AMD Radeon Pro WX 9100 vs NVIDIA CMP 40HX Comparison

The NVIDIA CMP 40HX and AMD Radeon Pro WX 9100 occupy very different corners of the hardware landscape, one built for a specialized mining task and the other for professional visualization. The database shows a clear performance hierarchy, but the story is more nuanced than a simple win/loss tally.

Head-to-Head Benchmarks

The recorded head-to-head results are unambiguous. In Geekbench OpenCL, the NVIDIA CMP 40HX scores 93,395 against the AMD Radeon Pro WX 9100's 66,605. That is a 40.2% advantage for the NVIDIA part. The gap widens slightly in Geekbench Vulkan, where the CMP 40HX posts 77,879 versus 54,711 for the WX 9100, a 42.3% lead. The NVIDIA card wins both recorded comparisons, giving it a 2-0 sweep in direct matchups.

These are not marginal victories. The 40.2% and 42.3% deltas indicate a generational leap in compute efficiency, even though the AMD card carries far more raw hardware. The CMP 40HX achieves this with 2,304 shading units, while the WX 9100 has 4,096. The NVIDIA card's higher clock speeds, 1470 MHz base and 1650 MHz boost versus 1200 MHz base and 1500 MHz boost for AMD, help explain part of the gap, but the architecture itself appears to extract more work per cycle.

The average benchmark score tells a similar story. The CMP 40HX sits at 85,637, placing it in the 93rd percentile of all GPUs in the database. The WX 9100 averages 64,212, which lands in the 89th percentile. The NVIDIA card's nearest rivals include the AMD Radeon PRO W7600 at 87,108 (1.7% higher), the NVIDIA Quadro GP100 at 87,445 (2.1% higher), and the AMD Radeon PRO W6600 at 81,995 (4.4% lower). The WX 9100, by contrast, clusters with far more modest hardware: the NVIDIA CMP 30HX at 63,842 (0.6% higher), the AMD Radeon RX 9060 XT LP at 63,830 (0.6% higher), and the AMD Radeon RX 7600M at 63,775 (0.7% higher). The WX 9100 is essentially tied with a group of midrange cards, while the CMP 40HX competes in a higher tier.

FAQ

Q: Which GPU wins in Geekbench OpenCL?

A: The NVIDIA CMP 40HX wins with a score of 93,395 against the AMD Radeon Pro WX 9100's 66,605, a 40.2% advantage.

Q: How large is the Vulkan performance gap?

A: The CMP 40HX scores 77,879 in Geekbench Vulkan, while the WX 9100 scores 54,711. That is a 42.3% difference in favor of NVIDIA.

Q: Does the AMD card have any benchmark where it wins?

A: No. The head-to-head results show 2 wins for the NVIDIA CMP 40HX and 0 for the AMD Radeon Pro WX 9100 across the recorded tests.

Q: Where does each card rank among all GPUs?

A: The CMP 40HX is in the 93rd percentile with an average score of 85,637. The WX 9100 is in the 89th percentile with an average of 64,212.

Q: What is the memory capacity difference?

A: The WX 9100 has 16 GB of HBM2, while the CMP 40HX has 8 GB of GDDR6. The bus widths differ substantially: 2048-bit for AMD versus 256-bit for NVIDIA.

Q: Which card has a higher boost clock?

A: The NVIDIA CMP 40HX boosts to 1650 MHz, while the AMD WX 9100 boosts to 1500 MHz.

Architecture Differences

The two GPUs come from different foundries and process nodes. NVIDIA uses TSMC's 12 nm process for the TU106 chip, while AMD uses GlobalFoundries' 14 nm process for the Vega 10 die. The transistor counts are close: 10,800 million for NVIDIA and 12,500 million for AMD. Die sizes are 445 mm² for the CMP 40HX and 495 mm² for the WX 9100. Interestingly, the transistor density is similar, 24.3M per mm² for NVIDIA and 25.3M per mm² for AMD, despite the process node difference.

The compute configurations diverge sharply. The WX 9100 has 4,096 shading units, 256 texture mapping units, and 64 render output units. The CMP 40HX has 2,304 shading units, 144 TMUs, and 64 ROPs. The AMD card has no ray tracing cores and no tensor cores. The NVIDIA card includes 36 RT cores and 288 tensor cores, reflecting its Turing architecture. The feature sets tell the story of different design goals: NVIDIA built for a specific compute workload with future flexibility, while AMD built a broad professional compute engine.

Memory architecture is another major split. The WX 9100 uses 16 GB of HBM2 on a 2048-bit bus, delivering 483.8 GB/s of bandwidth. The CMP 40HX uses 8 GB of GDDR6 on a 256-bit bus, delivering 448.0 GB/s. Despite the massive bus width advantage for AMD, the bandwidth difference is only about 8%, proof of the efficiency of GDDR6 at 14 Gbps effective versus HBM2 at 1890 Mbps effective.

