandwidth is listed as system dependent. The H200 NVL is a server accelerator with no display outputs, PCIe 5.0 x16 connectivity, and a 600 W TDP. It is designed for data center workloads, large model inference, and high-performance computing, not for rendering frames to a screen.

For gaming and everyday graphics on a portable device, the Arc B390 has the relevant feature set. For compute density, memory capacity, and raw throughput, the H200 NVL is the clear choice. The two products do not compete for the same buyer or the same workload.

Architecture Differences

The Intel Arc B390 uses the Panther Lake chip with Xe3-LPG architecture, built on a 3 nm process at Intel's own foundry. It belongs to the Arc Graphics-M (Panther Lake) generation. The NVIDIA H200 NVL uses the GH100 chip with Hopper architecture, built on a 5 nm process at TSMC. It belongs to the Server Hopper (Hxx) generation.

The Arc B390 has 1536 shading units, 48 TMUs, 24 ROPs, and 12 RT cores. It has no listed tensor cores. Its base clock is 300 MHz with a boost clock of 2500 MHz. Its memory configuration is entirely system shared, with no dedicated VRAM, no fixed bus width, and bandwidth that depends on the host system. Its pixel rate is 60.00 GPixel/s and its texture rate is 120.0 GTexel/s. It supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. Its TDP is 80 W, and it uses no power connectors. The physical slot width is IGP, and the bus interface is also IGP.

The H200 NVL has 16896 shading units, 528 TMUs, and 24 ROPs. It has 528 tensor cores and no listed RT cores. Its base clock is 1365 MHz with a boost clock of 1785 MHz. Its memory clock is 1593 MHz, which translates to 6.4 Gbps effective. It uses 141 GB of HBM3e on a 6144-bit bus, delivering 4.89 TB/s of bandwidth. Its pixel rate is 42.84 GPixel/s and its texture rate is 942.5 GTexel/s. It has no API support listed for DirectX, OpenGL, or Vulkan, which is typical for a compute-focused server part. Its TDP is 600 W, it requires an 8-pin EPS power connector, and a 1000 W power supply is suggested. It is a dual-slot card measuring 267 mm in length and 111 mm in height. The H200 NVL uses PCIe 5.0 x16 as its bus interface and has no display outputs.

The die size difference is substantial. The H200 NVL packs 80,000 million transistors onto an 814 mm² die, for a transistor density of 98.3M per mm². The Arc B390's transistor count and die size are listed as unknown. The H200 NVL's predecessor is Server Ada and its successor is Server Blackwell; the Arc B390 has no listed predecessor or successor.

FAQ

Q: Which device has more shading units?

A: The NVIDIA H200 NVL has 16896 shading units, compared to 1536 on the Intel Arc B390.

Q: Does the Intel Arc B390 support ray tracing?

A: Yes, the Arc B390 has 12 RT cores. The NVIDIA H200 NVL has no listed RT cores.

Q: What memory configuration does each device use?

A: The Arc B390 uses system-shared memory with no dedicated VRAM, no fixed bus width, and system-dependent bandwidth. The H200 NVL uses 141 GB of HBM3e on a 6144-bit bus with 4.89 TB/s bandwidth.

Q: Which device has a higher FP32 throughput?

A: The NVIDIA H200 NVL delivers 60.32 TFLOPS FP32, while the Intel Arc B390 delivers 7.680 TFLOPS FP32.

Q: What is the TDP difference?

A: The Arc B390 has an 80 W TDP and uses no power connectors. The H200 NVL has a 600 W TDP and requires an 8-pin EPS power connector, with a 1000 W power supply suggested.

Q: Does the H200 NVL have display outputs?

A: No, the H200 NVL has no display outputs. The Arc B390's display outputs are portable device dependent.

Specification Differences

The two devices differ in nearly every recorded specification. The Arc B390 uses a 3 nm Intel process; the H200 NVL uses a 5 nm TSMC process. The H200 NVL has 80,000 million transistors on an 814 mm² die with a density of 98.3M per mm²; the Arc B390's transistor count and die size are unknown. The Arc B390 clocks at 300 MHz base and 2500 MHz boost; the H200 NVL clocks at 1365 MHz base and 1785 MHz boost. Memory differs completely: the Arc B390 uses system-shared memory with system-dependent bandwidth, while the H200 NVL has 141 GB of HBM3e on a 6144-bit bus with 4.89 TB/s bandwidth.

Shading units: 1536 versus 16896. TMUs: 48 versus 528. ROPs are equal at 24. RT cores: 12 on the Arc B390, none listed on the H200 NVL. Tensor cores: none listed on the Arc B390, 528 on the H200 NVL. Pixel rate: 60.00 GPixel/s versus 42.84 GPixel/s. Texture rate: 120.0 GTexel/s versus 942.5 GTexel/s. FP32: 7.680 TFLOPS versus 60.32 TFLOPS. FP16: 15.36 TFLOPS versus 120.6 TFLOPS, both at 2:1 ratio. TDP: 80 W versus 600 W. Slot width: IGP versus dual-slot. Power connectors: none versus 8-pin EPS. Suggested PSU: not listed versus 1000 W. Bus interface: IGP versus PCIe 5.0 x16. Display outputs: portable device dependent versus none. API support: the Arc B390 lists DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4; the H200 NVL lists N/A for all three. Dimensions: the Arc B390 has no listed length or height, while the H200 NVL measures 267 mm by 111 mm. Release dates: the Arc B390 is dated 2026-01-26, the H200 NVL is dated 2024-11-17. Both are listed as active production parts.

The Verdict

The data separates these two products cleanly. The Intel Arc B390 is an integrated graphics solution for portable devices. Its 1482 3DMark Steel Nomad score and 9th percentile ranking place it among the weakest GPUs in the database, just ahead of the NVIDIA GT 520MX, 800M, GT 625 OEM, and GT 710 by margins of 1.3% to 2.7%. It supports modern graphics APIs, includes ray tracing hardware, and draws only 80 W with no external power. It is the appropriate choice for a system that needs basic 3D capability without a discrete card.

The NVIDIA H200 NVL is a server-class compute accelerator. Its Geekbench OpenCL score of 334891 and 100th percentile ranking put it at the top of the database, ahead of the AMD Instinct MI300X by 5.3% and behind only the NVIDIA B200 and B300 SXM6 AC among its nearest recorded rivals. It offers 141 GB of HBM3e memory, 4.89 TB/s bandwidth, 528 tensor cores, and FP16 throughput of 120.6 TFLOPS. It requires 600 W, an 8-pin EPS connector, and a 1000 W power supply. It has no display outputs and no consumer graphics API support. This is the device for compute-heavy server workloads.

There is no genuine competition between these two. The Arc B390 is for portable integrated graphics, the H200 NVL is for data center compute. The benchmark data confirms they serve entirely different purposes, and the specification differences reinforce that conclusion at every level.