AMD Instinct MI455X vs Intel Arc B390 Comparison

Where Each One Wins

The AMD Instinct MI455X and Intel Arc B390 occupy entirely separate performance domains, as the recorded data shows no shared benchmark entries. The MI455X is an accelerator-class product with 32,768 shading units, 1,024 texture mapping units, and 157.3 TFLOPS of FP32 compute. The Arc B390 is an integrated graphics processor with 1,536 shading units, 48 TMUs, and 7.680 TFLOPS of FP32 compute. The MI455X delivers 2,457.6 GTexel/s of texture rate versus 120.0 GTexel/s for the Arc B390, a 20.48x difference in texture throughput. Pixel rate tells a different story: the MI455X records 0 MPixel/s because it has no ROPs, while the Arc B390 posts 60.00 GPixel/s with 24 ROPs. The MI455X cannot drive displays, whereas the Arc B390 provides portable-device-dependent display outputs.

The use-case split is sharp. The MI455X targets compute-heavy workloads where texture throughput and FP32/FP16 performance dominate. Its 432 GB of HBM4 memory with a 24,576-bit bus and 23.3 TB/s bandwidth positions it for large data sets. The Arc B390 targets conventional graphics rendering with a full DirectX 12 Ultimate feature set, OpenGL 4.6, and Vulkan 1.4 support. The MI455X lists no graphics API support at all. The Arc B390 includes 12 ray tracing cores; the MI455X lists no ray tracing cores. The MI455X wins every compute and memory bandwidth category. The Arc B390 wins every graphics rendering category, including the only actual benchmark score in the database.

Benchmark results indicate the Arc B390 has a recorded 3DMark Steel Nomad DX12 score of 1,482. The MI455X has no recorded benchmark scores and an average benchmark score of zero. The percentile data reinforces the divide: the Arc B390 sits at the 9th percentile of all GPUs, while the MI455X sits at the 50th percentile, though that percentile reflects no measured performance data. The MI455X is a server accelerator with no display outputs and no consumer API support. The Arc B390 is an integrated part of the Panther Lake processor generation, designed to render frames on portable devices.

Architecture Differences

The two parts come from different foundries and process nodes. The MI455X uses a 2 nm TSMC process with 320,000 million transistors on a 2,990 mm² die, yielding a transistor density of 107.0M per mm². The Arc B390 uses Intel's 3 nm process with unknown transistor count and die size. The MI455X is built on CDNA 5.0 architecture under the Instinct (MIx) generation, using the MI450 256CU chip. The Arc B390 uses Xe3-LPG architecture under the Arc Graphics-M (Panther Lake) generation, built around the Panther Lake chip.

Memory architecture differs fundamentally. The MI455X has 432 GB of HBM4 memory on a 24,576-bit bus with 23.3 TB/s bandwidth and a 1900 MHz memory clock running at 7.6 Gbps effective. The Arc B390 uses system shared memory with system-dependent bandwidth and no dedicated memory clock. The MI455X draws 2,300 W with a suggested power supply of 2,700 W and uses an EAM Module slot with no power connectors of its own. The Arc B390 draws 80 W, is an IGP with no power connectors, and has no suggested PSU rating.

Clock behavior also diverges. The MI455X has a 1000 MHz base clock and 2400 MHz boost. The Arc B390 has a 300 MHz base clock and 2500 MHz boost. The MI455X reaches a higher boost clock by 100 MHz despite a much larger thermal envelope. FP16 processing differs in ratio: the MI455X delivers 157.3 TFLOPS at 1:1 with FP32, while the Arc B390 delivers 15.36 TFLOPS at 2:1, which is double its FP32 rate. The MI455X has no ROPs, no pixel rate, and no display outputs. The Arc B390 has 24 ROPs and a 60.00 GPixel/s pixel rate. The MI455X uses PCIe 6.0 x16 for host connectivity; the Arc B390 uses an IGP bus interface. The MI455X lists no DirectX, OpenGL, or Vulkan support. The Arc B390 lists DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Release timing places the Arc B390 first, with a release date of 2026-01-26, and the MI455X second, with a release date of 2026-07-22. The Arc B390 has an active production status; the MI455X has no recorded production status. The MI455X lists Radeon Instinct as its predecessor. The Arc B390 lists no predecessor. Neither part has a listed successor or launch MSRP.

Head-to-Head Benchmarks

The database contains no direct head-to-head benchmark entries between these two products. The only recorded benchmark belongs to the Arc B390: a 3DMark Steel Nomad DX12 score of 1,482. The nearest rivals for that score are all NVIDIA parts. The GT 520MX scores 1,463, a 1.3% delta. The GeForce 800M scores 1,460, a 1.5% delta. The GT 625 OEM scores 1,446, a 2.5% delta. The GT 710 scores 1,443, a 2.7% delta. The Arc B390 leads all four by small margins, between 1.3% and 2.7%.

The MI455X has no benchmark scores and no nearest rivals in the database. Its average benchmark score is zero. Its 50th percentile ranking against all GPUs is therefore unverified by any measured workload. The recorded data cannot confirm how the MI455X would perform in any graphics or compute test.

Compute figures provide the only cross-product comparison. The MI455X FP32 throughput of 157.3 TFLOPS is 20.48x the Arc B390's 7.680 TFLOPS. The MI455X FP16 throughput of 157.3 TFLOPS is 10.24x the Arc B390's 15.36 TFLOPS at 2:1. Texture rate comparison shows 2,457.6 GTexel/s versus 120.0 GTexel/s, again a 20.48x gap. Memory bandwidth comparison is extreme: 23.3 TB/s versus system dependent, with the MI455X using dedicated HBM4 and the Arc B390 relying on shared system memory.

