AMD Ryzen Z1 Extreme GPU vs Intel Arc Pro B390 Comparison

Head-to-Head Benchmarks

The recorded database contains no direct head-to-head benchmark scores for the AMD Ryzen Z1 Extreme GPU versus the Intel Arc Pro B390. Both entries list an empty benchmark array, and the win counters for each side sit at zero. Without measured performance deltas, the comparison rests entirely on architectural specifications and theoretical throughput figures rather than observed application results.

The raw compute ceilings are close but not identical. The AMD part delivers 8.294 TFLOPS of FP32 throughput, while the Intel part reaches 7.680 TFLOPS. That translates to an 8.0% advantage for the AMD silicon in single-precision floating-point work. The gap narrows in FP16, where AMD posts 16.59 TFLOPS (2:1) against Intel's 15.36 TFLOPS (2:1), a 7.4% edge for the Ryzen Z1 Extreme GPU.

Texture processing favors AMD as well. The RDNA 3.0 part sustains 129.6 GTexel/s, versus 120.0 GTexel/s for the Xe3-LPG part, a 7.5% margin. Pixel throughput tells the opposite story. AMD's 86.40 GPixel/s beats Intel's 60.00 GPixel/s by a substantial 36.0%, making the Ryzen Z1 Extreme GPU the stronger choice for fill-rate-bound scenarios such as high-resolution rasterization without heavy shading.

Neither GPU has recorded benchmark scores, so both sit at the 50th percentile among all GPUs in the database by default. The absence of measured data means the percentile ranking carries no comparative weight here; it is a placeholder value rather than a performance verdict.

Architecture Differences

The two chips come from different foundries and process nodes. AMD builds the Ryzen Z1 Extreme GPU on TSMC's 4 nm process, while Intel fabricates the Arc Pro B390 on its own 3 nm node. The AMD die, codenamed Phoenix, measures 178 mm² and packs 25,390 million transistors, yielding a transistor density of 142.6 million per square millimeter. Intel does not disclose transistor count or die size for the Panther Lake-based Arc Pro B390, and its density figure is absent from the database.

Architecturally, AMD uses RDNA 3.0, a graphics-focused design, whereas Intel employs Xe3-LPG, the low-power graphics variant of its Xe3 architecture. Both belong to the same generation tier in their respective lineups: AMD's is classified as a Console GPU, Intel's as Arc Graphics-WM (Panther Lake).

The shading engine layout diverges sharply. AMD fields 768 shading units, 48 texture mapping units, and 32 raster output units. Intel counters with 1,536 shading units, exactly double, but pairs them with 48 TMUs and only 24 ROPs. The result is a processor with more raw ALU lanes but fewer back-end pixel-writing units. Both GPUs carry 12 ray tracing cores, so ray-traced workloads see identical hardware resources on paper.

Memory architecture differs fundamentally. The AMD part uses 16 GB of LPDDR5 on a 64-bit bus, delivering 51.20 GB/s of bandwidth at 6.4 Gbps effective. The Intel part relies entirely on system shared memory, with its size, type, bus width, and bandwidth all listed as system dependent. That makes the Intel GPU's memory performance contingent on the host platform's configuration, a variable the database cannot quantify.

Clock behavior also separates the two. AMD runs a base clock of 800 MHz and boosts to 2700 MHz. Intel starts much lower at 300 MHz base but boosts to 2500 MHz. The Intel part's lower base clock suggests a more aggressive power-management profile, consistent with its IGP classification.

Power targets diverge considerably. AMD's TDP is 30 W, while Intel's is 80 W. Neither requires external power connectors, and neither lists a suggested PSU. Intel's slot width is IGP, meaning it resides on the processor package rather than a discrete board, while AMD's dimensions are recorded as 280 mm length, 111 mm height, and 21 mm width, a physical card footprint.

API support matches exactly. Both expose DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. Display output differs: AMD lists a single USB Type-C connector, while Intel's outputs are described as portable device dependent.

The production status for both is Active. AMD launched on 2023-06-12 with a launch MSRP of 699 USD. Intel's release date is 2026-01-26, and it has no recorded launch MSRP. Intel's predecessor is HD Graphics-WM; AMD's predecessor and successor fields are empty.

FAQ

Q: Which GPU has higher FP32 compute throughput?

A: The AMD Ryzen Z1 Extreme GPU delivers 8.294 TFLOPS, which is 8.0% higher than the Intel Arc Pro B390's 7.680 TFLOPS.

Q: How do the shading unit counts compare?

A: Intel's Arc Pro B390 has 1,536 shading units, exactly twice the 768 found in the AMD Ryzen Z1 Extreme GPU.

Q: Does either GPU use dedicated video memory?

