AMD Instinct MI350P vs Intel Arc B770 Comparison

FAQ

Q: What are the primary architectural differences between the AMD Instinct MI350P and the Intel Arc B770?

A: The AMD Instinct MI350P is built on CDNA 4.0 architecture using a 3 nm TSMC process, while the Intel Arc B770 uses Xe2-HPG architecture on a 5 nm TSMC process. The MI350P is an accelerator with no display outputs, whereas the Arc B770 is a graphics card with 1x HDMI 2.1a and 3x DisplayPort 2.1 outputs.

Q: How do the memory subsystems compare?

A: The MI350P features 144 GB of HBM3e memory on an 8192-bit bus, delivering 8.19 TB/s of bandwidth. The Arc B770 has 16 GB of GDDR6 on a 256-bit bus, providing 512.0 GB/s. The MI350P offers substantially higher capacity and bandwidth.

Q: What is the difference in compute performance?

A: The MI350P delivers 36.04 TFLOPS FP32 and 36.04 TFLOPS FP16 (1:1 ratio). The Arc B770 provides 19.66 TFLOPS FP32 and 39.32 TFLOPS FP16 (2:1 ratio). The MI350P leads in FP32, while the Arc B770 leads in FP16 throughput.

Q: How do power requirements differ?

A: The MI350P has a TDP of 600 W and requires a 1000 W suggested PSU with a single 16-pin connector. The Arc B770 has a TDP of 225 W and requires a 550 W suggested PSU with one 6-pin and one 8-pin connector.

Q: What API support does each product offer?

A: The MI350P has no DirectX, OpenGL, or Vulkan support. The Arc B770 supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Q: How do the two compare in transistor count and die size?

A: The MI350P uses 73,000 million transistors on a 1190 mm² die, achieving a transistor density of 61.3M per mm². The Arc B770 has an unknown transistor count on a 368 mm² die.

Architecture Differences

The AMD Instinct MI350P and Intel Arc B770 represent fundamentally different design philosophies. The MI350P is a compute accelerator built for data center workloads, while the Arc B770 is a consumer-oriented graphics card with full display capabilities.

The MI350P uses CDNA 4.0 architecture, a compute-focused design with no graphics output. It is manufactured on a 3 nm TSMC process with 73,000 million transistors on a 1190 mm² die. The transistor density reaches 61.3M per mm². The chip is designated MI350 128CU, indicating 128 compute units. The card has 8192 shading units and 512 texture mapping units, but no ROPs, resulting in a pixel rate of 0 MPixel/s. The texture rate is 1,126.4 GTexel/s.

The Arc B770 uses Xe2-HPG architecture from the Battlemage generation. It is manufactured on a 5 nm TSMC process with a 368 mm² die. The transistor count is not recorded in the database. The chip is designated BMG-G31. The card has 4096 shading units, 256 TMUs, and 128 ROPs, giving a pixel rate of 307.2 GPixel/s and a texture rate of 614.4 GTexel/s. It also includes 32 ray tracing cores, a feature entirely absent from the MI350P's compute-focused design.

Clock behavior differs significantly. The MI350P runs at a base clock of 1000 MHz and boosts to 2200 MHz. The Arc B770 operates at a higher base clock of 2100 MHz and boosts to 2400 MHz. Memory clocks are identical at 2000 MHz, but effective speeds differ: 8 Gbps for the MI350P and 16 Gbps for the Arc B770.

The MI350P uses HBM3e memory with 144 GB capacity on an 8192-bit bus. The Arc B770 uses GDDR6 with 16 GB on a 256-bit bus. The bandwidth gap is substantial: 8.19 TB/s versus 512.0 GB/s.

The MI350P connects via PCIe 5.0 x16, while the Arc B770 uses PCIe 4.0 x16. The MI350P has no display outputs; the Arc B770 provides 1x HDMI 2.1a and 3x DisplayPort 2.1. API support follows the same split: the MI350P has no DirectX, OpenGL, or Vulkan support, while the Arc B770 supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Both cards are dual-slot designs. The MI350P measures 267 mm in length, 111 mm in height, and 40 mm in width. The Arc B770 has no recorded dimensions in the database.

The Verdict

The data indicates two distinct products for different workloads. The AMD Instinct MI350P is an accelerator with no graphics output, targeting compute environments that require massive memory capacity and bandwidth. The Intel Arc B770 is a graphics card with full display and API support, targeting rendering and gaming workloads.

The MI350P delivers 36.04 TFLOPS FP32 performance, which is 83% higher than the Arc B770's 19.66 TFLOPS. Its memory subsystem provides 8.19 TB/s of bandwidth, which is 16 times the Arc B770's 512.0 GB/s, and its 144 GB capacity is 9 times larger than 16 GB. The 8192-bit bus width compared to 256-bit further confirms the accelerator positioning.

The Arc B770 counters in FP16 compute with 39.32 TFLOPS versus 36.04 TFLOPS for the MI350P, a 9% advantage. It also offers 32 ray tracing cores, which the MI350P lacks entirely. The Arc B770 includes standard API support: DirectX 12 Ultimate, OpenGL 4.6, and Vulkan 1.4. The MI350P has none of these APIs.

Power requirements reflect the performance split. The MI350P consumes 600 W and needs a 1000 W suggested PSU. The Arc B770 consumes 225 W and needs a 550 W suggested PSU, representing a 62.5% reduction in TDP.

