Intel Arc Pro B390 vs NVIDIA Jetson T5000 Comparison

Head-to-Head Benchmarks

The recorded data contains no direct head-to-head benchmark scores for the Intel Arc Pro B390 and the NVIDIA Jetson T5000. Both entries show an average benchmark score of zero and a percentile ranking of 50 against all GPUs in the database, which indicates that neither part has accumulated standardized performance results in the current dataset. Without measured frame rates, render times, or synthetic scores, the comparison must rely on the architectural specifications and computed throughput values that are documented for each product.

The raw compute figures provide the clearest point of contrast. The Intel Arc Pro B390 delivers 7.680 TFLOPS of FP32 performance, while the NVIDIA Jetson T5000 delivers 8.064 TFLOPS. That places the Jetson T5000 approximately 5% ahead of the Arc Pro B390 in single-precision floating-point throughput. In FP16 workloads, the situation reverses: the Arc Pro B390 reaches 15.36 TFLOPS using a 2:1 ratio, whereas the Jetson T5000 matches its FP32 figure at 8.064 TFLOPS with a 1:1 ratio. The Intel part therefore offers nearly double the half-precision throughput, which is a decisive advantage for workloads that can use reduced precision.

Texture and pixel throughput tell a similar split story. The Jetson T5000 posts 126.0 GTexel/s against 120.0 GTexel/s for the Arc Pro B390, a narrow lead of about 5%. Pixel rate favors the Intel part, with 60.00 GPixel/s versus 50.40 GPixel/s for the NVIDIA product, an advantage of roughly 19%. These figures reflect the different balance of texture mapping units and render output units on each chip, and they predict which workloads will favor which GPU.

Clock behavior also separates the two. The Arc Pro B390 runs at a 300 MHz base clock and boosts to 2500 MHz, while the Jetson T5000 runs at 1386 MHz base and 1575 MHz boost. The Intel part has a substantially higher boost ceiling, which helps explain how it reaches competitive throughput despite having fewer shading units. The Jetson T5000 compensates with a higher base clock and a larger shader count of 2560 units against 1536 for the Intel chip.

Where Each One Wins

The Intel Arc Pro B390 wins in scenarios that leverage its FP16 capability and pixel throughput. With 15.36 TFLOPS of half-precision compute, the part is positioned for machine learning inference and other reduced-precision workloads that accept FP16 inputs. The 60.00 GPixel/s pixel rate also gives it an edge in rasterization-heavy tasks that saturate render output units, such as display composition or lower-resolution rendering where pixel fill is the limiting factor. The integrated nature of the GPU, tied to a Panther Lake processor, means it operates as an IGP with system-shared memory, making it suitable for compact platforms where a discrete card is not an option.

The NVIDIA Jetson T5000 wins in FP32 compute and texture throughput. Its 8.064 TFLOPS of single-precision performance is the highest raw FP32 number in this comparison, and the 126.0 GTexel/s texture rate is likewise the stronger figure. The part also brings 96 tensor cores, a feature the Intel chip does not list at all, which points to dedicated tensor acceleration for AI workloads. The Jetson T5000 carries 128 GB of LPDDR5X memory on a 256-bit bus with 273.2 GB/s of bandwidth, a massive capacity advantage over the Arc Pro B390's system-shared memory with dependent bandwidth. For edge server deployments that need to hold large models or datasets locally, that memory configuration is a clear differentiator.

The Jetson T5000 also holds the advantage in API support in one specific direction: it uses an NVIDIA Blackwell architecture with tensor cores, which aligns with CUDA-based software ecosystems. The Intel part counters with DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4 support, while the Jetson T5000 lists no DirectX, OpenGL, or Vulkan APIs. Any graphics-oriented application requiring those APIs will only run on the Intel side.

Architecture Differences

The two GPUs come from different manufacturers and different process nodes. Intel builds the Arc Pro B390 on a 3 nm process at Intel foundries, using the Panther Lake chip with Xe3-LPG architecture. NVIDIA builds the Jetson T5000 on a 5 nm process at TSMC, using the GB10B chip with Blackwell architecture. The Intel part belongs to the Arc Graphics-WM (Panther Lake) generation, while the NVIDIA part belongs to the Server Blackwell (Bxx) generation.

Shader resources differ significantly. The Arc Pro B390 contains 1536 shading units, 48 texture mapping units, 24 render output units, and 12 ray tracing cores. The Jetson T5000 contains 2560 shading units, 80 texture mapping units, 32 render output units, 20 ray tracing cores, and 96 tensor cores. The NVIDIA chip has more of every execution resource except for the missing tensor core count on the Intel side, which simply is not documented for the Arc Pro B390.

Memory architecture is fundamentally different. The Arc Pro B390 uses system-shared memory with a system-dependent bandwidth figure, making its performance reliant on the host platform's memory subsystem. The Jetson T5000 has dedicated 128 GB of LPDDR5X memory on a 256-bit bus, delivering 273.2 GB/s of bandwidth. The clock for that memory is listed at 1067 MHz with 8.5 Gbps effective data rate. This distinction matters for any workload that repeatedly accesses large datasets, since the Jetson T5000's dedicated pool avoids contention with CPU memory traffic.

Power and physical design also diverge. The Arc Pro B390 has an 80 W TDP, while the Jetson T5000 has a 120 W TDP. Both are integrated form factors with no power connectors, but the Jetson T5000 lists a suggested PSU of 300 W, while the Intel part has no suggested PSU figure. The Jetson T5000 measures 87 mm by 100 mm by 15 mm and uses a PCIe 5.0 x8 bus interface. The Arc Pro B390 uses an IGP bus interface with no listed dimensions.

