NVIDIA CMP 90HX vs NVIDIA GeForce RTX 5090 Comparison

The NVIDIA GeForce RTX 5090 is the clear performance leader in this comparison, but the NVIDIA CMP 90HX is a fundamentally different product with a specific historical purpose. The RTX 5090 dominates the only shared benchmark, delivering a 384.6% higher Geekbench OpenCL score, yet the CMP 90HX remains a distinct entry in the database due to its mining-specific design. The data shows that these two GPUs are separated by architecture, memory configuration, and intended use case, with the RTX 5090 being the only viable option for general computing and graphics workloads.

Head-to-Head Benchmarks

The single direct comparison available in the dataset is the Geekbench OpenCL test. Here, the NVIDIA GeForce RTX 5090 scores 334,370, while the NVIDIA CMP 90HX scores 69,000. This represents a delta of 384.6% in favor of the RTX 5090, a massive margin that reflects the generational and architectural gap between the two cards. The RTX 5090’s score places it in the 92nd percentile among all GPUs, while the CMP 90HX sits at the 90th percentile, but the raw score difference is the decisive factor.

The RTX 5090’s average benchmark score across all tests is 79,842, which is 0.3% ahead of the NVIDIA Tesla P100 PCIe 16 GB (79,605) and 1.1% ahead of the AMD Radeon RX 6850M XT (78,940). It trails the AMD Radeon Pro Vega 64X by 1.4%, which scores 80,959. The CMP 90HX has an average score of 69,000, which is 0.3% ahead of the Intel Arc A770 (68,809) and 0.6% ahead of the AMD Radeon Instinct MI25 (68,562), but it falls 1.2% behind the AMD Radeon Pro WX 8200 (69,870) and 1.4% behind the NVIDIA Quadro P6000 (69,986). These rival comparisons show that the RTX 5090 competes in a higher performance tier, while the CMP 90HX sits closer to mid-range workstation and consumer cards.

In other benchmark categories, the RTX 5090 demonstrates its breadth of capability. It scores 18,355 in 3DMark Steel Nomad DX12, 376,728 in Geekbench Vulkan, and 39,650 in Passmark G3D. Its Passmark GPU Compute score is 26,756, and it reaches 1,413 in Passmark G2D. The CMP 90HX has no results for these tests, so the RTX 5090 wins all available comparisons by default. The only test where the CMP 90HX appears is Geekbench OpenCL, and the RTX 5090 wins that contest by a wide margin.

Architecture Differences

The RTX 5090 is built on the Blackwell 2.0 architecture using the GB202 chip, manufactured on a 5 nm process at TSMC. It integrates 92,200 million transistors on a 750 mm² die, resulting in a transistor density of 122.9 million per mm². The CMP 90HX, by contrast, uses the Ampere architecture with the GA102 chip, built on an 8 nm process at Samsung. It contains 28,300 million transistors on a 628 mm² die, with a density of 45.1 million per mm². The RTX 5090’s process advantage is clear: a smaller node and higher density enable more compute resources in a similar physical footprint.

The memory subsystems are vastly different. The RTX 5090 features 32 GB of GDDR7 memory on a 512-bit bus, delivering 1.79 TB/s of bandwidth. The CMP 90HX has 10 GB of GDDR6X memory on a 320-bit bus, with 760.3 GB/s of bandwidth. The RTX 5090’s memory clock is 1750 MHz (28 Gbps effective), while the CMP 90HX runs at 1188 MHz (19 Gbps effective). This gives the RTX 5090 more than double the memory bandwidth, which is critical for high-resolution textures and compute workloads.

Core counts also favor the RTX 5090. It has 21,760 shading units, 680 TMUs, 176 ROPs, 170 RT cores, and 680 tensor cores. The CMP 90HX has 6,400 shading units, 200 TMUs, 80 ROPs, 50 RT cores, and 200 tensor cores. Pixel and texture rates follow suit: the RTX 5090 achieves 423.6 GPixel/s and 1,636.8 GTexel/s, while the CMP 90HX manages 136.8 GPixel/s and 342.0 GTexel/s. FP32 and FP16 performance are both rated at 104.8 TFLOPS for the RTX 5090, versus 21.89 TFLOPS for the CMP 90HX.

The bus interface differs significantly. The RTX 5090 uses PCIe 5.0 x16, while the CMP 90HX uses PCIe 1.0 x4, a legacy interface that severely limits data transfer rates. The CMP 90HX has no display outputs, confirming its mining-only design, whereas the RTX 5090 offers 1x HDMI 2.1b and 3x DisplayPort 2.1b outputs.

Where Each One Wins

The RTX 5090 wins every benchmark category where data exists. Its strongest advantage is in compute-heavy workloads, as shown by the 384.6% lead in Geekbench OpenCL. This makes it the superior choice for tasks like 3D rendering, video encoding, scientific simulation, and AI inference, where the 32 GB memory capacity and 1.79 TB/s bandwidth are decisive. The RTX 5090’s 92nd percentile ranking among all GPUs reinforces that it is a top-tier performer for professional and enthusiast applications.

