AMD Instinct MI350P vs AMD Radeon 8065S Comparison

Where Each One Wins

The AMD Instinct MI350P and AMD Radeon 8065S occupy entirely separate performance domains, and the recorded data makes that split unambiguous. The MI350P is built for data-center compute workloads, while the 8065S targets mobile graphics with display output support and API compatibility. Neither part wins across the board; each dominates the categories its architecture was designed to serve.

The MI350P holds every advantage in raw compute throughput. Its FP32 rating of 36.04 TFLOPS is more than double the 8065S's 15.36 TFLOPS, and its FP16 output matches at 36.04 TFLOPS (1:1) versus the 8065S's 15.36 TFLOPS (1:1). Texture rate also favors the Instinct card decisively: 1,126.4 GTexel/s against 480.0 GTexel/s. These are not marginal gaps; they represent a fundamental difference in processing scale. The MI350P uses 8,192 shading units and 512 TMUs, while the 8065S employs 2,560 shading units and 160 TMUs.

The 8065S wins in areas the MI350P cannot even enter. The Instinct card outputs 0 MPixel/s and has no display outputs, whereas the Radeon delivers 192.0 GPixel/s pixel rate and supports portable-device-dependent display connections. The 8065S also carries 40 ray tracing cores; the MI350P lists none. API support splits the same way: the 8065S supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, while the MI350P reports N/A across all three.

Power efficiency favors the mobile part by a wide margin. The 8065S operates at 55 W TDP with no power connectors, while the MI350P draws 600 W and requires a single 16-pin connector plus a 1000 W suggested PSU. The MI350P is a dual-slot accelerator; the 8065S is an integrated graphics processor. These are not competing products in any practical sense; they are specialized tools for different jobs.

The Verdict

The data dictates a simple conclusion: pick the AMD Instinct MI350P for compute-heavy, server-side workloads where raw throughput and massive memory capacity matter, and pick the AMD Radeon 8065S for mobile systems needing graphics output, ray tracing, and modern API support.

The MI350P's 144 GB of HBM3e memory on an 8192-bit bus with 8.19 TB/s bandwidth has no equivalent in the 8065S, which relies on system-shared memory with system-dependent bandwidth. Any workload that fits in the Instinct's memory pool will benefit from bandwidth that the Radeon cannot approach. FP32 and FP16 compute at 36.04 TFLOPS, combined with a 1,126.4 GTexel/s texture rate, positions the MI350P as the clear choice for simulation, AI training, or scientific computing.

The 8065S answers a different question. Its 3000 MHz boost clock, 192.0 GPixel/s pixel rate, and 40 ray tracing cores make it suitable for graphics rendering and gaming in portable devices. The 12 Ultimate API support and Vulkan 1.4 compatibility mean modern graphics features are available, something the MI350P lacks entirely. The 55 W TDP makes it feasible for battery-powered systems, whereas the MI350P's 600 W requirement demands a workstation or server environment.

Neither part should be chosen for the other's strengths. Selecting the MI350P for graphics output would fail immediately because it has no display outputs. Selecting the 8065S for large-scale compute would hit memory and throughput ceilings quickly. The recorded data shows two specialized products, and the verdict follows directly from their specifications.

Head-to-Head Benchmarks

The benchmark database lists no direct head-to-head benchmark results between these two parts, and both have an average benchmark score of 0 with empty benchmark arrays. However, the specification data provides a quantitative basis for comparison, and the differences are substantial.

FP32 compute is the largest numerical gap. The MI350P delivers 36.04 TFLOPS, which is 134.6% higher than the 8065S's 15.36 TFLOPS. In practical terms, the Instinct card processes more than two FP32 operations for every one the Radeon handles. FP16 follows the same pattern: 36.04 TFLOPS (1:1) versus 15.36 TFLOPS (1:1), an identical 134.6% advantage. The 1:1 ratio on both parts means no half-rate penalty, but the absolute throughput difference remains.

Texture rate shows a similar story. The MI350P's 512 TMUs produce 1,126.4 GTexel/s, while the 8065S's 160 TMUs produce 480.0 GTexel/s. That is a 134.7% lead for the Instinct part. The Radeon counters in pixel throughput: 192.0 GPixel/s versus 0 MPixel/s for the MI350P, which has no ROP functionality listed. This is not a close contest in either direction; each part wins its category by the full margin available.

Memory bandwidth is where the separation becomes extreme. The MI350P accesses 144 GB of HBM3e across an 8192-bit bus, achieving 8.19 TB/s. The 8065S uses system-shared memory with system-dependent bandwidth, meaning its memory performance is not fixed in the database. Even the fastest shared-memory configurations would need to approach 8.19 TB/s to match the Instinct card, which is unlikely given the 55 W TDP envelope. The MI350P's 8,192-bit memory bus is 32 times wider than a typical 256-bit discrete GPU bus, though the 8065S's bus width is listed as system-shared, so no direct bit-width comparison is possible.

Clock speeds favor the Radeon in frequency terms. The 8065S boosts to 3000 MHz from a 1295 MHz base, while the MI350P boosts to 2200 MHz from a 1000 MHz base. The 36.4% higher boost clock on the 8065S helps its graphics workloads but cannot overcome the Instinct's 3.2x shading unit advantage and 3.2x TMU advantage.

