NVIDIA CMP 30HX vs NVIDIA GeForce RTX 5090 Comparison

The NVIDIA GeForce RTX 5090 utterly dominates the NVIDIA CMP 30HX in every measurable benchmark category. In the only two shared tests, the RTX 5090 delivers leads of 412.8% and 502.9%, making the comparison less a contest and more a demonstration of generational and architectural gulf. The data shows a clear verdict: the RTX 5090 is a flagship consumer and professional compute card, while the CMP 30HX is a niche, end-of-life mining product with a fraction of the capability.

Head-to-Head Benchmarks

The head-to-head data presents a one-sided picture, with the RTX 5090 winning both available comparisons. In the Geekbench OpenCL test, the RTX 5090 scores 334,370 against the CMP 30HX’s 65,199. This translates to a delta of 412.8%, meaning the RTX 5090 is over five times faster in raw compute throughput. This is not a marginal improvement; it is a categorical leap that reflects the massive disparity in shading units (21,760 vs 1,408) and memory bandwidth (1.79 TB/s vs 336.0 GB/s).

The Vulkan benchmark tells an even more extreme story. The RTX 5090 posts 376,728 points, while the CMP 30HX manages just 62,484. The delta here is 502.9%, indicating the RTX 5090 is roughly six times faster in this API. Vulkan’s low-level nature often highlights architectural efficiency, and the RTX 5090’s Blackwell 2.0 architecture with dedicated RT and Tensor cores (170 and 680, respectively) simply overwhelms the Turing-based CMP 30HX, which has no such dedicated hardware. The CMP 30HX’s scores are so far behind that they fall outside the RTX 5090’s nearest rival delta range—the RTX 5090’s closest competitor in average score, the AMD Radeon Pro Vega 64X, trails by only 1.4%, while the CMP 30HX’s closest rival, the AMD Radeon RX 9060 XT LP, is within 0% delta.

Looking at the broader benchmark suite, the RTX 5090’s average benchmark score is 79,842, placing it in the 92nd percentile of all GPUs. The CMP 30HX averages 63,842, sitting in the 89th percentile. While the percentile gap looks modest, the absolute score difference is substantial. The RTX 5090 also has a full suite of PassMark results—DirectX 9 (395), DirectX 10 (226), DirectX 11 (341), DirectX 12 (185), G2D (1,413), G3D (39,650), and GPU Compute (26,756)—whereas the CMP 30HX has no PassMark data recorded. This absence itself is telling: the CMP 30HX was never designed for the standard graphics workloads these tests measure.

Where Each One Wins

The RTX 5090 wins everywhere the data provides a comparison. In compute-heavy APIs like OpenCL and Vulkan, its advantage is overwhelming, with deltas exceeding 400%. Its FP32 throughput of 104.8 TFLOPS dwarfs the CMP 30HX’s 5.027 TFLOPS, making it the obvious choice for any parallel processing task, from rendering to scientific simulation. The RTX 5090 also supports FP16 at a 1:1 ratio (104.8 TFLOPS), while the CMP 30HX’s FP16 (10.05 TFLOPS) is achieved at a 2:1 ratio, meaning the RTX 5090’s half-precision performance is not just higher but also more efficient in terms of core utilization.

The CMP 30HX has no benchmark wins to claim. Its only advantages are in physical specifications that favor low power draw and a mining-specific design. It has a TDP of 125 W versus the RTX 5090’s 575 W, and its suggested PSU is 300 W versus 950 W. For a use case where compute per watt is paramount and display output is irrelevant, the CMP 30HX’s 6 GB of GDDR6 memory on a 192-bit bus provides 336.0 GB/s of bandwidth—sufficient for its intended mining role but trivial compared to the RTX 5090’s 32 GB of GDDR7 on a 512-bit bus. The CMP 30HX also has no display outputs, making it unusable for any gaming or graphics workstation scenario. In every functional category that matters to a general-purpose GPU buyer, the RTX 5090 wins decisively.

Architecture Differences

The two cards represent opposite ends of NVIDIA’s design philosophy. The RTX 5090 uses the GB202 chip on a 5 nm TSMC process, packing 92,200 million transistors into a 750 mm² die. This yields a transistor density of 122.9M per mm². The architecture is Blackwell 2.0, the latest generation, and it includes 170 RT cores and 680 Tensor cores, enabling hardware-accelerated ray tracing and AI workloads. The CMP 30HX, by contrast, uses the TU116 chip on a 12 nm TSMC process, with just 6,600 million transistors on a 284 mm² die—a density of only 23.2M per mm². Its Turing architecture predates Blackwell by several generations and lacks RT and Tensor cores entirely, as evidenced by null values for those fields.

The memory subsystems are equally divergent. The RTX 5090 features 32 GB of GDDR7 memory with a 512-bit bus, delivering 1.79 TB/s of bandwidth. The CMP 30HX has 6 GB of GDDR6 on a 192-bit bus, yielding 336.0 GB/s. This is a 5.3x difference in bandwidth, which directly impacts memory-bound workloads. Clock speeds also differ: the RTX 5090 boosts to 2407 MHz, while the CMP 30HX boosts to 1785 MHz. The RTX 5090’s base clock of 2017 MHz is higher than the CMP 30HX’s boost clock, underscoring the efficiency of the smaller process node.

