AMD Instinct MI350X vs Intel Arc Pro B65 Comparison

Head-to-Head Benchmarks

The database contains no recorded benchmark scores for either the AMD Instinct MI350X or the Intel Arc Pro B65. Both cards hold an average benchmark score of 0 and sit at the 50th percentile among all GPUs in the database. With zero wins recorded for either side and no head-to-head entries, the performance comparison must rely entirely on architectural specifications and theoretical throughput calculations rather than measured application results.

The raw compute figures reveal a dramatic separation. The Instinct MI350X delivers 72.09 TFLOPS of FP32 throughput, which is 5.9 times the 12.29 TFLOPS produced by the Arc Pro B65. In FP16 workloads, the gap narrows slightly in relative terms because the Intel card executes FP16 at a 2:1 rate, reaching 24.58 TFLOPS, while the AMD accelerator sustains its 72.09 TFLOPS at a 1:1 ratio. The MI350X still leads by a factor of 2.9 in FP16, but the Intel card's doubled FP16 rate means it processes half-precision data more efficiently relative to its own FP32 capability.

Texture throughput follows a similar pattern. The MI350X achieves 2,252.8 GTexel/s against 384.0 GTexel/s for the Arc Pro B65, a 5.9-fold advantage that directly reflects the disparity in texture mapping units, 1,024 versus 160. Pixel throughput tells a different story entirely, the AMD card records 0 MPixel/s because it has no ROPs, while the Intel card produces 192.0 GPixel/s from its 80 raster output units. This makes the Arc Pro B65 the only one of the two capable of traditional rasterized frame output.

Memory bandwidth amplifies the compute gap. The MI350X accesses 8.19 TB/s through an 8192-bit HBM3e interface, compared to 608.0 GB/s across a 256-bit GDDR6 bus for the Arc Pro B65. That is a 13.5-fold bandwidth advantage, which matters enormously for data-intensive workloads such as large language model inference or scientific simulation. The capacity differential is equally stark: 288 GB versus 32 GB, a 9-fold difference in on-board memory footprint.

Architecture Differences

The two GPUs share a foundry, TSMC, but employ different process nodes. The Instinct MI350X uses a 3 nm node, while the Arc Pro B65 uses 5 nm. Transistor counts differ by nearly an order of magnitude. The AMD chip packs 185,000 million transistors on a 2380 mm² die, yielding a density of 77.7 million transistors per square millimeter. The Intel die contains 19,600 million transistors across 272 mm², a density of 72.1 million per square millimeter. The AMD part is therefore both substantially larger and slightly denser, an 8.7-fold transistor advantage overall.

Architecturally, the MI350X uses CDNA 4.0, AMD's compute-optimized design with 16,384 shading units and 1,024 TMUs but zero ROPs, reflecting its purpose as an accelerator with no display output. The Arc Pro B65 uses Xe2-HPG, Intel's Battlemage architecture for professional graphics, with 2,560 shading units, 160 TMUs, 80 ROPs, and 20 ray tracing cores. The presence of ray tracing hardware on the Intel card and its complete absence on the AMD card marks a fundamental design split: one targets rasterized and ray-traced visualization, the other targets pure compute.

Memory technologies diverge completely. The MI350X uses HBM3e across a 8192-bit bus, while the Arc Pro B65 uses GDDR6 across a 256-bit bus. Clock behavior also contrasts sharply. The AMD card runs a 1000 MHz base and 2200 MHz boost, while the Intel card operates at a flat 2400 MHz for both base and boost. Memory clocks differ as well, 2000 MHz with 8 Gbps effective on the AMD side versus 2375 MHz with 19 Gbps effective on the Intel side, though the HBM3e's massive bus width dwarfs the GDDR6's higher per-pin data rate.

API support separates the two completely. The MI350X lists N/A for DirectX, OpenGL, and Vulkan, confirming it is not a graphics card in the conventional sense. The Arc Pro B65 supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, making it fully compatible with modern graphics applications and game engines. Power delivery differs accordingly: the AMD module draws a 1000 W TDP with no power connectors, as it is an OAM module designed for server trays, while the Intel card is a dual-slot board with a single 8-pin connector and a 200 W TDP. The suggested power supply ratings reflect this, 1400 W for the AMD accelerator versus 550 W for the Intel card.

Where Each One Wins

The Instinct MI350X wins in every metric that matters for high-performance computing. Its FP32 throughput of 72.09 TFLOPS makes it suitable for large-scale scientific computation, and its FP16 performance at the same 72.09 TFLOPS with a 1:1 ratio indicates strong half-precision capability, often used in AI training and inference. The 288 GB HBM3e pool with 8.19 TB/s bandwidth allows models and datasets to reside entirely in GPU memory, eliminating the need for constant host-device transfers. The 1024 TMUs and 2,252.8 GTexel/s texture rate support heavy texture sampling in compute shaders, and the PCIe 5.0 x16 interface provides modern host connectivity. The 3 nm process node and 185,000 million transistor count give it the raw silicon resources for massive parallel workloads.

