AMD Instinct MI350X vs Intel Iris Xe Graphics 80EU Mobile Comparison

The Verdict

The database places both the AMD Instinct MI350X and the Intel Iris Xe Graphics 80EU Mobile at the 50th percentile among all GPUs, but this parity in rank obscures a fundamental divergence in purpose and capability. The MI350X is a 1000 W OAM accelerator module designed for compute-heavy datacenter workloads, while the Iris Xe is a 15 W integrated graphics processor for mobile systems. The recorded specifications show no overlap in their target use cases.

From the data, the MI350X delivers 72.09 TFLOPS of FP32 throughput, which is 38.8 times the 1.856 TFLOPS of the Iris Xe. Its memory subsystem provides 288 GB of HBM3e with 8.19 TB/s of bandwidth, a figure that dwarfs the system-shared memory of the Intel part. The MI350X uses 16384 shading units against 640 for the Iris Xe. Any workload that fits in the accelerator's memory pool or scales across its compute units belongs to the AMD part. The Iris Xe, conversely, supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4, while the MI350X lists no graphics API support at all. The Intel part is the only one of the two with display outputs, making it the sole candidate for rendering to a screen.

The data indicates the MI350X is for server racks, scientific computing, and AI inference or training. The Iris Xe is for laptops and portable devices where graphics output and low power draw are mandatory. There is no scenario in the recorded specifications where one substitutes for the other.

Architecture Differences

The MI350X uses the CDNA 4.0 architecture, fabricated on a 3 nm process at TSMC. The chip, labeled MI350 256CU, contains 185,000 million transistors on a die size of 2380 mm², yielding a transistor density of 77.7 million per square millimeter. The Iris Xe uses Intel's Generation 12.2 architecture on a 10 nm process, with the chip described as Raptor Lake. No transistor count or die size is recorded for the Intel part. The MI350X belongs to the Instinct (MIx) generation, while the Iris Xe belongs to the HD Graphics-M (Raptor Lake) generation.

The MI350X has no ROPs, recording a pixel rate of 0 MPixel/s. It has 1024 texture mapping units and a texture rate of 2,252.8 GTexel/s. The Iris Xe has 20 ROPs and 40 TMUs, with a pixel rate of 29.00 GPixel/s and a texture rate of 58.00 GTexel/s. The MI350X lists no RT cores and no tensor cores in the data. The Iris Xe also lists none. The MI350X supports FP16 at a 1:1 ratio with FP32, both at 72.09 TFLOPS. The Iris Xe supports FP16 at a 2:1 ratio, delivering 3.712 TFLOPS against 1.856 TFLOPS for FP32.

The MI350X has no display outputs and no graphics API support (DirectX, OpenGL, and Vulkan are all marked N/A). The Iris Xe supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The MI350X uses a PCIe 5.0 x16 bus interface, while the Iris Xe uses a Ring Bus. The MI350X is an OAM Module with a slot width of OAM, requiring no power connectors but a suggested PSU of 1400 W. The Iris Xe is an IGP with no power connector data and no suggested PSU.

Head-to-Head Benchmarks

The database records no benchmark scores for either GPU. The head-to-head benchmark array is empty, and both parts show an average benchmark score of 0. The wins counter is 0 for each side. This absence of measured performance data forces the analysis to rely entirely on the recorded hardware specifications.

The FP32 compute ratio is the clearest differentiator. The MI350X delivers 72.09 TFLOPS, which is 38.8 times the 1.856 TFLOPS of the Iris Xe. This gap reflects the 16384 shading units versus 640, a 25.6 times difference in raw shader count, combined with the MI350X boost clock of 2200 MHz against 1450 MHz for the Iris Xe. The base clocks differ by a wider margin: 1000 MHz for the MI350X versus 300 MHz for the Intel part.

Texture throughput follows a similar pattern. The MI350X reaches 2,252.8 GTexel/s, which is 38.8 times the 58.00 GTexel/s of the Iris Xe. The pixel rate tells a different story: the MI350X records 0 MPixel/s because it has no ROPs, while the Iris Xe achieves 29.00 GPixel/s. For any rasterization or display output task, the Intel part is the only one with functional capability. The MI350X is not a graphics card in the traditional sense; it has no pixel pipeline.

Memory bandwidth is another categorical gap. The MI350X provides 8.19 TB/s across an 8192-bit bus using HBM3e. The Iris Xe uses system shared memory with system dependent bandwidth, so no direct numeric comparison is possible. The MI350X memory clock is 2000 MHz with 8 Gbps effective speed. The Iris Xe memory clock is listed as System Shared. The MI350X holds 288 GB of memory, a capacity that no integrated GPU can approach.

Specification Differences

The two parts differ in nearly every recorded specification field. The process node is 3 nm for the MI350X versus 10 nm for the Iris Xe. The foundry is TSMC for AMD and Intel for the Intel part. The MI350X has a transistor count of 185,000 million and a die size of 2380 mm²; the Iris Xe has neither recorded. The MI350X has a transistor density of 77.7 million per square millimeter; the Iris Xe has none.

