AMD Instinct MI350X vs Intel Arc A380M Comparison

FAQ

Q: What are the core differences between the AMD Instinct MI350X and the Intel Arc A380M?

A: The AMD Instinct MI350X is a server accelerator built on the CDNA 4.0 architecture, while the Intel Arc A380M is a mobile graphics module based on the Xe-HPG architecture. The MI350X uses a 3 nm process with an 185,000 million transistor count, whereas the A380M uses a 6 nm process with 7,200 million transistors.

Q: How do the memory subsystems compare?

A: The MI350X features 288 GB of HBM3e memory on an 8192-bit bus, delivering 8.19 TB/s of bandwidth. The A380M has 6 GB of GDDR6 memory on a 96-bit bus, providing 186.0 GB/s of bandwidth.

Q: Which GPU has a higher FP32 compute throughput?

A: The MI350X delivers 72.09 TFLOPS FP32 performance. The A380M provides 4.096 TFLOPS FP32. The MI350X is approximately 17.6 times higher in raw FP32 throughput.

Q: What are the power requirements?

A: The MI350X has a TDP of 1000 W and a suggested PSU of 1400 W. The A380M has a TDP of 35 W and has no suggested PSU listed.

Q: What form factors do these use?

A: The MI350X uses an OAM Module slot width and has no display outputs. The A380M uses an MXM Module slot width with display outputs described as "Portable Device Dependent."

Q: What API support does each provide?

A: The MI350X lists N/A for DirectX, OpenGL, and Vulkan. The A380M supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Architecture Differences

The AMD Instinct MI350X and Intel Arc A380M represent fundamentally different design goals. The MI350X uses AMD's CDNA 4.0 architecture, optimized for compute acceleration in data centers. The A380M uses Intel's Xe-HPG architecture, designed for graphics workloads in portable devices.

The manufacturing processes show a clear generation gap. The MI350X is fabricated on TSMC's 3 nm process, while the A380M uses TSMC's 6 nm process. This contributes to the massive difference in transistor counts: 185,000 million for the MI350X versus 7,200 million for the A380M. The die sizes reflect this disparity, with the MI350X measuring 2380 mm² compared to 157 mm² for the A380M. Transistor density also differs, with the MI350X achieving 77.7M transistors per mm² and the A380M reaching 45.9M per mm².

The compute resources are heavily skewed toward the AMD part. The MI350X contains 16,384 shading units and 1,024 TMUs, while the A380M has 1,024 shading units and 64 TMUs. The MI350X reports 0 ROPs, and the A380M has 32 ROPs. The A380M includes 8 ray tracing cores, while the MI350X lists no RT core count. Neither GPU lists tensor cores in the recorded data.

Memory architecture differs significantly between the designs. The MI350X uses 288 GB of HBM3e on an 8192-bit interface, targeting large-scale compute workloads that require massive memory capacity. The A380M uses 6 GB of GDDR6 on a 96-bit interface, suited for mobile graphics tasks. The MI350X's memory clock is 2000 MHz with 8 Gbps effective, while the A380M runs at 1937 MHz with 15.5 Gbps effective.

The MI350X's CDNA 4.0 architecture prioritizes FP32 and FP16 throughput equally at 72.09 TFLOPS each with a 1:1 ratio. The A380M's Xe-HPG architecture delivers 4.096 TFLOPS FP32 and 8.192 TFLOPS FP16 at a 2:1 ratio, showing a different approach to compute allocation.

Power delivery and physical design reflect their intended environments. The MI350X draws 1000 W TDP with a 1400 W suggested PSU and uses an OAM module form factor with no power connectors listed. The A380M draws 35 W TDP and uses an MXM-A (3.1) bus interface. The MI350X measures 102 mm in length and 165 mm in width, while the A380M has no recorded dimensions.

Head-to-Head Benchmarks

The recorded benchmark data shows no direct head-to-head benchmark entries, no wins for either GPU, and no average benchmark scores. Both parts share a 50th percentile rank against all GPUs in the database. The absence of direct measurements means the comparison relies on the architectural and specification data recorded for each part.

The FP32 compute results provide the clearest separation between the two. The MI350X delivers 72.09 TFLOPS, which is 17.6 times the 4.096 TFLOPS of the A380M. This gap indicates the MI350X is designed for workloads that demand sustained floating-point throughput, such as large-scale training or inference tasks.

Texture processing shows a similar pattern. The MI350X achieves a texture rate of 2,252.8 GTexel/s, while the A380M reaches 128.0 GTexel/s. The MI350X leads by a factor of 17.6 in this metric as well, consistent with its much larger TMU count of 1,024 versus 64.

