Intel Arc A550M vs NVIDIA CMP 90HX Comparison

Head-to-Head Benchmarks

The recorded data contains one direct benchmark comparison between these two accelerators, and it is decisive. In the Geekbench OpenCL test, the NVIDIA CMP 90HX scores 69,000 points, while the Intel Arc A550M scores 49,894 points. The delta is a 38.3% advantage for the NVIDIA part, which represents a substantial performance gap in compute workloads. This is not a marginal difference; it places the CMP 90HX firmly in a different performance class.

The database confirms the NVIDIA CMP 90HX sits at the 90th percentile among all GPUs, while the Intel Arc A550M sits at the 86th percentile. While both are above the midpoint, the gap between the 90th and 86th percentile is meaningful, especially when considering the CMP 90HX is a dedicated mining product with no display outputs, whereas the Arc A550M is a mobile part designed for portable devices. The percentile difference aligns with the raw score difference in the OpenCL test, reinforcing that the NVIDIA part is the stronger compute performer.

The Intel Arc A550M does have a second benchmark entry, a Vulkan score of 49,580, which is close to its OpenCL result of 49,894. This consistency across APIs suggests the Arc A550M's compute performance is stable regardless of the workload interface. The NVIDIA CMP 90HX, however, only has an OpenCL entry, so no Vulkan comparison is available in the database. The absence of a Vulkan score for the CMP 90HX means the comparison relies entirely on the OpenCL result, which still favors NVIDIA by a wide margin.

Looking at the nearest rivals for each part provides additional context. The NVIDIA CMP 90HX is 0.3% ahead of the Intel Arc A770, which scores 68,809, and 0.6% ahead of the AMD Radeon Instinct MI25 at 68,562. It trails the AMD Radeon Pro WX 8200 by 1.2% (69,870) and the NVIDIA Quadro P6000 by 1.4% (69,986). These are tight margins, indicating the CMP 90HX is positioned within a cluster of high-end workstation and compute accelerators. The Intel Arc A550M, by contrast, is 0.4% behind the NVIDIA GeForce RTX 5070 Ti (49,957), 0.5% behind the AMD Radeon RX Vega 64 (50,001), and 2.4% behind the AMD Radeon RX 6900 XT (50,951). It is 2.6% ahead of the AMD Radeon RX 6800 XT (48,477). The Arc A550M's rivals are a mix of consumer gaming cards and older high-end parts, and its position among them is roughly mid-pack.

FAQ

Q: Which GPU wins the head-to-head benchmark?

A: The NVIDIA CMP 90HX wins the only direct comparison, the Geekbench OpenCL test, with a score of 69,000 versus 49,894 for the Intel Arc A550M, a 38.3% difference.

Q: How does the NVIDIA CMP 90HX compare to its closest competitors?

A: The CMP 90HX is 0.3% ahead of the Intel Arc A770, 0.6% ahead of the AMD Radeon Instinct MI25, 1.2% behind the AMD Radeon Pro WX 8200, and 1.4% behind the NVIDIA Quadro P6000.

Q: Where does the Intel Arc A550M stand relative to its nearest rivals?

A: The Arc A550M is 0.4% behind the NVIDIA GeForce RTX 5070 Ti, 0.5% behind the AMD Radeon RX Vega 64, 2.4% behind the AMD Radeon RX 6900 XT, and 2.6% ahead of the AMD Radeon RX 6800 XT.

Q: Are there other benchmark results for these GPUs?

A: The NVIDIA CMP 90HX has only a Geekbench OpenCL score. The Intel Arc A550M has both a Geekbench OpenCL score of 49,894 and a Geekbench Vulkan score of 49,580.

Q: Which GPU has the higher overall percentile ranking?

A: The NVIDIA CMP 90HX ranks in the 90th percentile among all GPUs, while the Intel Arc A550M ranks in the 86th percentile.

Architecture Differences

The two GPUs come from different manufacturers, foundries, and architectural generations. The NVIDIA CMP 90HX is built on the GA102 chip, uses the Ampere architecture, and belongs to the Mining GPUs generation. It is fabricated on an 8 nm process at Samsung. The Intel Arc A550M uses the DG2-512 chip, employs the Xe-HPG architecture, and is part of the Alchemist (Arc 5 Mobile) generation. It is fabricated on a 6 nm process at TSMC.

