NVIDIA CMP 90HX vs NVIDIA RTX A5500 Mobile Comparison

Where Each One Wins

The recorded benchmark data splits these two Ampere-based NVIDIA cards cleanly. The NVIDIA RTX A5500 Mobile wins the only head-to-head test available, Geekbench OpenCL, with a score of 124287 against 69000 for the NVIDIA CMP 90HX. That is an 80.1% advantage, a dominant margin that places the A5500 Mobile in a different performance tier despite both being built on the same 8 nm Samsung process.

The RTX A5500 Mobile also posts a Geekbench Vulkan score of 103601, a result the CMP 90HX simply cannot match because the CMP 90HX has no display outputs and was not designed for graphics workloads. The database contains no Vulkan score for the CMP 90HX, which is consistent with its role as a compute-only mining part. When you look at average benchmark scores, the A5500 Mobile sits at 113944, while the CMP 90HX averages 69000. The A5500 Mobile lands in the 94th percentile of all GPUs in the database, compared to the 90th percentile for the CMP 90HX.

The CMP 90HX does not win any of the recorded benchmark comparisons. Its only listed score, 69000 in Geekbench OpenCL, places it near the Intel Arc A770 (68809, a 0.3% delta) and the AMD Radeon Instinct MI25 (68562, a 0.6% delta). That grouping shows the CMP 90HX is competitive with mid-range workstation and consumer cards from the same era, but it is not in the same league as the A5500 Mobile. The A5500 Mobile, by contrast, sits within 0.4% of the NVIDIA Tesla V100 SXM2 16 GB (114395) and within 2.7% of the NVIDIA RTX 4000 SFF Ada Generation (117088), while running 2.9% ahead of the AMD Radeon PRO W7900 (110725). Those rival deltas show the A5500 Mobile holding its own against much newer hardware.

For any use case that requires OpenCL compute, the RTX A5500 Mobile is the clear winner. For a pure mining card with no display outputs, the CMP 90HX has no benchmark wins to claim.

Architecture Differences

Both cards use NVIDIA's Ampere architecture and are fabricated on Samsung's 8 nm process, but the underlying chips differ in important ways. The RTX A5500 Mobile uses the GA103 chip with 22,000 million transistors on a 496 mm² die, resulting in a transistor density of 44.4 million per mm². The CMP 90HX uses the larger GA102 chip with 28,300 million transistors on a 628 mm² die, yielding a slightly higher transistor density of 45.1 million per mm². The CMP 90HX physically has more silicon, but the A5500 Mobile is configured for more parallel work.

The A5500 Mobile packs 7424 shading units, 232 texture mapping units, 96 raster operation units, 58 ray tracing cores, and 232 tensor cores. The CMP 90HX has 6400 shading units, 200 texture mapping units, 80 raster operation units, 50 ray tracing cores, and 200 tensor cores. That means the A5500 Mobile has roughly 16% more shaders, 16% more TMUs, 20% more ROPs, and 16% more tensor cores than the CMP 90HX. The A5500 Mobile also carries more ray tracing cores, 58 versus 50, despite being a mobile-oriented part.

Memory architecture differs substantially. The A5500 Mobile uses 16 GB of GDDR6 on a 256-bit bus with 512.0 GB/s of bandwidth. The CMP 90HX uses 10 GB of GDDR6X on a 320-bit bus with 760.3 GB/s of bandwidth. The CMP 90HX wins on raw memory bandwidth by a wide margin, 760.3 GB/s versus 512.0 GB/s, and it uses faster GDDR6X memory rated at 19 Gbps effective versus 16 Gbps effective for the A5500 Mobile's GDDR6. But the A5500 Mobile has 60% more capacity, which matters for larger datasets.

Clock speeds also favor the CMP 90HX. It runs at a 1500 MHz base clock and 1710 MHz boost, while the A5500 Mobile runs at 975 MHz base and 1500 MHz boost. The CMP 90HX's higher clocks and wider memory bus explain its competitive memory throughput, but the A5500 Mobile's higher core counts push its FP32 output to 22.27 TFLOPS versus 21.89 TFLOPS for the CMP 90HX. The FP16 figures are identical to the FP32 numbers on both cards, 22.27 TFLOPS and 21.89 TFLOPS respectively, since both implement 1:1 FP16/FP32 rates.

Power and physical design diverge even further. The A5500 Mobile is rated at 165 W TDP with no power connectors, because it is designed for portable devices and draws power through the motherboard. The CMP 90HX is a 320 W dual-slot card that requires two 8-pin power connectors and a 700 W suggested power supply. The CMP 90HX measures 285 mm in length and 112 mm in height. The A5500 Mobile has no listed dimensions because it is a mobile part.

Interface and outputs also differ. The A5500 Mobile uses PCIe 4.0 x16 and has display outputs described as "Portable Device Dependent." The CMP 90HX uses PCIe 1.0 x4 and has no display outputs at all. Both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Head-to-Head Benchmarks

The only direct benchmark comparison recorded between these two cards is Geekbench OpenCL, and it is not close. The RTX A5500 Mobile scores 124287, while the CMP 90HX scores 69000. That is an 80.1% delta in favor of the A5500 Mobile. For context, the A5500 Mobile's OpenCL result is 20% higher than its own Vulkan score of 103601, which suggests the card is particularly strong in compute-heavy OpenCL workloads.

