AMD Radeon Pro WX 9100 vs NVIDIA CMP 90HX Comparison

The benchmark data clearly separates these two professional-grade GPUs, with the NVIDIA CMP 90HX taking the only direct head-to-head win, though the AMD Radeon Pro WX 9100 counters with a broader software ecosystem and a significantly higher peak score in a different API. The CMP 90HX posts a Geekbench OpenCL score of 69,000 against the WX 9100’s 66,605, a 3.6% margin that places it in the 90th percentile of all GPUs, while the AMD card sits at the 89th percentile with an average benchmark score of 64,212 across three tests.

Head-to-Head Benchmarks

The sole direct comparison available is Geekbench OpenCL, and the NVIDIA CMP 90HX wins decisively. Its score of 69,000 beats the AMD Radeon Pro WX 9100’s 66,605 by exactly 3.6%, a modest but consistent advantage. This margin is not a fluke of a single workload; the CMP 90HX’s average benchmark score of 69,000 is identical to its OpenCL result, indicating uniform performance across its tested workload. The WX 9100, by contrast, shows more variance: its OpenCL score of 66,605 is its middle result, sitting between a Metal score of 71,319 and a Vulkan score of 54,711.

The 3.6% gap in OpenCL is smaller than what the raw specifications might suggest. The CMP 90HX delivers 21.89 TFLOPS of FP32 compute versus the WX 9100’s 12.29 TFLOPS, a 78% raw throughput difference that does not translate into a proportional benchmark lead. This suggests the WX 9100’s architecture extracts more real-world efficiency from its compute units, or that the OpenCL test is not fully stressing the CMP 90HX’s strengths. Regardless, the CMP 90HX wins the only head-to-head test, giving it a 1-0 record in direct matchups.

Context from the nearest rivals reinforces the CMP 90HX’s position. Its 69,000 score places it 0.3% ahead of the Intel Arc A770 (68,809) and 0.6% ahead of the AMD Radeon Instinct MI25 (68,562), while trailing the AMD Radeon Pro WX 8200 by 1.2% and the NVIDIA Quadro P6000 by 1.4%. This cluster of scores within roughly 2% of each other shows that the CMP 90HX sits in a highly competitive performance tier, not a dominant one. The WX 9100’s nearest rivals tell a similar story: it leads the NVIDIA CMP 30HX by 0.6% (64,212 vs 63,842), the AMD Radeon RX 9060 XT LP by 0.6%, and the AMD Radeon RX 7600M by 0.7%, with the AMD Radeon Pro Vega 56 trailing by 0.8%.

The WX 9100’s Metal score of 71,319 is the single highest benchmark result between the two cards, surpassing the CMP 90HX’s best by 3.4%. This is a critical data point: while the CMP 90HX wins in OpenCL, the WX 9100 demonstrates superior performance in Apple’s Metal API, likely reflecting its broader driver optimization for that platform. Its Vulkan score of 54,711 is substantially lower, 17.9% below its own OpenCL result and 20.7% below the CMP 90HX’s OpenCL score, indicating that Vulkan workloads are not this card’s strength.

The Verdict

The data supports a clear split decision. The NVIDIA CMP 90HX is the better choice for users whose workloads are dominated by OpenCL compute, as evidenced by its 3.6% head-to-head win and its top-10-percentile ranking. Its consistent 69,000 score across its tested benchmark suggests predictable performance without API-dependent swings. The AMD Radeon Pro WX 9100, however, is the more versatile card for mixed-API environments, particularly where Metal is used, given its 71,319 Metal score that exceeds anything the CMP 90HX can produce in OpenCL.

For raw compute density, the CMP 90HX is the clear winner on paper. Its 21.89 TFLOPS FP32 output is 78% higher than the WX 9100’s 12.29 TFLOPS, and its FP16 performance of 21.89 TFLOPS (1:1) matches its FP32, whereas the WX 9100’s FP16 of 24.58 TFLOPS (2:1) is actually higher than its FP32. This means the AMD card has a 12.3% FP16 advantage over the NVIDIA card in peak throughput, despite the NVIDIA card’s FP32 lead. For workloads that leverage FP16, the WX 9100 may close or even reverse the gap seen in OpenCL.

