AMD Radeon Pro 575 vs NVIDIA CMP 50HX Comparison

Where Each One Wins

The NVIDIA CMP 50HX wins every recorded head-to-head benchmark against the AMD Radeon Pro 575. In the database, the CMP 50HX claims two wins out of two shared tests, while the Radeon Pro 575 records zero wins. That is a clean sweep, but the shape of the victory matters. The Geekbench OpenCL gap is massive, with the NVIDIA part scoring 62.3% higher. The Vulkan gap is smaller but still decisive at 25.3%. If the workload leans on OpenCL, the CMP 50HX is in a different class. If the workload uses Vulkan, the advantage narrows but does not disappear.

The AMD part does have one recorded benchmark that the NVIDIA card cannot match because the CMP 50HX has no Metal benchmark in the database. The Radeon Pro 575 scores 46192 in Geekbench Metal. That is a meaningful data point for macOS-centric workflows, since Metal is the native graphics API on Apple platforms. The CMP 50HX has no display outputs and no Metal score recorded, so any comparison on that front is impossible from the available data. The Radeon Pro 575 also posts an average benchmark score of 39555, while the CMP 50HX averages 51790. The percentile ranking tells a similar story: the CMP 50HX sits at the 86th percentile of all GPUs, while the Radeon Pro 575 sits at the 82nd percentile. Both are above average, but the NVIDIA part sits higher.

The use-case split is therefore clear. The CMP 50HX is the compute-oriented part that dominates in cross-platform APIs. The Radeon Pro 575 is the mobile, Apple-oriented part whose only unique recorded strength is Metal performance. For general compute, the CMP 50HX wins. For Metal-specific tasks, the Radeon Pro 575 has the only recorded score.

Architecture Differences

The two GPUs come from different architectural generations and different foundries. The NVIDIA CMP 50HX uses the TU102 chip built on Turing architecture, fabricated by TSMC on a 12 nm process. It packs 18,600 million transistors into a 754 mm² die, giving a transistor density of 24.7M per mm². The AMD Radeon Pro 575 uses the Ellesmere chip on GCN 4.0 architecture, fabricated by GlobalFoundries on a 14 nm process. It contains 5,700 million transistors on a 232 mm² die, with a transistor density of 24.6M per mm². The density is nearly identical, which is remarkable given the huge difference in scale. The NVIDIA chip is roughly three times larger in transistor count and die area.

The memory subsystems are completely different. The CMP 50HX has 10 GB of GDDR6 on a 320 bit bus, delivering 560.0 GB/s of bandwidth. The Radeon Pro 575 has 4 GB of GDDR5 on a 256 bit bus, delivering 217.0 GB/s. That is a 2.6x bandwidth advantage for the NVIDIA part, which directly feeds its compute advantage. Memory clock figures also differ: the CMP 50HX runs at 1750 MHz with 14 Gbps effective, while the Radeon Pro 575 runs at 1695 MHz with 6.8 Gbps effective.

Compute resources show a wide gulf. The CMP 50HX has 3584 shading units, 192 texture mapping units, 80 ROPs, 56 RT cores, and 448 tensor cores. The Radeon Pro 575 has 2048 shading units, 128 TMUs, and 32 ROPs, with no RT cores or tensor cores listed. The NVIDIA part also has hardware features for ray tracing and tensor operations that the AMD part lacks entirely. Pixel rate and texture rate tell the same story: 123.6 GPixel/s and 296.6 GTexel/s for the CMP 50HX, versus 35.07 GPixel/s and 140.3 GTexel/s for the Radeon Pro 575.

Power and physical design differ as well. The CMP 50HX is a dual-slot card rated at 250 W, requiring two 8-pin power connectors and a 600 W suggested PSU. It has no display outputs. The Radeon Pro 575 is an MXM module rated at 150 W with no power connectors, and its display outputs are listed as portable device dependent. The bus interfaces are also inverted in an interesting way: the CMP 50HX uses PCIe 1.0 x4, while the Radeon Pro 575 uses PCIe 3.0 x16. That is a curious combination, since the newer, faster card is hobbled by an older, narrower bus interface, which may constrain data transfer in some workloads. The Radeon Pro 575 also supports DirectX 12 (12_0) and Vulkan 1.3, while the CMP 50HX supports DirectX 12 Ultimate (12_2) and Vulkan 1.4. Both support OpenGL 4.6.

Head-to-Head Benchmarks

The Geekbench OpenCL result is the single biggest recorded gap between these two cards. The CMP 50HX scores 56135, while the Radeon Pro 575 scores 34596. That is a 62.3% advantage for the NVIDIA part. OpenCL is a general-purpose compute API, so this result reflects raw throughput across a wide range of workloads. The CMP 50HX has more than 1.7x the shading units, more than 2.5x the bandwidth, and more than 2x the FP32 throughput, so the margin is consistent with the hardware.

The Geekbench Vulkan result is closer but still one-sided. The CMP 50HX scores 47445, while the Radeon Pro 575 scores 37878. That is a 25.3% advantage for the NVIDIA part. Vulkan is a lower-level graphics and compute API, and the narrower gap suggests that the Radeon Pro 575's GCN architecture handles Vulkan relatively better than it handles OpenCL. Still, the CMP 50HX wins.