The Arc Pro B65 wins in every metric that matters for conventional graphics work. Its 80 ROPs and 192.0 GPixel/s pixel fill rate enable actual frame rendering, and its 20 ray tracing cores bring hardware-accelerated ray tracing, a feature entirely absent from the AMD accelerator. The 4x DisplayPort 2.1 outputs allow direct connection to monitors, while the MI350X has no display outputs at all. The Intel card's 12.29 TFLOPS FP32 and 24.58 TFLOPS FP16 are modest by comparison but still represent capable compute for a professional graphics card. Its 32 GB GDDR6 memory with 608.0 GB/s bandwidth supports large textures and complex scenes, and its 2400 MHz sustained boost clock means consistent performance without thermal throttling concerns at its 200 W TDP. The dual-slot form factor and single 8-pin power connector allow installation in standard workstation chassis, unlike the OAM module format of the AMD part.

The 5 nm process node and 72.1 million transistors per square millimeter density show the Arc Pro B65 is efficient for its size, but the 272 mm² die is a fraction of the 2380 mm² MI350X. The Intel card's 2560 shading units are 6.4 times fewer than the AMD card's 16384, and its 160 TMUs are 6.4 times fewer. The FP32 gap of 5.9 times mirrors this ratio precisely, confirming that the architectural scaling is consistent across compute units.

Specification Differences

| Specification | AMD Instinct MI350X | Intel Arc Pro B65 |

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

| Architecture | CDNA 4.0 | Xe2-HPG |

| Generation | Instinct (MIx) | Battlemage (Pro Series) |

| Process Node | 3 nm | 5 nm |

| Transistors | 185,000 million | 19,600 million |

| Die Size | 2380 mm² | 272 mm² |

| Transistor Density | 77.7M / mm² | 72.1M / mm² |

| Base Clock | 1000 MHz | 2400 MHz |

| Boost Clock | 2200 MHz | 2400 MHz |

| Memory Size | 288 GB | 32 GB |

| Memory Type | HBM3e | GDDR6 |

| Memory Bus Width | 8192 bit | 256 bit |

| Memory Bandwidth | 8.19 TB/s | 608.0 GB/s |

| Memory Clock | 2000 MHz, 8 Gbps effective | 2375 MHz, 19 Gbps effective |

| Shading Units | 16384 | 2560 |

| TMUs | 1024 | 160 |

| ROPs | 0 | 80 |

| Ray Tracing Cores | N/A | 20 |

| Pixel Rate | 0 MPixel/s | 192.0 GPixel/s |

| Texture Rate | 2,252.8 GTexel/s | 384.0 GTexel/s |

| FP32 | 72.09 TFLOPS | 12.29 TFLOPS |

| FP16 | 72.09 TFLOPS (1:1) | 24.58 TFLOPS (2:1) |

| TDP | 1000 W | 200 W |

| Slot Width | OAM Module | Dual-slot |

| Power Connectors | None | 1x 8-pin |

| Suggested PSU | 1400 W | 550 W |

| Bus Interface | PCIe 5.0 x16 | PCIe 5.0 x16 |

| Display Outputs | No outputs | 4x DisplayPort 2.1 |

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

| OpenGL Support | N/A | 4.6 |

| Vulkan Support | N/A | 1.4 |

| Dimensions | 102 mm length, 165 mm width | Not recorded |

| Release Date | 2025-06-11 | 2026-03-31 |

| Production Status | Not recorded | Active |

The bus interface matches at PCIe 5.0 x16 for both. The AMD card measures 102 mm in length and 165 mm in width, while the Intel card's dimensions are not recorded. The MI350X lists a predecessor, Radeon Instinct, while the Arc Pro B65 has no predecessor recorded. Neither card has a recorded launch MSRP.

FAQ

Q: Which GPU has higher FP32 compute throughput?

A: The AMD Instinct MI350X delivers 72.09 TFLOPS of FP32, which is 5.9 times the 12.29 TFLOPS produced by the Intel Arc Pro B65.

Q: Does the AMD Instinct MI350X support ray tracing?

A: No. The MI350X has no ray tracing cores recorded, and its API support for DirectX, OpenGL, and Vulkan is listed as N/A. The Intel Arc Pro B65 includes 20 ray tracing cores and supports DirectX 12 Ultimate (12_2).

Q: What memory configuration does each card use?

A: The AMD Instinct MI350X uses 288 GB of HBM3e across an 8192-bit bus with 8.19 TB/s bandwidth. The Intel Arc Pro B65 uses 32 GB of GDDR6 across a 256-bit bus with 608.0 GB/s bandwidth.

Q: Can the AMD Instinct MI350X output video to displays?

A: No. The MI350X has no display outputs, while the Intel Arc Pro B65 includes 4x DisplayPort 2.1 outputs.

Q: What are the power requirements for each card?

A: The AMD Instinct MI350X has a 1000 W TDP with no power connectors, as it is an OAM module, and requires a 1400 W suggested power supply. The Intel Arc Pro B65 has a 200 W TDP with one 8-pin connector and a 550 W suggested power supply.

Q: Which card has a higher transistor density?

A: The AMD Instinct MI350X achieves 77.7 million transistors per square millimeter on a 3 nm process, slightly higher than the Intel Arc Pro B65's 72.1 million per square millimeter on a 5 nm process. The AMD card's total transistor count is 185,000 million versus 19,600 million for the Intel card.