The A5500 Mobile's average benchmark score of 113944 sits just 0.4% below the Tesla V100 SXM2 16 GB and 2.7% below the RTX 4000 SFF Ada Generation. It beats the AMD Radeon PRO W7900 by 2.9% and the NVIDIA GB10 by 2.9% on average. The CMP 90HX's average of 69000 puts it within 1.4% of the Quadro P6000 (69986) and within 1.2% of the Radeon Pro WX 8200 (69870). Those rival comparisons show the A5500 Mobile competing with much more recent and expensive hardware, while the CMP 90HX trades blows with older workstation cards.

The 80.1% delta is the single largest recorded gap between these two cards, and it appears in the only test where both have data. The A5500 Mobile wins that test by a landslide. The CMP 90HX's higher memory bandwidth, 760.3 GB/s versus 512.0 GB/s, does not translate into a benchmark win in the recorded data. Its higher clocks, 1500 MHz base and 1710 MHz boost, also do not overcome the A5500 Mobile's core count advantage. In practice, the A5500 Mobile's 7424 shading units and 232 tensor cores deliver more usable compute throughput than the CMP 90HX's 6400 shaders and 200 tensor cores, despite the CMP 90HX's higher clocks.

FAQ

Q: Which card has more shading units?

A: The NVIDIA RTX A5500 Mobile has 7424 shading units, while the NVIDIA CMP 90HX has 6400.

Q: Does the CMP 90HX have display outputs?

A: No, the CMP 90HX has no display outputs. The RTX A5500 Mobile has outputs described as "Portable Device Dependent."

Q: Which card has higher memory bandwidth?

A: The CMP 90HX has 760.3 GB/s of bandwidth using GDDR6X on a 320-bit bus. The A5500 Mobile has 512.0 GB/s using GDDR6 on a 256-bit bus.

Q: What is the performance gap in Geekbench OpenCL?

A: The A5500 Mobile scores 124287 versus 69000 for the CMP 90HX, an 80.1% advantage for the A5500 Mobile.

Q: Which card uses more power?

A: The CMP 90HX is rated at 320 W TDP and requires two 8-pin power connectors and a 700 W suggested power supply. The A5500 Mobile is rated at 165 W and uses no power connectors.

Q: Do both cards support the same APIs?

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

The Verdict

The data points to a clear choice for almost every buyer. The NVIDIA RTX A5500 Mobile wins the only head-to-head benchmark, 124287 versus 69000 in Geekbench OpenCL, an 80.1% margin. It also has a Vulkan score of 103601, a 94th percentile ranking, and an average benchmark score of 113944. The CMP 90HX has no benchmark wins, a 90th percentile ranking, and an average score of 69000.

The A5500 Mobile is the better card for any workload that involves OpenCL compute, graphics, or ray tracing. It has more shading units, more texture units, more ROPs, more ray tracing cores, and more tensor cores than the CMP 90HX. It also has 16 GB of memory versus 10 GB, which gives it more headroom for large datasets. The CMP 90HX counters with higher clocks, faster GDDR6X memory, and more bandwidth, but none of that shows up in the recorded benchmark results.

The CMP 90HX makes sense only if you need a mining-specific card with no display outputs, a dual-slot 285 mm form factor, and a 320 W power envelope. Its PCIe 1.0 x4 interface and lack of outputs make it unsuitable for normal desktop use. The A5500 Mobile, despite being a mobile part, outperforms it decisively in the only test both cards share.

Pick the RTX A5500 Mobile for compute, graphics, or any mixed workload. Pick the CMP 90HX only if you have a specific use case that requires its mining-oriented design and higher memory bandwidth, and even then, the benchmark data offers no evidence that it outperforms the A5500 Mobile in any recorded test.

Specification Differences

| Specification | NVIDIA RTX A5500 Mobile | NVIDIA CMP 90HX |

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

| Chip | GA103 | GA102 |

| Process node | 8 nm | 8 nm |

| Transistors | 22,000 million | 28,300 million |

| Die size | 496 mm² | 628 mm² |

| Transistor density | 44.4M / mm² | 45.1M / mm² |

| Base clock | 975 MHz | 1500 MHz |

| Boost clock | 1500 MHz | 1710 MHz |

| Memory clock | 16 Gbps effective | 19 Gbps effective |

| Memory size | 16 GB | 10 GB |

| Memory type | GDDR6 | GDDR6X |

| Memory bus width | 256 bit | 320 bit |

| Memory bandwidth | 512.0 GB/s | 760.3 GB/s |

| Shading units | 7424 | 6400 |

| TMUs | 232 | 200 |

| ROPs | 96 | 80 |

| Ray tracing cores | 58 | 50 |

| Tensor cores | 232 | 200 |

| Pixel rate | 144.0 GPixel/s | 136.8 GPixel/s |

| Texture rate | 348.0 GTexel/s | 342.0 GTexel/s |

| FP32 | 22.27 TFLOPS | 21.89 TFLOPS |

| FP16 | 22.27 TFLOPS (1:1) | 21.89 TFLOPS (1:1) |

| TDP | 165 W | 320 W |

| Slot width | Not listed | Dual-slot |

| Power connectors | None | 2x 8-pin |

| Suggested PSU | Not listed | 700 W |

| Bus interface | PCIe 4.0 x16 | PCIe 1.0 x4 |

| Display outputs | Portable Device Dependent | No outputs |

| Length | Not listed | 285 mm (11.2 inches) |

| Height | Not listed | 112 mm (4.4 inches) |

| Release date | 2022-03-21 | 2021-07-27 |

| Production status | End-of-life | End-of-life |