A: Only AMD does. It has 16 GB of LPDDR5 on a 64-bit bus with 51.20 GB/s bandwidth. Intel's memory is system shared, with bandwidth listed as system dependent.

Q: What are the TDP figures for each?

A: AMD's TDP is 30 W, while Intel's is 80 W. Neither requires external power connectors.

Q: Do they support the same graphics APIs?

A: Yes. Both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Q: Which GPU has the higher pixel fill rate?

A: AMD's 86.40 GPixel/s exceeds Intel's 60.00 GPixel/s by 36.0%.

Specification Differences

| Field | AMD Ryzen Z1 Extreme GPU | Intel Arc Pro B390 |

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

| Architecture | RDNA 3.0 | Xe3-LPG |

| Process node | 4 nm | 3 nm |

| Foundry | TSMC | Intel |

| Transistors | 25,390 million | unknown |

| Die size | 178 mm² | unknown |

| Transistor density | 142.6M / mm² | null |

| Base clock | 800 MHz | 300 MHz |

| Boost clock | 2700 MHz | 2500 MHz |

| Memory size | 16 GB | System Shared |

| Memory type | LPDDR5 | System Shared |

| Memory bus width | 64 bit | System Shared |

| Memory bandwidth | 51.20 GB/s | System Dependent |

| Shading units | 768 | 1536 |

| ROPs | 32 | 24 |

| Pixel rate | 86.40 GPixel/s | 60.00 GPixel/s |

| Texture rate | 129.6 GTexel/s | 120.0 GTexel/s |

| FP32 | 8.294 TFLOPS | 7.680 TFLOPS |

| FP16 | 16.59 TFLOPS (2:1) | 15.36 TFLOPS (2:1) |

| TDP | 30 W | 80 W |

| Slot width | null | IGP |

| Bus interface | null | IGP |

| Display outputs | 1x USB Type-C | Portable Device Dependent |

| Dimensions | 280 mm x 111 mm x 21 mm | null |

| Release date | 2023-06-12 | 2026-01-26 |

| Launch MSRP | 699 USD | null |

| Predecessor | null | HD Graphics-WM |

Fields that match, including TMUs (48), RT cores (12), API list, production status, and the absence of tensor cores, are excluded from the table.

Where Each One Wins

The AMD Ryzen Z1 Extreme GPU wins in raw compute throughput. Its 8.294 TFLOPS FP32 output tops Intel's by 8.0%, and its 16.59 TFLOPS FP16 output leads by 7.4%. Applications that stress shader math, general-purpose GPGPU workloads, or FP16-heavy inference tasks will lean toward the AMD part based on these recorded figures.

AMD also takes the fill-rate categories. The 36.0% pixel-rate advantage (86.40 GPixel/s versus 60.00 GPixel/s) makes it the stronger candidate for scenarios where outputting fragments to a framebuffer dominates, such as high-resolution display rendering or post-processing passes. Its texture rate advantage of 7.5% (129.6 GTexel/s versus 120.0 GTexel/s) adds another layer of rasterization strength.

The power envelope favors AMD decisively. With a 30 W TDP against Intel's 80 W, the Ryzen Z1 Extreme GPU delivers higher theoretical throughput while drawing less than half the power budget. For thermally constrained platforms, notebooks, or handheld devices, that efficiency gap is the single largest differentiator in the database.

The Intel Arc Pro B390 wins on shading unit count, fielding 1,536 units versus 768. That doubling suggests headroom in workloads that scale with ALU width, even though the recorded FP32 throughput is lower, implying the Intel architecture does not translate its extra lanes into proportionally higher peak FLOPs at the listed clocks.

Intel also wins on process technology, using a 3 nm node from its own foundry compared to AMD's 4 nm TSMC process. The database does not record the Intel die size or transistor count, so the density comparison cannot be made, but the node advantage stands.

The discrete versus integrated distinction shapes use cases. AMD ships as a card with a 280 mm length, 111 mm height, and 21 mm width, with a single USB Type-C display output. Intel is an IGP with no physical dimensions, no dedicated memory, and display outputs that depend on the portable device. Systems requiring a self-contained graphics card with dedicated 16 GB of LPDDR5 will use the AMD part; systems integrating graphics on the processor die will use the Intel part.

The memory bandwidth split reinforces the separation. AMD's dedicated 51.20 GB/s is fixed and predictable, while Intel's system dependent bandwidth varies with the host platform. GPU workloads sensitive to memory bandwidth, particularly those with large working sets, will favor the fixed-bandwidth AMD solution.

Launch timing also differs: AMD released on 2023-06-12 with a 699 USD launch MSRP, while Intel's Arc Pro B390 arrives on 2026-01-26 with no recorded MSRP. The Intel part is the newer silicon, but the database contains no measured benchmarks for either GPU, so the performance ranking rests on theoretical peak rates and architectural features alone.