AMD Steam Machine GPU vs Intel Arc Pro B65 Comparison

Head-to-Head Benchmarks

The database contains no recorded head-to-head benchmark results for the AMD Steam Machine GPU and the Intel Arc Pro B65. Both parts register an average benchmark score of 0, and neither has any individual benchmark entries listed. The win count stands at 0 for each product. This absence of measured data means any direct performance comparison must be inferred from the architectural and specification records rather than from empirical test results.

The AMD Steam Machine GPU posts a raw FP32 throughput of 17.56 TFLOPS, while the Intel Arc Pro B65 delivers 12.29 TFLOPS in FP32. On paper, AMD holds a 42.9% lead in single-precision compute based on these figures. However, the Intel part counters with an FP16 rate of 24.58 TFLOPS (2:1 ratio), compared to AMD's 17.56 TFLOPS (1:1 ratio), giving Intel a 40.0% advantage in half-precision workloads. These numbers suggest AMD favors FP32-centric rendering pipelines, while Intel positions itself for mixed-precision or FP16-heavy compute tasks.

Pixel throughput favors Intel: 192.0 GPixel/s versus 156.8 GPixel/s, a 22.4% margin. Texture fill rates show a similar pattern, with Intel at 384.0 GTexel/s against AMD's 274.4 GTexel/s, a 40.0% difference. These fill-rate gaps indicate Intel's wider ROP and TMU counts translate into higher rasterization throughput under equal clock assumptions.

Memory bandwidth splits the two clearly. Intel's 608.0 GB/s nearly doubles AMD's 288.0 GB/s, a 111.1% advantage. This stems from Intel's 256-bit bus versus AMD's 128-bit bus, combined with slightly faster effective memory speeds (19 Gbps versus 18 Gbps). For bandwidth-sensitive scenarios such as 4K texturing or large dataset processing, the Intel card holds a substantial edge.

Boost clocks differ: AMD reaches 2450 MHz, Intel holds a flat 2400 MHz. The AMD part also lists a game clock of 2250 MHz, a figure Intel does not record. Base clocks tell a different story, with Intel starting at 2400 MHz and AMD at 1720 MHz, a 39.5% higher floor for Intel. This suggests Intel maintains consistent frequency across sustained loads, while AMD relies on boost headroom.

Architecture Differences

The AMD Steam Machine GPU uses the Navi 33 chip built on RDNA 3.0 architecture, codenamed Hotpink Bonefish. It belongs to the Console GPU (Valve) generation and is fabricated on a 6 nm TSMC process. The die measures 204 mm² and contains 13,300 million transistors, yielding a density of 65.2 million transistors per square millimeter.

Intel Arc Pro B65 uses the BMG-G21 chip based on Xe2-HPG architecture, part of the Battlemage (Pro Series) generation. This GPU is fabricated on a 5 nm TSMC process, with a die size of 272 mm² and 19,600 million transistors. Density reaches 72.1 million transistors per square millimeter, 10.6% higher than AMD's implementation.

The transistor counts differ markedly: Intel packs 47.4% more transistors into a 33.3% larger die. This reflects the different design philosophies. AMD's RDNA 3.0 focuses on compute-per-area efficiency with a smaller chip, while Intel's Xe2-HPG spreads more hardware across a larger footprint.

Shading unit counts diverge: Intel carries 2560 shading units against AMD's 1792, a 42.9% advantage. TMU counts follow: 160 versus 112, a 42.9% difference. ROPs also favor Intel, 80 versus 64. Yet AMD's RT core count is higher at 28, compared to Intel's 20, a 40.0% advantage for AMD in ray tracing hardware units.

The FP32-to-FP16 ratio reveals architectural intent. AMD runs 1:1, meaning equal throughput for both precisions. Intel runs 2:1, doubling FP16 throughput relative to FP32. This makes Intel more suited for AI inference, image processing, or other half-precision tasks, while AMD keeps balanced throughput across precisions.

Both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, so API-level feature parity holds. Neither lists tensor cores, so dedicated AI acceleration hardware is absent from both records.

Where Each One Wins

AMD Steam Machine GPU wins in raw FP32 compute. The 17.56 TFLOPS figure exceeds Intel's 12.29 TFLOPS by 42.9%. Game rendering pipelines that rely heavily on FP32 shader math should favor the AMD part. The higher RT core count (28 versus 20) also suggests better ray tracing throughput per clock, assuming similar core efficiency.

AMD also draws less power. The 110 W TDP compares favorably to Intel's 200 W, an 81.8% higher power draw for Intel. For compact console-style builds or systems with modest power delivery, the AMD card fits better. It requires no power connectors, while Intel demands a single 8-pin connector and a 550 W suggested PSU.

Intel Arc Pro B65 wins in memory capacity and bandwidth. The 32 GB frame buffer quadruples AMD's 8 GB, and the 608.0 GB/s bandwidth more than doubles AMD's 288.0 GB/s. For large textures, compute workloads with big datasets, or multi-monitor high-resolution output, Intel's memory subsystem provides clear headroom.