Clock speeds diverge sharply. The MI350X base clock is 1000 MHz with a boost of 2200 MHz. The Iris Xe base is 300 MHz with a boost of 1450 MHz. The MI350X memory clock is 2000 MHz (8 Gbps effective), while the Iris Xe uses System Shared memory. The MI350X has 288 GB of HBM3e on an 8192-bit bus with 8.19 TB/s bandwidth. The Iris Xe has system shared memory, type, bus width, and bandwidth.

Compute resources differ by an order of magnitude. The MI350X has 16384 shading units, 1024 TMUs, and 0 ROPs. The Iris Xe has 640 shading units, 40 TMUs, and 20 ROPs. The MI350X pixel rate is 0 MPixel/s; the Iris Xe is 29.00 GPixel/s. The MI350X texture rate is 2,252.8 GTexel/s; the Iris Xe is 58.00 GTexel/s. FP32 throughput is 72.09 TFLOPS versus 1.856 TFLOPS. FP16 is 72.09 TFLOPS (1:1) versus 3.712 TFLOPS (2:1).

Power and physical specifications are incompatible. The MI350X TDP is 1000 W with a suggested PSU of 1400 W, an OAM Module slot width, and dimensions of 102 mm length and 165 mm width. The Iris Xe TDP is 15 W with no suggested PSU, an IGP slot width, and no recorded dimensions. The MI350X uses PCIe 5.0 x16; the Iris Xe uses Ring Bus. The MI350X has no display outputs; the Iris Xe has portable device dependent outputs. The MI350X has no graphics API support; the Iris Xe supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The MI350X release date is June 11, 2025; the Iris Xe release date is January 3, 2023. The Iris Xe has a successor listed as Arc Graphics-M; the MI350X has no successor. The MI350X predecessor is Radeon Instinct; the Iris Xe has no predecessor.

FAQ

Q: Which GPU has higher FP32 compute performance?

A: The AMD Instinct MI350X delivers 72.09 TFLOPS of FP32, which is 38.8 times the 1.856 TFLOPS of the Intel Iris Xe Graphics 80EU Mobile.

Q: Can the AMD Instinct MI350X output video to a display?

A: No. The database records no display outputs for the MI350X, and its pixel rate is 0 MPixel/s because it has no ROPs. The Intel Iris Xe has portable device dependent display outputs and a pixel rate of 29.00 GPixel/s.

Q: What is the memory configuration of each GPU?

A: The MI350X has 288 GB of HBM3e memory on an 8192-bit bus with 8.19 TB/s bandwidth. The Iris Xe uses system shared memory with system dependent bandwidth and no dedicated capacity.

Q: Which GPU supports graphics APIs like DirectX and Vulkan?

A: The Intel Iris Xe supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The AMD MI350X lists DirectX, OpenGL, and Vulkan as N/A.

Q: What is the power consumption difference?

A: The MI350X has a TDP of 1000 W with a suggested PSU of 1400 W. The Iris Xe has a TDP of 15 W and no suggested PSU recorded.

Q: How do the release dates compare?

A: The AMD Instinct MI350X was released on June 11, 2025. The Intel Iris Xe Graphics 80EU Mobile was released on January 3, 2023, and has a successor listed as Arc Graphics-M.

Where Each One Wins

The AMD Instinct MI350X wins in every computational throughput metric recorded in the database. Its FP32 output of 72.09 TFLOPS is nearly 39 times that of the Intel part, and its FP16 output matches that figure at a 1:1 ratio. The texture rate of 2,252.8 GTexel/s is 38.8 times higher. The memory subsystem provides 288 GB of HBM3e at 8.19 TB/s, eliminating any bottleneck from system shared memory. The 8192-bit bus width is unmatched by any integrated solution. The 3 nm process and 185,000 million transistors indicate a design optimized for density and parallel execution. The PCIe 5.0 x16 interface allows high bandwidth host communication. This GPU is for workloads that saturate compute units and memory bandwidth: large matrix operations, scientific simulation, and AI model training or inference where the dataset fits within 288 GB.

The Intel Iris Xe Graphics 80EU Mobile wins in the categories that the MI350X does not compete in. It has 20 ROPs and a pixel rate of 29.00 GPixel/s, making it capable of rasterizing frames. It supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4, which means it can run modern graphics applications. It has display outputs, though the specifics are portable device dependent. Its 15 W TDP fits into mobile thermal envelopes, while the MI350X requires 1000 W and a 1400 W suggested PSU. The Iris Xe uses a Ring Bus interface, which is typical for integrated graphics in a laptop processor. Its FP16 throughput of 3.712 TFLOPS, while far below the MI350X, still exceeds its own FP32 rate by a factor of 2, indicating some efficiency for half-precision workloads. The release date of January 3, 2023, and the active production status suggest it is a current product for mobile systems.

The data supports a clean split: the MI350X for compute acceleration without any display or graphics API requirements, and the Iris Xe for mobile graphics output, API compatibility, and low power operation. Neither part encroaches on the other's domain.