AMD Radeon Pro W5700X vs NVIDIA CMP 30HX Comparison

Head-to-Head Benchmarks

The recorded data shows a decisive performance gap between the NVIDIA CMP 30HX and the AMD Radeon Pro W5700X in the two benchmark tests where both cards have comparable results. In Geekbench OpenCL, the NVIDIA CMP 30HX scores 65,199 against the AMD Radeon Pro W5700X's 43,810, a 48.8% advantage. The Vulkan test tells a similar story, with the NVIDIA card posting 62,484 versus 45,246 for the AMD card, a 38.1% lead. These are substantial margins, not marginal differences, and they place the NVIDIA CMP 30HX firmly ahead in raw compute workloads that leverage these APIs.

The win tally stands at 2 for NVIDIA CMP 30HX and 0 for AMD Radeon Pro W5700X in direct head-to-head comparisons. However, the AMD card has an additional benchmark result in its favor: Geekbench Metal, where it scores 75,427. This test is not available for the NVIDIA card, as the CMP 30HX has no display outputs and is not designed for Apple's graphics stack. The Metal score is notably higher than the AMD card's own OpenCL and Vulkan scores, suggesting the Radeon Pro W5700X performs significantly better under Metal, which is relevant for macOS environments.

When looking at average benchmark scores across all recorded tests, the picture shifts. The NVIDIA CMP 30HX has an average of 63,842, while the AMD Radeon Pro W5700X averages 54,828. This represents a 16.4% difference in favor of NVIDIA based on their respective test suites. But the comparison is complicated by the fact that the AMD card includes a Metal result that has no NVIDIA counterpart, while the NVIDIA card lacks a Metal test entirely. The averages reflect each card's own benchmark set, not a perfectly matched suite.

In terms of performance percentiles, the NVIDIA CMP 30HX sits at the 89th percentile among all GPUs, while the AMD Radeon Pro W5700X ranks at the 87th. This narrow 2-point percentile gap is interesting given how large the OpenCL and Vulkan deltas are. It suggests that the AMD card's Metal score helps its overall standing, and that the broader GPU landscape contains many cards that fall between these two in mixed workloads.

Architecture Differences

The two cards come from fundamentally different design philosophies and manufacturing processes. The NVIDIA CMP 30HX uses the TU116 chip built on a 12 nm process at TSMC, with 6,600 million transistors packed into a 284 mm² die. The transistor density is 23.2 million per square millimeter. The AMD Radeon Pro W5700X uses the Navi 10 chip on TSMC's 7 nm process, with 10,300 million transistors on a 251 mm² die, yielding a density of 41.0 million per square millimeter. The AMD chip is denser and packs more transistors into a smaller area, which is a direct consequence of the more advanced 7 nm node.

The NVIDIA card is based on the Turing architecture, while the AMD card uses RDNA 1.0. Neither card includes ray tracing cores or tensor cores, so those features are absent from both. The compute resources differ significantly: the NVIDIA card has 1,408 shading units, 88 texture mapping units, and 48 render output units. The AMD card has 2,560 shading units, 160 TMUs, and 64 ROPs. These are substantially higher counts across the board for AMD, yet the benchmark results show NVIDIA winning in OpenCL and Vulkan, which indicates that architectural efficiency and clock behavior play a larger role than raw unit counts.

Memory configurations diverge sharply. The NVIDIA CMP 30HX has 6 GB of GDDR6 on a 192-bit bus with a bandwidth of 336.0 GB/s. The AMD Radeon Pro W5700X has 16 GB of GDDR6 on a 256-bit bus with bandwidth of 448.0 GB/s. The AMD card offers more capacity and higher bandwidth, which matters for large datasets and high-resolution textures. The memory clock is identical at 1750 MHz, translating to 14 Gbps effective for both. The bus interface also differs: the NVIDIA card uses PCIe 1.0 x4, while the AMD card uses Apple MPX, a proprietary connector for Mac Pro systems.

The power envelope is another major divergence. The NVIDIA CMP 30HX has a TDP of 125 W and requires a single 8-pin power connector with a suggested PSU of 300 W. The AMD Radeon Pro W5700X has a TDP of 205 W, no listed power connectors (relying on the MPX interface), and a suggested PSU of 550 W. The physical footprint differs as well: the NVIDIA card is dual-slot, 229 mm long, 111 mm tall, and 35 mm wide. The AMD card is quad-slot and 305 mm long, with no listed height or width. The AMD card is longer and much thicker, which reflects its higher power delivery needs and larger cooler assembly.

Where Each One Wins

The NVIDIA CMP 30HX wins decisively in OpenCL and Vulkan compute workloads. The 48.8% OpenCL lead and 38.1% Vulkan lead indicate that this card is significantly stronger for general-purpose GPU compute using those APIs. This is consistent with its purpose: the CMP 30HX is a mining GPU, designed for algorithmic workloads that rely heavily on raw compute throughput in these API environments. Its 5.027 TFLOPS FP32 performance, while modest compared to the AMD card's 10.44 TFLOPS, is delivered more efficiently in practice, as the benchmark results show.