Pixel rate presents a different outcome. The MI350X records 0 MPixel/s, while the A380M delivers 64.00 GPixel/s. This reflects the MI350X's lack of ROPs and its role as a compute accelerator rather than a rasterization-focused GPU. The A380M, with 32 ROPs, produces meaningful pixel throughput for graphics rendering.

Memory bandwidth shows the MI350X's dominance in data movement. The 8.19 TB/s bandwidth of the MI350X is 44 times the 186.0 GB/s of the A380M. The 8192-bit bus on the MI350X versus the 96-bit bus on the A380M drives this difference.

The performance profile indicates that neither GPU is positioned for the other's primary use case. The MI350X offers no display outputs and no graphics API support, confirming its compute-only orientation. The A380M provides full graphics API coverage and portable-device-dependent display outputs, confirming its mobile graphics role.

Specification Differences

The two GPUs differ across nearly every recorded specification field.

The process node differs: the MI350X uses 3 nm, the A380M uses 6 nm. Both use TSMC as the foundry. Transistor counts are 185,000 million for the MI350X and 7,200 million for the A380M. Die size is 2380 mm² for the MI350X and 157 mm² for the A380M. Transistor density is 77.7M per mm² versus 45.9M per mm².

Clock speeds show a mixed comparison. The MI350X has a base clock of 1000 MHz and a boost clock of 2200 MHz. The A380M has a higher base clock of 1550 MHz but a lower boost clock of 2000 MHz. Memory clocks are 2000 MHz with 8 Gbps effective for the MI350X and 1937 MHz with 15.5 Gbps effective for the A380M.

Memory specifications differ completely. The MI350X uses 288 GB of HBM3e with an 8192-bit bus and 8.19 TB/s bandwidth. The A380M uses 6 GB of GDDR6 with a 96-bit bus and 186.0 GB/s bandwidth.

Compute unit counts differ. The MI350X has 16,384 shading units, 1,024 TMUs, and 0 ROPs. The A380M has 1,024 shading units, 64 TMUs, and 32 ROPs. The A380M includes 8 RT cores; the MI350X lists none.

Rates and throughput values differ. The MI350X has a pixel rate of 0 MPixel/s and a texture rate of 2,252.8 GTexel/s. The A380M has a pixel rate of 64.00 GPixel/s and a texture rate of 128.0 GTexel/s. FP32 is 72.09 TFLOPS for the MI350X and 4.096 TFLOPS for the A380M. FP16 is 72.09 TFLOPS at 1:1 ratio for the MI350X and 8.192 TFLOPS at 2:1 ratio for the A380M.

Power figures contrast sharply. The MI350X has a TDP of 1000 W and a suggested PSU of 1400 W. The A380M has a TDP of 35 W and no suggested PSU recorded.

Form factors and interfaces differ. The MI350X uses an OAM Module slot width with a PCIe 5.0 x16 bus interface and no power connectors. The A380M uses an MXM Module slot width with an MXM-A (3.1) bus interface.

Display support differs. The MI350X has no display outputs. The A380M has portable-device-dependent display outputs. API support shows the MI350X with N/A across DirectX, OpenGL, and Vulkan, while the A380M supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Release dates differ, with the A380M released on 2023-01-23 and the MI350X on 2025-06-11. The MI350X lists "Radeon Instinct" as its predecessor, while the A380M lists no predecessor. The A380M has an active production status, while the MI350X has none recorded.

The Verdict

The recorded data positions the AMD Instinct MI350X and Intel Arc A380M in separate market segments with no meaningful overlap. The MI350X delivers 72.09 TFLOPS FP32, 288 GB of HBM3e memory, and 8.19 TB/s bandwidth, all within a 1000 W TDP envelope. The A380M provides 4.096 TFLOPS FP32, 6 GB of GDDR6, and 186.0 GB/s bandwidth at 35 W TDP.

Users requiring massive compute throughput for data center workloads should consider the MI350X. Its 17.6 times higher FP32 performance and 44 times higher memory bandwidth relative to the A380M make it suited for compute-heavy tasks. The absence of display outputs and graphics API support confirms its role as a compute accelerator rather than a graphics card.

Users needing mobile graphics capability should consider the A380M. It provides 64.00 GPixel/s pixel throughput, 8 ray tracing cores, and full DirectX 12 Ultimate support. Its 35 W TDP and MXM form factor suit portable device integration. The MI350X cannot serve this role because it has no display outputs and no graphics API support.

The 50th percentile ranking for both GPUs in the database indicates neither part has a measured performance advantage over the wider GPU population. Rather than a direct competition, these are complementary products for different applications. The data confirms the MI350X for compute density and the A380M for mobile graphics, with each excelling only within its own domain.