The average benchmark scores reinforce the head-to-head data. The CMP 50HX averages 51790 across its recorded benchmarks, while the Radeon Pro 575 averages 39555. The CMP 50HX also sits 1.6% above the AMD Radeon RX 6900 XT in its nearest rival list, while the Radeon Pro 575 sits essentially tied with the NVIDIA RTX A500 Mobile at a delta of 0%. The CMP 50HX's nearest rivals include the RX 6900 XT, RX Vega 64, RTX 5070 Ti, and Intel Arc A550M, all within 4.1% of its average score. The Radeon Pro 575's nearest rivals are all pro or mobile parts, including the Pro 575X, Pro WX 7100, and Pro 580, all within 1.9%. The two cards compete in different segments, even though they are being compared directly here.

The Verdict

The data points to one straightforward conclusion for compute: the NVIDIA CMP 50HX is the stronger card. It wins both shared benchmarks, has a higher average score, sits at a higher percentile, and offers substantially more memory bandwidth, shading units, and FP32 throughput. Any workload that can use OpenCL or Vulkan will run faster on the CMP 50HX, with the OpenCL advantage being particularly large at 62.3%.

The AMD Radeon Pro 575 has one unique recorded strength: Metal. Its Geekbench Metal score of 46192 is the only benchmark in the database where it has a result and the CMP 50HX does not. For macOS environments where Metal is the primary API, the Radeon Pro 575 is the only one of the two with recorded data. It is also a much lower power part at 150 W versus 250 W, and it comes as an MXM module with no external power connectors, making it suitable for portable or all-in-one systems. The CMP 50HX has no display outputs and requires a 600 W PSU, so it cannot serve as a standard graphics card for a desktop workstation.

The choice depends on the platform and the API. For compute-heavy workloads on OpenCL or Vulkan, the CMP 50HX is the clear winner. For Metal-based workflows on Apple hardware, the Radeon Pro 575 is the only part with a recorded score. The CMP 50HX is also an end-of-life mining GPU with no display outputs, so its practical use is limited to compute or mining rigs. The Radeon Pro 575 is an end-of-life mobile pro GPU with display support dependent on the host device.

FAQ

Q: Which GPU wins in Geekbench OpenCL?

A: The NVIDIA CMP 50HX scores 56135 versus 34596 for the AMD Radeon Pro 575, a 62.3% advantage.

Q: Which GPU wins in Geekbench Vulkan?

A: The NVIDIA CMP 50HX scores 47445 versus 37878 for the AMD Radeon Pro 575, a 25.3% advantage.

Q: Does the AMD Radeon Pro 575 have any benchmark where it outperforms the CMP 50HX?

A: In the recorded head-to-head tests, no. The Radeon Pro 575 has a Geekbench Metal score of 46192, but the CMP 50HX has no Metal benchmark recorded, so no direct comparison is possible.

Q: How do their average benchmark scores compare?

A: The CMP 50HX averages 51790, while the Radeon Pro 575 averages 39555. The CMP 50HX sits at the 86th percentile of all GPUs, and the Radeon Pro 575 sits at the 82nd percentile.

Q: What memory configurations do they use?

A: The CMP 50HX uses 10 GB of GDDR6 on a 320 bit bus with 560.0 GB/s bandwidth. The Radeon Pro 575 uses 4 GB of GDDR5 on a 256 bit bus with 217.0 GB/s bandwidth.

Q: Do both cards support the same APIs?

A: No. The CMP 50HX supports DirectX 12 Ultimate (12_2) and Vulkan 1.4, while the Radeon Pro 575 supports DirectX 12 (12_0) and Vulkan 1.3. Both support OpenGL 4.6.

Specification Differences

| Specification | NVIDIA CMP 50HX | AMD Radeon Pro 575 |

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

| Architecture | Turing | GCN 4.0 |

| Process node | 12 nm | 14 nm |

| Foundry | TSMC | GlobalFoundries |

| Transistors | 18,600 million | 5,700 million |

| Die size | 754 mm² | 232 mm² |

| Transistor density | 24.7M / mm² | 24.6M / mm² |

| Memory size | 10 GB | 4 GB |

| Memory type | GDDR6 | GDDR5 |

| Memory bus width | 320 bit | 256 bit |

| Memory bandwidth | 560.0 GB/s | 217.0 GB/s |

| Memory clock | 1750 MHz, 14 Gbps effective | 1695 MHz, 6.8 Gbps effective |

| Shading units | 3584 | 2048 |

| TMUs | 192 | 128 |

| ROPs | 80 | 32 |

| RT cores | 56 | None |

| Tensor cores | 448 | None |

| Pixel rate | 123.6 GPixel/s | 35.07 GPixel/s |

| Texture rate | 296.6 GTexel/s | 140.3 GTexel/s |

| FP32 | 11.07 TFLOPS | 4.489 TFLOPS |

| FP16 | 22.15 TFLOPS (2:1) | 4.489 TFLOPS (1:1) |

| TDP | 250 W | 150 W |

| Slot width | Dual-slot | MXM Module |

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

| Suggested PSU | 600 W | Not listed |

| Bus interface | PCIe 1.0 x4 | PCIe 3.0 x16 |

| Display outputs | No outputs | Portable Device Dependent |

| DirectX support | 12 Ultimate (12_2) | 12 (12_0) |

| Vulkan support | 1.4 | 1.3 |

| Dimensions | 267 mm x 116 mm x 35 mm | Not listed |

| Release date | 2021-06-23 | 2017-06-04 |