The CMP 90HX does not win any benchmark comparisons, but its design points to a specific use case: cryptocurrency mining. The lack of display outputs and the PCIe 1.0 x4 interface indicate that it was engineered solely for compute tasks that do not require display output or high-bandwidth host communication. Its 90th percentile ranking is respectable, but the 69,000 Geekbench OpenCL score is comparable to mid-range cards like the Intel Arc A770 (68,809) and AMD Radeon Instinct MI25 (68,562). In practical terms, the CMP 90HX is not suited for gaming, content creation, or any workload requiring visual output.

For users with general-purpose needs, the RTX 5090 is the only choice. It offers modern API support including DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, matching the CMP 90HX’s API list, but with the hardware to actually use those features. The CMP 90HX also supports these APIs, but its lower core counts and memory bandwidth limit its effectiveness in real-world applications.

FAQ

Q: What is the performance difference in the only shared benchmark?

A: The RTX 5090 scores 334,370 in Geekbench OpenCL, while the CMP 90HX scores 69,000. The RTX 5090 is 384.6% faster, a massive margin driven by its newer architecture and higher resource counts.

Q: How does the RTX 5090 compare to its nearest rivals?

A: The RTX 5090’s average score of 79,842 is 0.3% ahead of the Tesla P100 PCIe 16 GB (79,605), 0.6% ahead of the Tesla P100 PCIe 12 GB (79,396), and 1.1% ahead of the RX 6850M XT (78,940). It is 1.4% behind the Radeon Pro Vega 64X (80,959).

Q: What are the memory differences between the two cards?

A: The RTX 5090 has 32 GB of GDDR7 on a 512-bit bus with 1.79 TB/s bandwidth. The CMP 90HX has 10 GB of GDDR6X on a 320-bit bus with 760.3 GB/s bandwidth.

Q: Is the CMP 90HX suitable for gaming?

A: No. The CMP 90HX has no display outputs, making it impossible to connect to a monitor. Its PCIe 1.0 x4 interface also limits data throughput, further reducing its utility for gaming.

Q: What is the production status of each card?

A: The RTX 5090 is listed as Active, while the CMP 90HX is End-of-life.

Q: How do the transistor counts compare?

A: The RTX 5090 has 92,200 million transistors on a 5 nm process, while the CMP 90HX has 28,300 million on an 8 nm process. The RTX 5090’s density is 122.9M per mm² versus 45.1M per mm² for the CMP 90HX.

The Verdict

The data is unambiguous: the RTX 5090 is the superior GPU in every measurable way. Its 384.6% lead in Geekbench OpenCL, combined with 32 GB of memory versus 10 GB, and a modern PCIe 5.0 x16 interface versus PCIe 1.0 x4, makes it the only rational choice for any user needing a functional graphics card. The RTX 5090’s 92nd percentile ranking and high average score of 79,842 place it among the fastest GPUs in the database, while the CMP 90HX’s 90th percentile and 69,000 average score reflect a more modest capability.

The CMP 90HX is an end-of-life product designed for a niche mining market. Its lack of display outputs and legacy bus interface mean it cannot serve as a desktop GPU. Even in its intended compute role, its 21.89 TFLOPS FP32 performance is dwarfed by the RTX 5090’s 104.8 TFLOPS. For anyone building a system, the RTX 5090 is the clear pick, provided its 575 W TDP and 950 W suggested PSU are acceptable. The CMP 90HX, with its 320 W TDP and 700 W suggested PSU, is a historical artifact rather than a competitive alternative.

Specification Differences

| Specification | NVIDIA GeForce RTX 5090 | NVIDIA CMP 90HX |

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

| Architecture | Blackwell 2.0 | Ampere |

| Process Node | 5 nm | 8 nm |

| Foundry | TSMC | Samsung |

| Transistors | 92,200 million | 28,300 million |

| Die Size | 750 mm² | 628 mm² |

| Transistor Density | 122.9M / mm² | 45.1M / mm² |

| Base Clock | 2017 MHz | 1500 MHz |

| Boost Clock | 2407 MHz | 1710 MHz |

| Memory Size | 32 GB | 10 GB |

| Memory Type | GDDR7 | GDDR6X |

| Memory Bus | 512 bit | 320 bit |

| Memory Bandwidth | 1.79 TB/s | 760.3 GB/s |

| Shading Units | 21760 | 6400 |

| TMUs | 680 | 200 |

| ROPs | 176 | 80 |

| RT Cores | 170 | 50 |

| Tensor Cores | 680 | 200 |

| Pixel Rate | 423.6 GPixel/s | 136.8 GPixel/s |

| Texture Rate | 1,636.8 GTexel/s | 342.0 GTexel/s |

| FP32 | 104.8 TFLOPS | 21.89 TFLOPS |

| FP16 | 104.8 TFLOPS (1:1) | 21.89 TFLOPS (1:1) |

| TDP | 575 W | 320 W |

| Power Connectors | 1x 16-pin | 2x 8-pin |

| Suggested PSU | 950 W | 700 W |

| Bus Interface | PCIe 5.0 x16 | PCIe 1.0 x4 |

| Display Outputs | 1x HDMI 2.1b, 3x DisplayPort 2.1b | No outputs |

| Production Status | Active | End-of-life |

| Release Date | 2025-01-29 | 2021-07-27 |

| Launch MSRP | 1,999 USD | None |