AMD Instinct MI350X vs Intel Arc A380E Comparison

Head-to-Head Benchmarks

The database contains no recorded benchmark scores for either the AMD Instinct MI350X or the Intel Arc A380E. Both GPUs hold a percentile rank of 50 against all GPUs in the database, with an average benchmark score of 0. The head-to-head benchmark list is empty, and neither product registers wins in any measured workload. Consequently, direct performance comparisons must be derived from the recorded specifications, which reveal two products with fundamentally different design goals.

The AMD Instinct MI350X delivers 72.09 TFLOPS of FP32 compute and 72.09 TFLOPS of FP16 (1:1 ratio). The Intel Arc A380E provides 4.096 TFLOPS of FP32 and 8.192 TFLOPS of FP16 (2:1 ratio). In raw FP32 throughput, the MI350X is 17.6 times higher. For FP16, the MI350X achieves 8.8 times the throughput. These ratios come directly from the recorded peak compute figures, not from any synthesized benchmark. The texture rate also diverges sharply: the MI350X processes 2,252.8 GTexel/s versus 128.0 GTexel/s for the A380E, a 17.6-fold advantage. Pixel rate is the one area where the Intel part leads, as the MI350X records 0 MPixel/s while the A380E reaches 64.00 GPixel/s. This reflects the MI350X having zero raster output units, whereas the A380E includes 32 ROPs.

Memory bandwidth tells a similar story. The MI350X accesses 288 GB of HBM3e across an 8192-bit bus, yielding 8.19 TB/s of bandwidth. The A380E uses 6 GB of GDDR6 on a 96-bit bus, producing 186.0 GB/s. The MI350X bandwidth is 44.0 times higher. Clock behavior also differs: the MI350X runs at a 1000 MHz base and 2200 MHz boost, while the A380E operates at a fixed 2000 MHz for both base and boost. The A380E memory clock is 1937 MHz with 15.5 Gbps effective data rate, slightly lower in base clock but with a higher effective memory rate per pin due to the GDDR6 interface.

Neither GPU has a recorded launch MSRP, so no price-based comparison is possible. The data shows two accelerators with opposite design philosophies: one optimized for massive parallel compute with no display output, the other for embedded graphics with full display support. The absence of benchmark data means all conclusions here rest on the specification sheet, which is unambiguous about the compute hierarchy.

FAQ

Q: Which GPU has higher FP32 performance?

A: The AMD Instinct MI350X records 72.09 TFLOPS of FP32, while the Intel Arc A380E records 4.096 TFLOPS. The MI350X delivers 17.6 times the FP32 throughput.

Q: Does the Intel Arc A380E support graphics APIs?

A: Yes. The A380E lists DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4 support. The MI350X records N/A for DirectX, OpenGL, and Vulkan, and has no display outputs.

Q: What memory configurations do the two GPUs use?

A: The MI350X uses 288 GB of HBM3e with an 8192-bit bus and 8.19 TB/s bandwidth. The A380E uses 6 GB of GDDR6 with a 96-bit bus and 186.0 GB/s bandwidth.

Q: How do the transistor counts compare?

A: The MI350X contains 185,000 million transistors on a 2380 mm² die, while the A380E contains 7,200 million transistors on a 157 mm² die. The MI350X has roughly 25.7 times more transistors.

Q: Are both GPUs currently in production?

A: The MI350X has no recorded production status. The A380E is marked as end-of-life, with its successor listed as Battlemage.

Q: What is the power requirement difference?

A: The MI350X has a TDP of 1000 W with a suggested PSU of 1400 W. The A380E has a TDP of 75 W with a suggested PSU of 250 W.

Architecture Differences

The AMD Instinct MI350X uses the CDNA 4.0 architecture on a 3 nm TSMC process, while the Intel Arc A380E uses the Xe-HPG architecture on a 6 nm TSMC process. The MI350X chip is designated MI350 256CU, indicating 256 compute units. The A380E chip is DG2-128. The MI350X belongs to the Instinct (MIx) generation, whereas the A380E belongs to the Alchemist (Arc 3) generation. These architectural generations target different workloads: CDNA 4.0 is optimized for compute acceleration, while Xe-HPG is designed for graphics rendering.

The MI350X has 16,384 shading units, 1,024 TMUs, and 0 ROPs. It records no RT cores or tensor cores. The A380E has 1,024 shading units, 64 TMUs, and 32 ROPs, plus 8 RT cores. The MI350X has no display outputs, while the A380E provides 4x DisplayPort 2.0. The MI350X uses an OAM Module slot width with no power connectors, whereas the A380E is a single-slot card also without power connectors. The bus interfaces differ: PCIe 5.0 x16 for the MI350X versus PCIe 4.0 x8 for the A380E.

The API support gap is absolute. The MI350X records N/A for DirectX, OpenGL, and Vulkan, confirming it is not a graphics-oriented product. The A380E fully supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, making it suitable for rendering workloads. The MI350X also records 0 MPixel/s pixel rate, consistent with having no ROPs, while the A380E reaches 64.00 GPixel/s. The FP16 ratio differs: the MI350X achieves 72.09 TFLOPS at a 1:1 ratio with FP32, while the A380E achieves 8.192 TFLOPS at a 2:1 ratio, indicating different compute precision strategies.