The Arc B390's pixel rate of 60.00 GPixel/s has no MI455X counterpart because the MI455X lists 0 MPixel/s. The MI455X has no ROPs to generate pixels. The Arc B390's 24 ROPs and 12 ray tracing cores are the only graphics-specific hardware in either part. The MI455X uses its 1,024 TMUs for texture work, while the Arc B390 uses 48 TMUs. Both parts share a 2,500 MHz boost clock range, with the MI455X at 2,400 MHz and the Arc B390 at 2,500 MHz.

FAQ

Q: Which product has better raw compute performance?

A: The MI455X delivers 157.3 TFLOPS of FP32 and FP16 (1:1), compared to 7.680 TFLOPS FP32 and 15.36 TFLOPS FP16 (2:1) for the Arc B390. The MI455X is 20.48x faster in FP32 and 10.24x faster in FP16.

Q: Does the MI455X support graphics APIs?

A: No. The MI455X lists N/A for DirectX, OpenGL, and Vulkan, and it has no display outputs. The Arc B390 supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Q: What is the Arc B390's recorded benchmark score?

A: The Arc B390 scores 1,482 in 3DMark Steel Nomad DX12. Its nearest rival, the NVIDIA GeForce GT 520MX, scores 1,463, putting the Arc B390 1.3% ahead.

Q: How much memory does each product have?

A: The MI455X has 432 GB of HBM4 memory on a 24,576-bit bus with 23.3 TB/s bandwidth. The Arc B390 uses system shared memory with system-dependent bandwidth.

Q: What process nodes do the two products use?

A: The MI455X uses a 2 nm TSMC process with 320,000 million transistors on a 2,990 mm² die. The Arc B390 uses Intel's 3 nm process with unknown transistor count and die size.

Q: Which product has a higher boost clock?

A: The Arc B390 boosts to 2,500 MHz, while the MI455X boosts to 2,400 MHz. The MI455X has a higher base clock at 1,000 MHz versus 300 MHz for the Arc B390.

The Verdict

The data describes two products with no functional overlap. The MI455X is a 2,300 W accelerator with 32,768 shading units, 432 GB of HBM4, 23.3 TB/s bandwidth, and 157.3 TFLOPS of FP32 compute. It has no ROPs, no pixel rate, no display outputs, and no graphics API support. It is built for compute workloads that can use its 2,457.6 GTexel/s texture rate and 1:1 FP16 throughput. Its 50th percentile ranking against all GPUs carries no benchmark verification because the database contains no measured scores for it.

The Arc B390 is an 80 W integrated graphics processor with 1,536 shading units, 24 ROPs, 12 ray tracing cores, and a 60.00 GPixel/s pixel rate. It supports DirectX 12 Ultimate, OpenGL 4.6, and Vulkan 1.4, and it has a verified 3DMark Steel Nomad DX12 score of 1,482, placing it 1.3% to 2.7% ahead of four NVIDIA rivals in the 9th percentile of all GPUs. Its system shared memory and system-dependent bandwidth limit memory performance, but its 2,500 MHz boost clock and portable-device-dependent display outputs suit integrated graphics duty.

The MI455X wins on compute density, memory capacity, memory bandwidth, and texture throughput. The Arc B390 wins on graphics rendering, API support, pixel output, and ray tracing. The MI455X requires a 2,700 W suggested power supply and uses an EAM Module slot with PCIe 6.0 x16. The Arc B390 uses no external power connectors and attaches via IGP. The MI455X is a data-center compute module. The Arc B390 is a client integrated GPU. The database shows no scenario where both would be candidates for the same workload.

Specification Differences

| Field | AMD Instinct MI455X | Intel Arc B390 |

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

| Architecture | CDNA 5.0 | Xe3-LPG |

| Process Node | 2 nm | 3 nm |

| Foundry | TSMC | Intel |

| Transistors | 320,000 million | unknown |

| Die Size | 2990 mm² | unknown |

| Base Clock | 1000 MHz | 300 MHz |

| Boost Clock | 2400 MHz | 2500 MHz |

| Memory Size | 432 GB | System Shared |

| Memory Type | HBM4 | System Shared |

| Memory Bus | 24576 bit | System Shared |

| Memory Bandwidth | 23.3 TB/s | System Dependent |

| Memory Clock | 1900 MHz 7.6 Gbps effective | none |

| Shading Units | 32768 | 1536 |

| TMUs | 1024 | 48 |

| ROPs | 0 | 24 |

| Ray Tracing Cores | none | 12 |

| Pixel Rate | 0 MPixel/s | 60.00 GPixel/s |

| Texture Rate | 2,457.6 GTexel/s | 120.0 GTexel/s |

| FP32 | 157.3 TFLOPS | 7.680 TFLOPS |

| FP16 | 157.3 TFLOPS (1:1) | 15.36 TFLOPS (2:1) |

| TDP | 2300 W | 80 W |

| Slot Width | EAM Module | IGP |

| Power Connectors | None | None |

| Suggested PSU | 2700 W | none |

| Bus Interface | PCIe 6.0 x16 | IGP |

| Display Outputs | No outputs | Portable Device Dependent |

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

| OpenGL | N/A | 4.6 |

| Vulkan | N/A | 1.4 |

| Production Status | none | Active |

| Release Date | 2026-07-22 | 2026-01-26 |

| Predecessor | Radeon Instinct | none |

| Avg Benchmark Score | 0 | 1482 |

| Percentile vs All GPUs | 50 | 9 |