Other architectural details reinforce the gap. The RTX 5090 has 680 TMUs and 176 ROPs, versus 88 TMUs and 48 ROPs on the CMP 30HX. Pixel rate is 423.6 GPixel/s versus 85.68 GPixel/s, and texture rate is 1,636.8 GTexel/s versus 157.1 GTexel/s. The RTX 5090 supports PCIe 5.0 x16, while the CMP 30HX is limited to PCIe 1.0 x4—a severely outdated interface that would bottleneck even modest data transfers. The CMP 30HX’s power connector is a single 8-pin, whereas the RTX 5090 requires a 16-pin connector, reflecting the former’s 125 W TDP and the latter’s 575 W TDP. The RTX 5090 also has display outputs (1x HDMI 2.1b and 3x DisplayPort 2.1b), while the CMP 30HX has none.

The Verdict

The data leaves no room for ambiguity. The RTX 5090 is the superior product by every performance metric recorded, with benchmark deltas of 412.8% and 502.9% over the CMP 30HX in OpenCL and Vulkan, respectively. Its 92nd percentile ranking versus the CMP 30HX’s 89th percentile further confirms its higher standing among all GPUs. The RTX 5090 should be chosen by anyone needing maximum compute performance, ray tracing capability, or large memory capacity for modern workloads. Its 32 GB of GDDR7 memory and 104.8 TFLOPS of FP32 performance make it suitable for high-end gaming, content creation, and AI research. Its active production status and 2025 release date indicate it is a current-generation product.

The CMP 30HX, however, is a specialized mining card that is now end-of-life, having been released in 2021. Its 6 GB memory and 5.027 TFLOPS are far below modern standards, and its lack of display outputs means it cannot function as a standard graphics card. Its only advantages are lower power consumption (125 W vs 575 W) and a lower suggested PSU (300 W vs 950 W), which matter only in mining rigs where efficiency per watt is critical. For any other purpose, the RTX 5090 is the obvious choice. If the requirement is purely a low-power compute card for a mining operation, the CMP 30HX’s data shows it is adequate, but for any general-purpose or gaming task, the RTX 5090 wins outright.

FAQ

Q: How much faster is the RTX 5090 than the CMP 30HX in Geekbench OpenCL?

A: The RTX 5090 scores 334,370 versus the CMP 30HX’s 65,199, a delta of 412.8%.

Q: Does the CMP 30HX support display outputs?

A: No, the CMP 30HX has no display outputs, while the RTX 5090 offers 1x HDMI 2.1b and 3x DisplayPort 2.1b.

Q: What are the memory configurations of each card?

A: The RTX 5090 has 32 GB of GDDR7 on a 512-bit bus with 1.79 TB/s bandwidth. The CMP 30HX has 6 GB of GDDR6 on a 192-bit bus with 336.0 GB/s bandwidth.

Q: Which card has dedicated ray tracing cores?

A: The RTX 5090 has 170 RT cores and 680 Tensor cores. The CMP 30HX has no RT or Tensor cores.

Q: What is the average benchmark score difference between the two?

A: The RTX 5090 averages 79,842, while the CMP 30HX averages 63,842. The RTX 5090 ranks in the 92nd percentile of all GPUs, versus the 89th percentile for the CMP 30HX.

Q: What is the TDP of each card?

A: The RTX 5090 has a TDP of 575 W with a suggested PSU of 950 W. The CMP 30HX has a TDP of 125 W with a suggested PSU of 300 W.

Specification Differences

| Specification | NVIDIA GeForce RTX 5090 | NVIDIA CMP 30HX |

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

| Chip | GB202 | TU116 |

| Architecture | Blackwell 2.0 | Turing |

| Process Node | 5 nm | 12 nm |

| Transistors | 92,200 million | 6,600 million |

| Die Size | 750 mm² | 284 mm² |

| Base Clock | 2017 MHz | 1530 MHz |

| Boost Clock | 2407 MHz | 1785 MHz |

| Memory Size | 32 GB | 6 GB |

| Memory Type | GDDR7 | GDDR6 |

| Memory Bus | 512 bit | 192 bit |

| Memory Bandwidth | 1.79 TB/s | 336.0 GB/s |

| Shading Units | 21760 | 1408 |

| TMUs | 680 | 88 |

| ROPs | 176 | 48 |

| RT Cores | 170 | None |

| Tensor Cores | 680 | None |

| Pixel Rate | 423.6 GPixel/s | 85.68 GPixel/s |

| Texture Rate | 1,636.8 GTexel/s | 157.1 GTexel/s |

| FP32 Performance | 104.8 TFLOPS | 5.027 TFLOPS |

| FP16 Performance | 104.8 TFLOPS (1:1) | 10.05 TFLOPS (2:1) |

| TDP | 575 W | 125 W |

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

| Suggested PSU | 950 W | 300 W |

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

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

| DirectX Support | 12 Ultimate (12_2) | 12 (12_1) |

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

| Release Date | 2025-01-29 | 2021-02-24 |