Theoretical peak performance favors AMD. The WX 9100 computes 12.29 TFLOPS in FP32 and 24.58 TFLOPS in FP16. The CMP 40HX reaches 7.603 TFLOPS in FP32 and 15.21 TFLOPS in FP16. Yet the benchmark results show NVIDIA winning decisively, meaning the theoretical numbers do not translate to real-world compute efficiency in these specific loads.

Other differences matter for deployment. The CMP 40HX has no display outputs, while the WX 9100 offers 6x mini-DisplayPort 1.4a. The NVIDIA card uses a PCIe 1.0 x4 interface, oddly limited for a modern GPU, while AMD uses PCIe 3.0 x16. Power draw is 185 W for NVIDIA versus 230 W for AMD, with the CMP 40HX needing a single 8-pin connector and the WX 9100 requiring a 6-pin and an 8-pin. The suggested PSU is 450 W for NVIDIA and 550 W for AMD.

The Verdict

The data points to a clear winner for raw compute performance: the NVIDIA CMP 40HX. It beats the AMD Radeon Pro WX 9100 by over 40% in both recorded benchmarks, despite having fewer shading units and less memory. The NVIDIA card also draws less power and requires a smaller power supply. Its 93rd percentile ranking versus the WX 9100's 89th percentile confirms that the performance gap is real and significant.

However, the WX 9100 is not without merit. It offers twice the memory capacity at 16 GB, which could matter for workloads that require large datasets in VRAM. It also has display outputs, making it usable for tasks that require visual output. The CMP 40HX, with no outputs and a PCIe 1.0 x4 interface, is clearly specialized for a narrow role.

For anyone choosing between these two based on the recorded benchmarks, the NVIDIA card is the faster compute device. Its 40.2% OpenCL lead and 42.3% Vulkan lead are decisive. The AMD card's advantages are memory capacity and connectivity, not speed.

Specification Differences

| Specification | NVIDIA CMP 40HX | AMD Radeon Pro WX 9100 |

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

| Chip | TU106 | Vega 10 |

| Architecture | Turing | GCN 5.0 |

| Process Node | 12 nm | 14 nm |

| Foundry | TSMC | GlobalFoundries |

| Transistors | 10,800 million | 12,500 million |

| Die Size | 445 mm² | 495 mm² |

| Base Clock | 1470 MHz | 1200 MHz |

| Boost Clock | 1650 MHz | 1500 MHz |

| Memory Size | 8 GB | 16 GB |

| Memory Type | GDDR6 | HBM2 |

| Memory Bus Width | 256 bit | 2048 bit |

| Memory Bandwidth | 448.0 GB/s | 483.8 GB/s |

| Shading Units | 2304 | 4096 |

| TMUs | 144 | 256 |

| RT Cores | 36 | null |

| Tensor Cores | 288 | null |

| FP32 | 7.603 TFLOPS | 12.29 TFLOPS |

| FP16 | 15.21 TFLOPS | 24.58 TFLOPS |

| TDP | 185 W | 230 W |

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

| Suggested PSU | 450 W | 550 W |

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

| Display Outputs | No outputs | 6x mini-DisplayPort 1.4a |

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

| Vulkan | 1.4 | 1.3 |

| Release Date | 2021-02-24 | 2017-07-09 |

| Launch MSRP | 699 USD | 1,599 USD |

Where Each One Wins

The NVIDIA CMP 40HX wins in every recorded benchmark category. Its Geekbench OpenCL score of 93,395 is 40.2% ahead of the WX 9100, and its Vulkan score of 77,879 is 42.3% ahead. It also wins on efficiency metrics: lower TDP at 185 W versus 230 W, lower suggested PSU at 450 W versus 550 W, and a smaller die at 445 mm² versus 495 mm². The NVIDIA card's Turing architecture brings ray tracing cores and tensor cores, features the AMD GPU lacks entirely.

The AMD Radeon Pro WX 9100 wins where capacity and connectivity matter. Its 16 GB of HBM2 memory doubles the CMP 40HX's 8 GB, and its 483.8 GB/s bandwidth edges out NVIDIA's 448.0 GB/s. The 6x mini-DisplayPort outputs make it a functional workstation card, while the CMP 40HX cannot display anything. The WX 9100 also has higher theoretical compute peaks at 12.29 TFLOPS FP32 and 24.58 TFLOPS FP16, though these do not translate to benchmark victories.

The verdict is straightforward: the CMP 40HX is the faster compute device by a wide margin, while the WX 9100 is the more versatile professional card with greater memory and display capabilities. Choose based on whether raw compute speed or hardware flexibility matters more.