Intel also leads in fill rates. Pixel rate of 192.0 GPixel/s and texture rate of 384.0 GTexel/s outpace AMD's 156.8 GPixel/s and 274.4 GTexel/s respectively. Rasterization-heavy scenes with many overdraws or high texel density favor Intel.

The FP16 throughput advantage (24.58 TFLOPS versus 17.56 TFLOPS) gives Intel an edge in half-precision compute. Applications that convert FP32 to FP16 for speed, such as certain image processing or scientific workloads, will run faster on Intel.

Specification Differences

| Specification | AMD Steam Machine GPU | Intel Arc Pro B65 |

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

| Chip | Navi 33 | BMG-G21 |

| Architecture | RDNA 3.0 | Xe2-HPG |

| Generation | Console GPU (Valve) | Battlemage (Pro Series) |

| Process Node | 6 nm | 5 nm |

| Transistors | 13,300 million | 19,600 million |

| Die Size | 204 mm² | 272 mm² |

| Transistor Density | 65.2M / mm² | 72.1M / mm² |

| Base Clock | 1720 MHz | 2400 MHz |

| Boost Clock | 2450 MHz | 2400 MHz |

| Game Clock | 2250 MHz | Not listed |

| Memory Size | 8 GB | 32 GB |

| Memory Type | GDDR6 | GDDR6 |

| Memory Bus Width | 128 bit | 256 bit |

| Memory Bandwidth | 288.0 GB/s | 608.0 GB/s |

| Shading Units | 1792 | 2560 |

| TMUs | 112 | 160 |

| ROPs | 64 | 80 |

| RT Cores | 28 | 20 |

| Pixel Rate | 156.8 GPixel/s | 192.0 GPixel/s |

| Texture Rate | 274.4 GTexel/s | 384.0 GTexel/s |

| FP32 | 17.56 TFLOPS | 12.29 TFLOPS |

| FP16 | 17.56 TFLOPS (1:1) | 24.58 TFLOPS (2:1) |

| TDP | 110 W | 200 W |

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

| Suggested PSU | Not listed | 550 W |

| Bus Interface | Not listed | PCIe 5.0 x16 |

| Display Outputs | 1x HDMI 2.1, 1x DisplayPort 2.1 | 4x DisplayPort 2.1 |

| Slot Width | Not listed | Dual-slot |

| Release Date | 2026-06-28 | 2026-03-31 |

The release dates show Intel launched roughly three months earlier (March 2026 versus June 2026). Both parts remain in active production per the database records.

FAQ

Q: Which GPU has higher FP32 compute?

A: The AMD Steam Machine GPU delivers 17.56 TFLOPS, which is 42.9% higher than the Intel Arc Pro B65's 12.29 TFLOPS.

Q: How much memory does each card carry?

A: The AMD part has 8 GB GDDR6 on a 128-bit bus, while the Intel part has 32 GB GDDR6 on a 256-bit bus. Intel's bandwidth of 608.0 GB/s is 111.1% higher than AMD's 288.0 GB/s.

Q: What are the power requirements?

A: AMD lists a 110 W TDP with no power connectors. Intel lists a 200 W TDP, requires one 8-pin connector, and suggests a 550 W PSU.

Q: Do both support the same graphics APIs?

A: Yes. Both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Q: Which card has more ray tracing cores?

A: AMD has 28 RT cores, while Intel has 20. This gives AMD a 40.0% higher RT core count.

Q: What is the process node difference?

A: AMD uses TSMC 6 nm with 13,300 million transistors on a 204 mm² die. Intel uses TSMC 5 nm with 19,600 million transistors on a 272 mm² die. Intel's density is 72.1M transistors per mm² versus AMD's 65.2M.

The Verdict

The recorded data points to two different design targets. The AMD Steam Machine GPU prioritizes raw FP32 throughput and low power draw. Its 17.56 TFLOPS FP32 performance, 110 W TDP, and connector-free design fit a console-style enclosure where thermal and power constraints are tight. The higher RT core count also suggests stronger ray tracing hardware per watt.

The Intel Arc Pro B65 targets capacity and bandwidth. The 32 GB memory pool and 608.0 GB/s bandwidth serve professional workloads, large frame buffers, or multi-display setups. The 200 W TDP and 8-pin connector demand a proper power supply, but the fill rates (192.0 GPixel/s, 384.0 GTexel/s) and FP16 throughput (24.58 TFLOPS) indicate sustained throughput for content creation or compute tasks.

For gaming at typical resolutions, the AMD part's higher FP32 compute and lower power draw make it the practical choice in constrained builds. For professional use, large datasets, or bandwidth-bound workloads, the Intel part's memory subsystem and fill rates dominate. The absence of benchmark results means these conclusions rest on specification analysis. Builders should weigh their workload's dominant pressure: compute throughput versus memory capacity and bandwidth.