NVIDIA CMP 90HX vs NVIDIA GeForce RTX 5080 Comparison

Head-to-Head Benchmarks

The only shared benchmark between these two cards in the database is Geekbench OpenCL, and the result is decisive. The NVIDIA GeForce RTX 5080 scores 235901, while the NVIDIA CMP 90HX scores 69000. That translates to a delta of -70.8% for the CMP 90HX, meaning the RTX 5080 is roughly 3.4 times faster in this compute-oriented workload. This is not a narrow margin; it is a generational gap made visible in a single metric.

Looking at the nearest rivals for each card helps contextualize the CMP 90HX’s position. The CMP 90HX sits at the 90th percentile among all GPUs, with an average benchmark score of 69000. Its closest competitors are the Intel Arc A770 at 68809 (a 0.3% delta) and the AMD Radeon Pro WX 8200 at 69870 (a -1.2% delta). In other words, the CMP 90HX is essentially performance-parity with those cards, trailing the AMD Radeon Pro WX 8200 by just 1.2% and the NVIDIA Quadro P6000 by 1.4%. These are small, almost negligible differences in raw OpenCL compute.

The RTX 5080, by contrast, occupies the 87th percentile, with an average benchmark score of 56083 across all its recorded tests. Its nearest rivals include the AMD Radeon 8060S at 55757 (0.6% ahead) and the AMD Radeon RX 9070 GRE at 57367 (2.2% ahead). Note that the RTX 5080’s average score is pulled down by its many PassMark sub-tests, which are not directly comparable to the CMP 90HX’s single OpenCL result. The head-to-head OpenCL score is the cleanest apples-to-apples comparison, and there the RTX 5080 dominates.

The data shows a single win for the RTX 5080 and zero for the CMP 90HX in head-to-head tests. That is not a statistical fluke; it reflects the architectural and process-node advantages that the RTX 5080 brings to the table. The CMP 90HX is not a slow card by 2021 standards, but against a 2025 Blackwell part, the performance delta is stark.

Architecture Differences

The two GPUs come from different eras and foundries. The CMP 90HX uses the GA102 chip, built on Samsung’s 8 nm process, with 28,300 million transistors on a 628 mm² die. That yields a transistor density of 45.1 million per square millimeter. The RTX 5080 uses the GB203 chip, fabricated by TSMC on a 5 nm node, packing 45,600 million transistors into a much smaller 378 mm² die, achieving a density of 120.6 million per square millimeter. The density difference is nearly 2.7 times, which explains a large part of the performance gap.

The CMP 90HX is based on Ampere architecture, while the RTX 5080 is Blackwell 2.0. This is not just a naming difference; the feature set and execution resources are fundamentally different. The CMP 90HX has 6400 shading units, 200 texture mapping units, and 80 ROPs. The RTX 5080 has 10752 shading units, 336 TMUs, and 112 ROPs. That is roughly 68% more shading units, 68% more TMUs, and 40% more ROPs.

Ray tracing and tensor core counts also diverge significantly. The CMP 90HX has 50 RT cores and 200 tensor cores. The RTX 5080 has 84 RT cores and 336 tensor cores. The RTX 5080’s RT core count is 68% higher, and its tensor core count is also 68% higher, matching the shading unit increase. This suggests a uniform scaling of compute resources across the board.

Memory configurations are another major split. The CMP 90HX has 10 GB of GDDR6X on a 320-bit bus, delivering 760.3 GB/s of bandwidth. The RTX 5080 has 16 GB of GDDR7 on a 256-bit bus, but achieves 960.0 GB/s due to faster memory clocks. The RTX 5080’s memory runs at 1875 MHz (30 Gbps effective) versus the CMP 90HX’s 1188 MHz (19 Gbps effective). The result is a 26% bandwidth advantage for the RTX 5080 despite a narrower bus.

Clock speeds also favor the RTX 5080. Its base clock is 2295 MHz and boost is 2617 MHz, compared to the CMP 90HX’s 1500 MHz base and 1710 MHz boost. That is a 53% higher boost clock. Combined with more cores, the RTX 5080 achieves 56.28 TFLOPS FP32 versus 21.89 TFLOPS for the CMP 90HX, a 2.57 times increase.

Pixel and texture rates tell the same story. The RTX 5080 hits 293.1 GPixel/s and 879.3 GTexel/s, while the CMP 90HX manages 136.8 GPixel/s and 342.0 GTexel/s. The RTX 5080 is 2.14 times faster in pixel fill and 2.57 times faster in texture fill.

The bus interface is another notable difference. The CMP 90HX uses PCIe 1.0 x4, which is a severely limited interface for a modern GPU. The RTX 5080 uses PCIe 5.0 x16, offering vastly more bandwidth for data transfer. This alone would bottleneck the CMP 90HX in any workload that requires frequent host-device communication.

Where Each One Wins

The CMP 90HX is a mining-oriented product, and its design reflects that. It has no display outputs, meaning it cannot drive a monitor. It was released in July 2021 as part of NVIDIA’s Mining GPUs generation, and it is now end-of-life. Its strengths lie in raw compute throughput for its era, but it is not designed for gaming or workstation use in a traditional sense.

The RTX 5080 is a mainstream GeForce card, released in January 2025, and it is still in active production. It has 1x HDMI 2.1b and 3x DisplayPort 2.1b outputs, making it suitable for gaming, content creation, and professional visualization. Its higher FP32 and FP16 throughput (both 56.28 TFLOPS) indicates strong compute capability for AI and simulation workloads, not just graphics.