Specification Differences

The two GPUs differ across nearly every recorded specification. Process node: 3 nm for the MI350X versus 6 nm for the A380E. Transistors: 185,000 million versus 7,200 million. Die size: 2380 mm² versus 157 mm². Transistor density: 77.7M per mm² versus 45.9M per mm². Base clock: 1000 MHz versus 2000 MHz. Boost clock: 2200 MHz versus 2000 MHz. Memory clock: 2000 MHz (8 Gbps effective) versus 1937 MHz (15.5 Gbps effective).

Memory size: 288 GB versus 6 GB. Memory type: HBM3e versus GDDR6. Bus width: 8192 bit versus 96 bit. Bandwidth: 8.19 TB/s versus 186.0 GB/s. Shading units: 16,384 versus 1,024. TMUs: 1,024 versus 64. ROPs: 0 versus 32. RT cores: not listed versus 8. Pixel rate: 0 MPixel/s versus 64.00 GPixel/s. Texture rate: 2,252.8 GTexel/s versus 128.0 GTexel/s. FP32: 72.09 TFLOPS versus 4.096 TFLOPS. FP16: 72.09 TFLOPS versus 8.192 TFLOPS.

TDP: 1000 W versus 75 W. Slot width: OAM Module versus single-slot. Suggested PSU: 1400 W versus 250 W. Bus interface: PCIe 5.0 x16 versus PCIe 4.0 x8. Display outputs: none versus 4x DisplayPort 2.0. API support: N/A versus full DirectX, OpenGL, Vulkan. Dimensions: 102 mm length and 165 mm width for the MI350X, versus 254 mm length, 127 mm height, and 20 mm width for the A380E. Release dates differ: 2025-06-11 for the MI350X, 2024-03-31 for the A380E. The A380E predecessor is Xe Graphics, while the MI350X predecessor is Radeon Instinct. The A380E successor is Battlemage; the MI350X has no listed successor.

The Verdict

The data indicates two non-overlapping products. The AMD Instinct MI350X is a compute accelerator with no graphics capabilities: no display outputs, no ROPs, no graphics API support. It delivers 72.09 TFLOPS of FP32 and FP16, 8.19 TB/s of memory bandwidth, and 288 GB of HBM3e. It targets workloads that demand massive parallel compute and large memory capacity. The 1000 W TDP and OAM Module form factor confirm a data center orientation.

The Intel Arc A380E is a graphics processor with 4.096 TFLOPS of FP32, 64.00 GPixel/s pixel rate, 8 RT cores, full DirectX 12 Ultimate, OpenGL 4.6, and Vulkan 1.4 support, plus 4x DisplayPort 2.0 outputs. Its 75 W TDP and single-slot design suit embedded or low-power graphics applications. The A380E is end-of-life, with Battlemage as its successor.

There are no benchmark wins for either side in the database. The verdict must rely on specifications. For compute density, the MI350X leads by orders of magnitude in FP32, FP16, texture rate, and memory bandwidth. For any graphics or display workload, the A380E is the only viable option, as the MI350X cannot output video or execute graphics APIs. The MI350X has 17.6 times the FP32 throughput and 44.0 times the memory bandwidth, but it also consumes 13.3 times the power. The A380E offers 32 ROPs and 8 RT cores, which the MI350X lacks entirely.

Where Each One Wins

The AMD Instinct MI350X wins in every compute-oriented metric. FP32 throughput at 72.09 TFLOPS exceeds the A380E by 17.6 times. FP16 throughput at 72.09 TFLOPS exceeds the A380E by 8.8 times. Texture rate at 2,252.8 GTexel/s exceeds 128.0 GTexel/s by 17.6 times. Memory bandwidth at 8.19 TB/s exceeds 186.0 GB/s by 44.0 times. Memory capacity at 288 GB exceeds 6 GB by 48 times. The MI350X also has 16,384 shading units versus 1,024, and 1,024 TMUs versus 64. The 8192-bit memory bus versus 96 bits provides a massive bandwidth advantage. The 3 nm process and 77.7M transistors per mm² density indicate a more advanced manufacturing node.

The Intel Arc A380E wins in all graphics-specific metrics. It delivers 64.00 GPixel/s pixel rate, while the MI350X records 0 MPixel/s. The A380E has 32 ROPs and 8 RT cores, enabling rasterization and ray tracing that the MI350X cannot perform. The A380E supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, while the MI350X lists N/A for all three. The A380E provides 4x DisplayPort 2.0 outputs, while the MI350X has no outputs. The A380E also has a lower TDP at 75 W versus 1000 W, and a smaller footprint: 254 mm length, 127 mm height, 20 mm width versus 102 mm length, 165 mm width for the MI350X. The A380E runs at a higher base clock of 2000 MHz versus 1000 MHz, and its memory operates at 15.5 Gbps effective versus 8 Gbps effective, though the MI350X compensates with a far wider bus.

For FP16 compute, the A380E achieves an 8.192 TFLOPS rate at a 2:1 ratio, indicating half-rate FP16 execution, while the MI350X achieves full 1:1 FP16 rate. This means the MI350X delivers the same FP16 and FP32 throughput, whereas the A380E doubles its throughput in FP16 relative to FP32. The A380E also has a higher transistor density per watt, though the MI350X has a higher absolute density. The production status differs: the A380E is end-of-life, while the MI350X has no recorded status. The A380E has a successor in Battlemage, while the MI350X does not. The recorded data supports a clear split: compute acceleration belongs to the MI350X, graphics rendering belongs to the A380E.