The WX 9100 also offers more memory: 16 GB of HBM2 versus 10 GB of GDDR6X. While the CMP 90HX has the higher bandwidth at 760.3 GB/s versus 483.8 GB/s, the AMD card’s larger capacity can hold larger datasets without spilling to system memory. The CMP 90HX’s 320-bit bus and GDDR6X memory deliver 57% more bandwidth, but the WX 9100’s 2048-bit HBM2 interface provides a different capacity-versus-speed tradeoff. Users with memory-hungry workloads may prefer the WX 9100 despite its lower bandwidth.

Architecture Differences

The two cards are built on fundamentally different architectures from different nodes and foundries. The NVIDIA CMP 90HX uses the GA102 chip on Samsung’s 8 nm process, packing 28,300 million transistors into a 628 mm² die for a density of 45.1 million transistors per square millimeter. The AMD Radeon Pro WX 9100 uses the Vega 10 chip on GlobalFoundries’ 14 nm process, with 12,500 million transistors on a 495 mm² die, yielding a density of 25.3 million transistors per square millimeter. The NVIDIA chip is 2.26 times larger in transistor count and has 78% more die area, enabling its higher compute throughput.

The CMP 90HX is built on the Ampere architecture with 6,400 shading units, 200 texture mapping units, and 80 render output units. It also includes 50 ray tracing cores and 200 tensor cores, making it a fully featured modern GPU despite its mining-oriented branding. The WX 9100 uses the older GCN 5.0 architecture with 4,096 shading units, 256 TMUs, and 64 ROPs. It has no ray tracing cores and no tensor cores, reflecting its 2017 design. The AMD card compensates with more TMUs (256 vs 200), giving it a higher texture rate of 384.0 GTexel/s versus the CMP 90HX’s 342.0 GTexel/s, despite the NVIDIA card’s higher pixel rate of 136.8 GPixel/s versus 96.00 GPixel/s.

Memory technologies differ markedly. The CMP 90HX uses 10 GB of GDDR6X on a 320-bit bus with a 760.3 GB/s bandwidth. The WX 9100 uses 16 GB of HBM2 on a 2048-bit bus with a 483.8 GB/s bandwidth. The NVIDIA card’s memory runs at 1188 MHz (19 Gbps effective) while the AMD card’s runs at 945 MHz (1890 Mbps effective). The CMP 90HX has a 57% bandwidth advantage but 37.5% less capacity. Clock speeds also differ: the CMP 90HX runs at 1500 MHz base and 1710 MHz boost, while the WX 9100 runs at 1200 MHz base and 1500 MHz boost, a 14% and 14% difference respectively.

Power and interface specifications diverge as well. The CMP 90HX has a 320 W TDP with dual 8-pin power connectors and requires a 700 W power supply. The WX 9100 has a 230 W TDP with a 6-pin plus 8-pin configuration and requires a 550 W power supply. The NVIDIA card uses a PCIe 1.0 x4 interface, which is a severe bottleneck for a modern GPU, while the AMD card uses PCIe 3.0 x16. The CMP 90HX has no display outputs, while the WX 9100 offers 6x mini-DisplayPort 1.4a, making the AMD card usable for visualization tasks.

FAQ

Q: Which card has higher FP32 compute performance?

A: The NVIDIA CMP 90HX with 21.89 TFLOPS versus the AMD Radeon Pro WX 9100’s 12.29 TFLOPS, a 78% advantage for the NVIDIA card.

Q: What is the memory size difference?

A: The AMD Radeon Pro WX 9100 has 16 GB of HBM2, while the NVIDIA CMP 90HX has 10 GB of GDDR6X. The AMD card offers 60% more capacity.

Q: How do they compare in OpenCL benchmarks?

A: The NVIDIA CMP 90HX scores 69,000 versus the AMD Radeon Pro WX 9100’s 66,605, giving the NVIDIA card a 3.6% lead in the only direct head-to-head test.