In terms of use cases, the CMP 90HX might still find a niche in compute clusters where display output is irrelevant and the card can be installed in a mining rig or a headless server. Its 320 W TDP and dual-slot design with 2x 8-pin connectors are straightforward for such deployments. However, its PCIe 1.0 x4 interface would hamper any workload that relies on frequent data transfers, such as database processing or real-time inference.

The RTX 5080 is the clear winner for any interactive or graphics-intensive task. Its 16 GB of GDDR7 memory is double the CMP 90HX’s 10 GB, which matters for large textures, high-resolution rendering, or large language model inference. The 84 RT cores provide hardware-accelerated ray tracing, a feature the CMP 90HX technically has but with fewer cores and older architecture. The RTX 5080 also supports PCIe 5.0, which future-proofs it for upcoming motherboards and storage solutions.

For pure compute tasks that fit within the CMP 90HX’s memory capacity and do not require display output, the CMP 90HX could still be a viable option if obtained at a low cost, but the data does not support a performance argument in its favor. The 70.8% deficit in OpenCL is too large to ignore.

The Verdict

The benchmark data is unambiguous: the NVIDIA GeForce RTX 5080 outperforms the NVIDIA CMP 90HX by a massive margin in the only shared test. The RTX 5080’s 235901 OpenCL score versus the CMP 90HX’s 69000 represents a 70.8% deficit for the older card. This is not a close contest.

Architecturally, the RTX 5080 is built on a smaller, denser process node (5 nm versus 8 nm), with more transistors (45,600 million versus 28,300 million), more shading units (10752 versus 6400), more RT cores (84 versus 50), and more tensor cores (336 versus 200). Its memory subsystem delivers 960.0 GB/s versus 760.3 GB/s, and its FP32 throughput is 56.28 TFLOPS versus 21.89 TFLOPS.

The CMP 90HX’s only advantage is its lower TDP (320 W versus 360 W) and its dual 8-pin power connectors, which might be more compatible with older power supplies. But the RTX 5080’s 1x 16-pin connector and 750 W suggested PSU are not unreasonable for a card of its performance class. The RTX 5080 is also longer (304 mm versus 285 mm) and taller (137 mm versus 112 mm), but that is a minor consideration for most cases.

Who should pick the CMP 90HX? Only someone with a specific need for a headless compute card with no display output, and who can tolerate the PCIe 1.0 x4 bottleneck. The data shows it is competitive with the Intel Arc A770 and AMD Radeon Pro WX 8200 in OpenCL, so within its own generation it is not a weak card. But it is end-of-life, and its performance relative to the RTX 5080 is severely outdated.

Who should pick the RTX 5080? Anyone who needs a modern, active GPU with display outputs, ray tracing, high compute throughput, and a large 16 GB memory pool. The RTX 5080 is in the 87th percentile of all GPUs, and its head-to-head result against the CMP 90HX is a 70.8% improvement. For gaming, professional rendering, or AI workloads, the RTX 5080 is the only reasonable choice based on this data.

FAQ

Q: Which GPU has a higher Geekbench OpenCL score?

A: The NVIDIA GeForce RTX 5080 scores 235901, while the NVIDIA CMP 90HX scores 69000, a 70.8% deficit for the CMP 90HX.

Q: How does the CMP 90HX compare to its nearest rivals?

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

Q: What is the memory configuration of each card?

A: The CMP 90HX has 10 GB of GDDR6X on a 320-bit bus, delivering 760.3 GB/s. The RTX 5080 has 16 GB of GDDR7 on a 256-bit bus, delivering 960.0 GB/s.

Q: Does the CMP 90HX support display outputs?

A: No, the CMP 90HX has no display outputs, making it unsuitable for any task requiring a monitor connection.

Q: What is the transistor count and die size difference?

A: The CMP 90HX has 28,300 million transistors on a 628 mm² die (8 nm Samsung). The RTX 5080 has 45,600 million transistors on a 378 mm² die (5 nm TSMC).

Q: Which card has more RT cores?

A: The RTX 5080 has 84 RT cores, while the CMP 90HX has 50 RT cores, a 68% increase for the newer card.

Specification Differences

| Specification | NVIDIA CMP 90HX | NVIDIA GeForce RTX 5080 |

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

| Architecture | Ampere | Blackwell 2.0 |

| Process Node | 8 nm (Samsung) | 5 nm (TSMC) |

| Transistors | 28,300 million | 45,600 million |

| Die Size | 628 mm² | 378 mm² |

| Transistor Density | 45.1M / mm² | 120.6M / mm² |

| Base Clock | 1500 MHz | 2295 MHz |

| Boost Clock | 1710 MHz | 2617 MHz |

| Memory Clock | 1188 MHz (19 Gbps effective) | 1875 MHz (30 Gbps effective) |

| Memory Size | 10 GB | 16 GB |

| Memory Type | GDDR6X | GDDR7 |

| Memory Bus | 320 bit | 256 bit |

| Memory Bandwidth | 760.3 GB/s | 960.0 GB/s |

| Shading Units | 6400 | 10752 |

| TMUs | 200 | 336 |

| ROPs | 80 | 112 |

| RT Cores | 50 | 84 |

| Tensor Cores | 200 | 336 |

| Pixel Rate | 136.8 GPixel/s | 293.1 GPixel/s |

| Texture Rate | 342.0 GTexel/s | 879.3 GTexel/s |

| FP32 | 21.89 TFLOPS | 56.28 TFLOPS |

| FP16 | 21.89 TFLOPS (1:1) | 56.28 TFLOPS (1:1) |

| TDP | 320 W | 360 W |

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

| Suggested PSU | 700 W | 750 W |

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

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

| Dimensions | 285 mm x 112 mm | 304 mm x 137 mm x 40 mm |

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

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