Release dates place the Arc B770 at the end of 2025 and the MI350P in early May 2026. Both products have a database percentile of 50, and no benchmark scores are recorded for either.

The choice depends on the workload. Compute environments needing large memory and bandwidth should use the MI350P. Graphics applications requiring display output, ray tracing, and standard API support should use the Arc B770.

Specification Differences

| Specification | AMD Instinct MI350P | Intel Arc B770 |

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

| Architecture | CDNA 4.0 | Xe2-HPG |

| Generation | Instinct (MIx) | Battlemage (Arc 7) |

| Process node | 3 nm | 5 nm |

| Die size | 1190 mm² | 368 mm² |

| Transistors | 73,000 million | unknown |

| Transistor density | 61.3M / mm² | null |

| Base clock | 1000 MHz | 2100 MHz |

| Boost clock | 2200 MHz | 2400 MHz |

| Memory speed | 2000 MHz 8 Gbps effective | 2000 MHz 16 Gbps effective |

| Memory size | 144 GB | 16 GB |

| Memory type | HBM3e | GDDR6 |

| Memory bus width | 8192 bit | 256 bit |

| Memory bandwidth | 8.19 TB/s | 512.0 GB/s |

| Shading units | 8192 | 4096 |

| TMUs | 512 | 256 |

| ROPs | 0 | 128 |

| Ray tracing cores | null | 32 |

| Pixel rate | 0 MPixel/s | 307.2 GPixel/s |

| Texture rate | 1,126.4 GTexel/s | 614.4 GTexel/s |

| FP32 performance | 36.04 TFLOPS | 19.66 TFLOPS |

| FP16 performance | 36.04 TFLOPS (1:1) | 39.32 TFLOPS (2:1) |

| TDP | 600 W | 225 W |

| Power connectors | 1x 16-pin | 1x 6-pin + 1x 8-pin |

| Suggested PSU | 1000 W | 550 W |

| Bus interface | PCIe 5.0 x16 | PCIe 4.0 x16 |

| Display outputs | No outputs | 1x HDMI 2.1a3x DisplayPort 2.1 |

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

| OpenGL support | N/A | 4.6 |

| Vulkan support | N/A | 1.4 |

| Release date | 2026-05-06 | 2025-12-31 |

| Predecessor | Radeon Instinct | Alchemist |

Head-to-Head Benchmarks

The database records no benchmark scores for either product, and the head-to-head benchmark list is empty. Both cards hold a 50th percentile ranking among all GPUs, with an average benchmark score of 0. The nearest rivals lists are empty for both products.

The specification data, however, allows direct comparison of theoretical compute limits. The MI350P delivers 36.04 TFLOPS FP32, which is 1.83 times the Arc B770's 19.66 TFLOPS. This represents an 83% advantage in single-precision floating-point throughput. The MI350P's 8192 shading units are double the Arc B770's 4096, and its 512 TMUs are double the Arc B770's 256.

Texture throughput follows the same pattern. The MI350P achieves 1,126.4 GTexel/s, which is 1.83 times the Arc B770's 614.4 GTexel/s. The Arc B770 counters with 128 ROPs and a pixel rate of 307.2 GPixel/s, while the MI350P has no ROPs and a pixel rate of 0 MPixel/s.

Memory bandwidth is the largest gap. The MI350P's 8.19 TB/s is exactly 16 times the Arc B770's 512.0 GB/s. The capacity difference is also dramatic: 144 GB versus 16 GB, or 9 times more memory. The bus width difference of 8192 bits versus 256 bits explains the bandwidth disparity.

The Arc B770 wins in FP16 throughput. Its 39.32 TFLOPS FP16 exceeds the MI350P's 36.04 TFLOPS by 9.1%. This advantage comes from the 2:1 FP16 ratio on the Arc B770 versus the 1:1 ratio on the MI350P. The MI350P's FP16 performance matches its FP32 performance, while the Arc B770's FP16 is double its FP32.

Clock speeds favor the Arc B770. The base clock of 2100 MHz is 2.1 times the MI350P's 1000 MHz. The boost clock of 2400 MHz is 9.1% higher than the MI350P's 2200 MHz. These higher clocks partially compensate for the Arc B770's lower compute unit count.

Ray tracing is exclusive to the Arc B770 with 32 dedicated cores. The MI350P has no ray tracing hardware recorded. The Arc B770 also provides standard graphics APIs, while the MI350P has no API support.

Power efficiency favors the Arc B770. The MI350P consumes 600 W to deliver 36.04 TFLOPS FP32, yielding about 60 GFLOPS per watt. The Arc B770 consumes 225 W to deliver 19.66 TFLOPS FP32, yielding about 87 GFLOPS per watt. The Arc B770 achieves roughly 45% higher FP32 efficiency per watt.

The MI350P requires a 1000 W suggested PSU, while the Arc B770 requires 550 W. The MI350P uses one 16-pin connector; the Arc B770 uses one 6-pin and one 8-pin connector. Both are dual-slot designs.

Release timing shows the Arc B770 launched on 2025-12-31 and the MI350P on 2026-05-06. The MI350P follows the Radeon Instinct line, while the Arc B770 follows Intel's Alchemist generation.