The transistor counts differ considerably. The NVIDIA part integrates 28,300 million transistors on a 628 mm² die, resulting in a transistor density of 45.1 million per mm². The Intel part has 21,700 million transistors on a 406 mm² die, yielding a higher density of 53.4 million per mm². The smaller, denser Intel process node allows for a more compact die despite the architectural differences.

Compute resources vary widely. The NVIDIA CMP 90HX has 6,400 shading units, 200 texture mapping units, and 80 raster operation units. It also includes 50 RT cores and 200 tensor cores. The Intel Arc A550M has 2,048 shading units, 128 TMUs, and 64 ROPs, along with 16 RT cores. The Intel part does not have tensor cores listed in the database, which is a notable difference, as the NVIDIA part relies on them for AI-accelerated workloads.

Clock speeds reveal an interesting contrast. The NVIDIA CMP 90HX has a base clock of 1500 MHz and a boost clock of 1710 MHz. The Intel Arc A550M has a much lower base clock of 900 MHz but a significantly higher boost clock of 2050 MHz. This suggests the Intel part is designed to ramp aggressively under load, while the NVIDIA part operates at a more constant, higher baseline.

Memory subsystems are also divergent. The NVIDIA CMP 90HX ships with 10 GB of GDDR6X on a 320-bit bus, delivering 760.3 GB/s of bandwidth. The memory clock is listed at 1188 MHz with 19 Gbps effective speed. The Intel Arc A550M has 8 GB of GDDR6 on a 128-bit bus, providing 224.0 GB/s of bandwidth, with a memory clock of 1750 MHz and 14 Gbps effective speed. The NVIDIA part has over three times the memory bandwidth, which is a critical factor in memory-bound compute workloads.

Rates and throughput figures follow the same trend. The NVIDIA CMP 90HX achieves a pixel rate of 136.8 GPixel/s and a texture rate of 342.0 GTexel/s. The Intel Arc A550M achieves 131.2 GPixel/s and 262.4 GTexel/s. The pixel rates are close, but the texture rate gap is substantial, favoring the NVIDIA part.

Compute throughput shows the largest divergence. The NVIDIA CMP 90HX delivers 21.89 TFLOPS for FP32 and the same 21.89 TFLOPS for FP16, a 1:1 ratio. The Intel Arc A550M delivers 8.397 TFLOPS for FP32 and 16.79 TFLOPS for FP16, a 2:1 ratio. The NVIDIA part is roughly 2.6 times faster than the Intel part in FP32, while the Intel part's FP16 advantage is halved due to its 2:1 architecture. For workloads that rely on FP32, this is a decisive gap.

Power and physical characteristics also differ sharply. The NVIDIA CMP 90HX has a TDP of 320 W, requires a dual-slot cooler, uses 2x 8-pin power connectors, and needs a 700 W suggested power supply. It is 285 mm long and 112 mm tall. The Intel Arc A550M has a TDP of 60 W, is an IGP form factor, has no power connectors listed, and no suggested PSU. It has no listed dimensions. The form factor difference is extreme: one is a large, power-hungry expansion card, the other is an integrated mobile GPU.

Interface and output differences are notable. The NVIDIA CMP 90HX uses a PCIe 1.0 x4 bus interface, which is unusual for a modern accelerator, and it has no display outputs. The Intel Arc A550M uses PCIe 4.0 x16, a current-generation interface, and its display outputs are listed as portable device dependent. The NVIDIA part is clearly not intended for any visual output; the Intel part is designed for laptops and portable systems.

Both GPUs support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, so API compatibility is equivalent. Both are end-of-life products. The NVIDIA CMP 90HX has a release date of July 27, 2021; the Intel Arc A550M has no release date listed in the database.