FAQ

Q: Which GPU has higher FP32 performance?

A: The AMD Instinct MI350P delivers 36.04 TFLOPS FP32, which is 134.6% higher than the AMD Radeon 8065S's 15.36 TFLOPS.

Q: Does the AMD Instinct MI350P support display outputs?

A: No. The MI350P lists no display outputs and a pixel rate of 0 MPixel/s. The Radeon 8065S supports portable-device-dependent display outputs and delivers 192.0 GPixel/s.

Q: What memory configuration does each GPU use?

A: The MI350P uses 144 GB of HBM3e on an 8192-bit bus with 8.19 TB/s bandwidth. The 8065S uses system-shared memory with system-dependent bandwidth.

Q: Which GPU supports ray tracing?

A: Only the AMD Radeon 8065S, which has 40 ray tracing cores. The MI350P lists no ray tracing cores.

Q: What are the power requirements for each part?

A: The MI350P has a 600 W TDP, uses a 1x 16-pin power connector, and suggests a 1000 W PSU. The 8065S has a 55 W TDP and requires no power connectors.

Q: Which GPU has newer API support?

A: The 8065S supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The MI350P reports N/A for DirectX, OpenGL, and Vulkan.

Architecture Differences

The two AMD parts come from different architecture families, and the data shows how that shapes their capabilities. The MI350P uses CDNA 4.0, AMD's compute-optimized architecture, built for data-center acceleration. The 8065S uses RDNA 3.5, the graphics-focused architecture, designed for mobile and consumer workloads.

Manufacturing processes differ as well. The MI350P is fabricated on a 3 nm process at TSMC, while the 8065S uses a 4 nm process, also at TSMC. The smaller node on the Instinct card helps pack 73,000 million transistors into a 1190 mm² die, yielding a transistor density of 61.3M per mm². The 8065S has a 308 mm² die with an unknown transistor count and no density figure listed.

Shading resources reflect the architectural split. The MI350P contains 8,192 shading units, 512 TMUs, and no ROPs. The 8065S contains 2,560 shading units, 160 TMUs, and 64 ROPs. The absence of ROPs on the MI350P explains its 0 MPixel/s pixel rate; it is not designed to rasterize graphics. The 8065S's 40 ray tracing cores add hardware acceleration for ray-traced effects, a feature entirely absent from the MI350P.

The chip identities also differ. The MI350P uses the MI350 128CU chip, and the 8065S uses the Gorgon Halo chip. These belong to different generations: the MI350P is in the Instinct (MIx) generation, while the 8065S is in the Navi Mobile (RX 8000M) generation. The MI350P's predecessor is Radeon Instinct, and the 8065S's predecessor is Polaris Mobile.

Memory architecture is a major differentiator. The MI350P has dedicated HBM3e memory with a fixed 8.19 TB/s bandwidth. The 8065S shares system memory, making its bandwidth dependent on the host platform. This fundamental difference means the MI350P's memory performance is predictable and extremely high, while the 8065S's memory performance varies with the system it is installed in.

Specification Differences

The recorded data shows several fields where the two parts differ directly. Clock speeds: the MI350P runs at 1000 MHz base and 2200 MHz boost, while the 8065S runs at 1295 MHz base and 3000 MHz boost. The Radeon's boost clock is 36.4% higher.

Memory: the MI350P has 144 GB of HBM3e on an 8192-bit bus with 8.19 TB/s bandwidth. The 8065S has system-shared memory, system-shared type, system-shared bus width, and system-dependent bandwidth. Memory clock differs too: 2000 MHz (8 Gbps effective) on the MI350P versus system-shared on the 8065S.

Compute units: the MI350P's 8,192 shading units, 512 TMUs, and 0 ROPs contrast with the 8065S's 2,560 shading units, 160 TMUs, and 64 ROPs. The 8065S adds 40 ray tracing cores; the MI350P has none.

Output rates: the MI350P produces 0 MPixel/s and 1,126.4 GTexel/s. The 8065S produces 192.0 GPixel/s and 480.0 GTexel/s. FP32 and FP16 are both 36.04 TFLOPS (1:1) on the MI350P and 15.36 TFLOPS (1:1) on the 8065S.

Power and physical design: the MI350P has a 600 W TDP, dual-slot width, a 1x 16-pin power connector, and a 1000 W suggested PSU. The 8065S has a 55 W TDP, IGP slot width, no power connectors, and no suggested PSU. The MI350P measures 267 mm in length, 111 mm in height, and 40 mm in width; the 8065S has no listed dimensions.

Interfaces and outputs: both use PCIe 5.0 x16. The MI350P has no display outputs; the 8065S has portable-device-dependent outputs. API support: the MI350P lists N/A for DirectX, OpenGL, and Vulkan; the 8065S lists DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. Process node: 3 nm for the MI350P versus 4 nm for the 8065S. Die size: 1190 mm² versus 308 mm². Transistors: 73,000 million versus unknown. Release dates: the MI350P releases on 2026-05-06, and the 8065S on 2025-12-31. Production status: the MI350P has none listed, while the 8065S is listed as active.