The AMD Radeon Pro W5700X wins in Metal workloads, posting a score of 75,427, which is far above its own OpenCL and Vulkan scores. This suggests the RDNA architecture is well optimized for Apple's Metal API, and the card is designed for Mac Pro systems where Metal is the primary graphics interface. The card also wins on memory capacity and bandwidth, offering 16 GB versus 6 GB and 448.0 GB/s versus 336.0 GB/s. For workloads that are memory-bound, such as large-scale rendering or data processing that exceeds the NVIDIA card's 6 GB capacity, the AMD card has a clear advantage.

The AMD card also wins on raw shader and texture throughput. Its 10.44 TFLOPS FP32 is more than double the NVIDIA card's 5.027 TFLOPS. The texture rate is 326.4 GTexel/s versus 157.1 GTexel/s, and the pixel rate is 130.6 GPixel/s versus 85.68 GPixel/s. In purely theoretical terms, the AMD card has a massive peak-compute advantage. Yet the benchmark results show NVIDIA winning the available compute tests, which means the AMD card's theoretical superiority does not translate into higher scores in OpenCL or Vulkan. This could be due to driver maturity, API overhead, or architectural differences in how compute shaders are scheduled.

For gaming and graphics rendering, the available data is limited. Neither card has a DirectX or Vulkan gaming benchmark in the database that directly compares them in rasterized workloads. The NVIDIA card supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The AMD card supports the same API versions. The NVIDIA card has no display outputs, so it cannot be used for rendering to a screen, while the AMD card has 1x HDMI 2.0b and 4x Thunderbolt outputs, making it suitable for display workloads.

Specification Differences

The two cards differ in nearly every major specification category. The process node is 12 nm for NVIDIA versus 7 nm for AMD. Transistor count is 6,600 million versus 10,300 million. Die size is 284 mm² versus 251 mm². Transistor density is 23.2M per mm² versus 41.0M per mm². The base clock is 1530 MHz for NVIDIA versus 1243 MHz for AMD, but the boost clock is 1785 MHz versus 2040 MHz, meaning the AMD card boosts much higher despite a lower base.

Memory size is 6 GB versus 16 GB. Bus width is 192-bit versus 256-bit. Bandwidth is 336.0 GB/s versus 448.0 GB/s. Shading units are 1,408 versus 2,560. TMUs are 88 versus 160. ROPs are 48 versus 64. Pixel rate is 85.68 GPixel/s versus 130.6 GPixel/s. Texture rate is 157.1 GTexel/s versus 326.4 GTexel/s. FP32 is 5.027 TFLOPS versus 10.44 TFLOPS. FP16 is 10.05 TFLOPS versus 20.89 TFLOPS.

TDP is 125 W versus 205 W. Slot width is dual-slot versus quad-slot. Power connectors are 1x 8-pin versus none listed. Suggested PSU is 300 W versus 550 W. Bus interface is PCIe 1.0 x4 versus Apple MPX. Display outputs are none versus 1x HDMI 2.0b and 4x Thunderbolt. Dimensions are 229 mm length versus 305 mm length, with the NVIDIA card also having listed height and width while the AMD card does not.

The release dates differ by over a year: the AMD Radeon Pro W5700X launched on 2019-12-10, while the NVIDIA CMP 30HX launched on 2021-02-24. Both are end-of-life products. The launch MSRP for the NVIDIA CMP 30HX was 799 USD, and for the AMD Radeon Pro W5700X it was 999 USD.

FAQ

Q: Which card has a higher average benchmark score?

A: The NVIDIA CMP 30HX has an average benchmark score of 63,842, while the AMD Radeon Pro W5700X has an average of 54,828. The NVIDIA card is about 16.4% higher on average, though the AMD card's average includes a Metal score that the NVIDIA card cannot produce.

Q: How large is the performance gap in Vulkan workloads?

A: In Geekbench Vulkan, the NVIDIA CMP 30HX scores 62,484 versus 45,246 for the AMD Radeon Pro W5700X, a 38.1% advantage for NVIDIA.

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

A: The AMD Radeon Pro W5700X has a Geekbench Metal score of 75,427, but the NVIDIA CMP 30HX has no Metal benchmark result in the database. In the two tests where both cards have scores (OpenCL and Vulkan), the NVIDIA card wins both.

Q: What memory capacity and bandwidth does each card offer?

A: The NVIDIA CMP 30HX has 6 GB of GDDR6 with 336.0 GB/s bandwidth. The AMD Radeon Pro W5700X has 16 GB of GDDR6 with 448.0 GB/s bandwidth.

Q: Which card has a higher transistor density?

A: The AMD Radeon Pro W5700X has a transistor density of 41.0M per mm², compared to 23.2M per mm² for the NVIDIA CMP 30HX. This reflects the AMD card's use of a 7 nm process versus NVIDIA's 12 nm.

Q: Are both cards still in production?

A: No, both are end-of-life products. The AMD Radeon Pro W5700X was released on 2019-12-10, and the NVIDIA CMP 30HX was released on 2021-02-24.