Q: Which card supports ray tracing?

A: Only the NVIDIA CMP 90HX, which includes 50 ray tracing cores. The AMD Radeon Pro WX 9100 has no ray tracing cores.

Q: What is the power consumption difference?

A: The NVIDIA CMP 90HX has a 320 W TDP, while the AMD Radeon Pro WX 9100 has a 230 W TDP. The NVIDIA card requires a 700 W power supply versus 550 W for the AMD card.

Q: Does either card support display outputs?

A: Only the AMD Radeon Pro WX 9100, which offers 6x mini-DisplayPort 1.4a. The NVIDIA CMP 90HX has no display outputs.

Where Each One Wins

The NVIDIA CMP 90HX wins in raw compute throughput, memory bandwidth, and modern feature support. Its 21.89 TFLOPS FP32 and 21.89 TFLOPS FP16 (1:1) deliver double the FP32 of the AMD card and nearly matching FP16 despite the AMD card’s higher 24.58 TFLOPS FP16 (2:1). Its 760.3 GB/s bandwidth is 57% higher, and its 136.8 GPixel/s pixel rate is 42% higher. The inclusion of 50 ray tracing cores and 200 tensor cores makes it future-proof for compute workloads that leverage these features. Its 69,000 OpenCL score places it in the 90th percentile of all GPUs, and its 3.6% win over the WX 9100 in the direct head-to-head confirms its compute superiority.

The AMD Radeon Pro WX 9100 wins in memory capacity, API versatility, and practical deployability. Its 16 GB of HBM2 exceeds the NVIDIA card’s 10 GB by 60%, and its PCIe 3.0 x16 interface is far superior to the CMP 90HX’s PCIe 1.0 x4, which severely limits data transfer for real-world workloads. Its 6x mini-DisplayPort outputs enable multi-display setups, while the NVIDIA card offers none. The WX 9100’s 71,319 Metal score is the highest single benchmark result between the two cards, and its 384.0 GTexel/s texture rate exceeds the NVIDIA card’s 342.0 GTexel/s. Its lower 230 W TDP and 550 W power supply requirement make it easier to integrate into existing systems.

Specification Differences

| Specification | NVIDIA CMP 90HX | AMD Radeon Pro WX 9100 |

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

| Architecture | Ampere | GCN 5.0 |

| Process Node | 8 nm | 14 nm |

| Foundry | Samsung | GlobalFoundries |

| Transistors | 28,300 million | 12,500 million |

| Die Size | 628 mm² | 495 mm² |

| Transistor Density | 45.1M / mm² | 25.3M / mm² |

| Base Clock | 1500 MHz | 1200 MHz |

| Boost Clock | 1710 MHz | 1500 MHz |

| Memory Size | 10 GB | 16 GB |

| Memory Type | GDDR6X | HBM2 |

| Memory Bus | 320 bit | 2048 bit |

| Memory Bandwidth | 760.3 GB/s | 483.8 GB/s |

| Shading Units | 6400 | 4096 |

| TMUs | 200 | 256 |

| ROPs | 80 | 64 |

| RT Cores | 50 | None |

| Tensor Cores | 200 | None |

| Pixel Rate | 136.8 GPixel/s | 96.00 GPixel/s |

| Texture Rate | 342.0 GTexel/s | 384.0 GTexel/s |

| FP32 | 21.89 TFLOPS | 12.29 TFLOPS |

| FP16 | 21.89 TFLOPS (1:1) | 24.58 TFLOPS (2:1) |

| TDP | 320 W | 230 W |

| Power Connectors | 2x 8-pin | 1x 6-pin + 1x 8-pin |

| Suggested PSU | 700 W | 550 W |

| Bus Interface | PCIe 1.0 x4 | PCIe 3.0 x16 |

| Display Outputs | None | 6x mini-DisplayPort 1.4a |

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

| Vulkan | 1.4 | 1.3 |

| Release Date | 2021-07-27 | 2017-07-09 |

| Launch MSRP | Not available | 1,599 USD |

| Average Benchmark Score | 69,000 | 64,212 |