Display output is another major split. The Arc Pro B390 supports display outputs described as portable device dependent, which means it can drive screens in compatible systems. The Jetson T5000 has no display outputs at all, confirming its role as a compute module rather than a graphics adapter. The Intel part also exposes graphics APIs, while the NVIDIA part lists N/A for DirectX, OpenGL, and Vulkan. Production status is active for both, with the Jetson T5000 releasing on 2025-08-26 and the Arc Pro B390 releasing on 2026-01-26. The Jetson T5000 succeeds Server Hopper and precedes Server Rubin, while the Arc Pro B390's predecessor is HD Graphics-WM.

FAQ

Q: Which GPU has higher FP32 performance?

A: The NVIDIA Jetson T5000 delivers 8.064 TFLOPS of FP32 compute, which is about 5% higher than the Intel Arc Pro B390's 7.680 TFLOPS.

Q: Which GPU has higher FP16 performance?

A: The Intel Arc Pro B390 delivers 15.36 TFLOPS of FP16 compute using a 2:1 ratio, while the NVIDIA Jetson T5000 delivers 8.064 TFLOPS at a 1:1 ratio. The Intel part is roughly 90% faster in half-precision throughput.

Q: How much memory does each GPU have?

A: The NVIDIA Jetson T5000 has 128 GB of LPDDR5X memory on a 256-bit bus with 273.2 GB/s bandwidth. The Intel Arc Pro B390 uses system-shared memory with system-dependent bandwidth, so its capacity and speed depend on the host system.

Q: Does either GPU support DirectX?

A: The Intel Arc Pro B390 supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The NVIDIA Jetson T5000 lists N/A for DirectX, OpenGL, and Vulkan, meaning no graphics API support is recorded for it.

Q: What is the TDP of each GPU?

A: The Intel Arc Pro B390 has an 80 W TDP. The NVIDIA Jetson T5000 has a 120 W TDP and lists a suggested PSU of 300 W.

Q: Do both GPUs have tensor cores?

A: The NVIDIA Jetson T5000 has 96 tensor cores. The Intel Arc Pro B390 does not list a tensor core count in its specifications.

Q: What process nodes are used?

A: The Intel Arc Pro B390 is built on a 3 nm process at Intel. The NVIDIA Jetson T5000 is built on a 5 nm process at TSMC.

The Verdict

The data points to two different design intents. The Intel Arc Pro B390 is a graphics-oriented integrated GPU with display outputs, full graphics API support, and a high boost clock of 2500 MHz. It leads in FP16 compute and pixel fill rate, and it fits into portable or compact systems where an IGP with system-shared memory is the only option. Its 80 W TDP and lack of dedicated memory make it a lower-power, platform-integrated solution.

The NVIDIA Jetson T5000 is a compute-oriented server module. It has no display outputs, no graphics APIs, and a 120 W TDP with a suggested 300 W PSU. It leads in FP32 compute, texture rate, shader count, ray tracing core count, and memory capacity with 128 GB of dedicated LPDDR5X. The 96 tensor cores and 8.064 TFLOPS of FP32 throughput position it for AI inference and data-center workloads that need large on-board memory and sustained single-precision compute.

Users who need a graphics adapter for a Panther Lake platform, with DirectX 12 Ultimate and Vulkan support, should choose the Arc Pro B390. Users who need a headless server accelerator with tensor cores and 128 GB of memory should choose the Jetson T5000. The Arc Pro B390 wins for FP16-heavy work and pixel-bound graphics; the Jetson T5000 wins for FP32-heavy compute and large-memory workloads. Neither product shows a benchmark score in the database, so the choice rests on the documented specifications above.

Specification Differences

| Specification | Intel Arc Pro B390 | NVIDIA Jetson T5000 |

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

| Architecture | Xe3-LPG | Blackwell |

| Process Node | 3 nm | 5 nm |

| Foundry | Intel | TSMC |

| Die Size | Unknown | 391 mm² |

| Base Clock | 300 MHz | 1386 MHz |

| Boost Clock | 2500 MHz | 1575 MHz |

| Memory Size | System Shared | 128 GB |

| Memory Type | System Shared | LPDDR5X |

| Memory Bus | System Shared | 256 bit |

| Memory Bandwidth | System Dependent | 273.2 GB/s |

| Shading Units | 1536 | 2560 |

| TMUs | 48 | 80 |

| ROPs | 24 | 32 |

| Ray Tracing Cores | 12 | 20 |

| Tensor Cores | Not listed | 96 |

| Pixel Rate | 60.00 GPixel/s | 50.40 GPixel/s |

| Texture Rate | 120.0 GTexel/s | 126.0 GTexel/s |

| FP32 | 7.680 TFLOPS | 8.064 TFLOPS |

| FP16 | 15.36 TFLOPS (2:1) | 8.064 TFLOPS (1:1) |

| TDP | 80 W | 120 W |

| Suggested PSU | None listed | 300 W |

| Bus Interface | IGP | PCIe 5.0 x8 |

| Display Outputs | Portable Device Dependent | No outputs |

| DirectX | 12 Ultimate (12_2) | N/A |

| OpenGL | 4.6 | N/A |

| Vulkan | 1.4 | N/A |

| Dimensions | Not listed | 87 mm x 100 mm x 15 mm |

| Release Date | 2026-01-26 | 2025-08-26 |

| Launch MSRP | None listed | 2,999 USD |

| Predecessor | HD Graphics-WM | Server Hopper |

| Successor | None listed | Server Rubin |