The Verdict

The data points to a clear split based on workload type. For raw compute throughput, especially FP32-heavy tasks, the NVIDIA CMP 90HX is the obvious choice. Its 21.89 TFLOPS FP32 performance, 760.3 GB/s memory bandwidth, and 90th percentile ranking make it a dominant force in compute benchmarks. The 38.3% lead over the Intel Arc A550M in OpenCL is decisive, and its position among rivals like the Intel Arc A770 and AMD Radeon Instinct MI25 shows it competes at the high end of the workstation and accelerator market.

However, the CMP 90HX comes with significant limitations. It has no display outputs, uses an outdated PCIe 1.0 x4 interface, consumes 320 W, and requires a 700 W power supply. It is a mining-specific product, end-of-life, with no launch MSRP recorded. It is a compute tool, not a general-purpose GPU. For anyone needing a display, this part is unsuitable.

The Intel Arc A550M is a different animal. With a 60 W TDP, no power connectors, and an IGP form factor, it is designed for mobile integration. Its 8.397 TFLOPS FP32 and 16.79 TFLOPS FP16 performance are respectable for a 60 W part, and its 86th percentile ranking places it above many consumer GPUs. Its Vulkan score of 49,580 is close to its OpenCL score, indicating consistent performance across APIs. It trails the RTX 5070 Ti by only 0.4% and beats the RX 6800 XT by 2.6%, which is strong for a low-power mobile part.

The choice depends on the intended use. If the workload is compute-only, requires maximum FP32 throughput, and can tolerate high power draw, the NVIDIA CMP 90HX is superior. If the workload runs on a portable device, needs display output, or operates under strict power limits, the Intel Arc A550M is the only viable option. The database does not provide enough data to declare a universal winner; it offers a performance hierarchy where the NVIDIA part leads in raw speed, and the Intel part leads in efficiency and integration.

Specification Differences

| Specification | NVIDIA CMP 90HX | Intel Arc A550M |

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

| Chip | GA102 | DG2-512 |

| Architecture | Ampere | Xe-HPG |

| Generation | Mining GPUs | Alchemist (Arc 5 Mobile) |

| Process Node | 8 nm | 6 nm |

| Foundry | Samsung | TSMC |

| Transistors | 28,300 million | 21,700 million |

| Die Size | 628 mm² | 406 mm² |

| Transistor Density | 45.1M / mm² | 53.4M / mm² |

| Base Clock | 1500 MHz | 900 MHz |

| Boost Clock | 1710 MHz | 2050 MHz |

| Memory Clock | 1188 MHz, 19 Gbps effective | 1750 MHz, 14 Gbps effective |

| Memory Size | 10 GB | 8 GB |

| Memory Type | GDDR6X | GDDR6 |

| Memory Bus Width | 320 bit | 128 bit |

| Memory Bandwidth | 760.3 GB/s | 224.0 GB/s |

| Shading Units | 6400 | 2048 |

| TMUs | 200 | 128 |

| ROPs | 80 | 64 |

| RT Cores | 50 | 16 |

| Tensor Cores | 200 | null |

| Pixel Rate | 136.8 GPixel/s | 131.2 GPixel/s |

| Texture Rate | 342.0 GTexel/s | 262.4 GTexel/s |

| FP32 | 21.89 TFLOPS | 8.397 TFLOPS |

| FP16 | 21.89 TFLOPS (1:1) | 16.79 TFLOPS (2:1) |

| TDP | 320 W | 60 W |

| Slot Width | Dual-slot | IGP |

| Power Connectors | 2x 8-pin | null |

| Suggested PSU | 700 W | null |

| Bus Interface | PCIe 1.0 x4 | PCIe 4.0 x16 |

| Display Outputs | No outputs | Portable Device Dependent |

| Dimensions | 285 mm x 112 mm | null |

| Release Date | 2021-07-27 | null |

| OpenCL Score | 69000 | 49894 |

| Vulkan Score | null | 49580 |

| Percentile | 90 | 86 |

The two GPUs share identical API support (DirectX 12 Ultimate, OpenGL 4.6, Vulkan 1.4), both are end-of-life, and neither has a launch MSRP recorded. Every other field in the database shows a difference, ranging from the fundamental (architecture, process node, foundry) to the operational (power draw, form factor, interface). The specification table above captures all divergences recorded in the database.