NVIDIA CMP 30HX

The NVIDIA CMP 30HX is a specialized Turing-architecture GPU designed exclusively for mining, and benchmark data shows it sits in a highly competitive performance tier. With an average benchmark score of 64,172, it ranks in the 91st percentile of all GPUs, placing it slightly ahead of several established AMD workstation and mobile parts. The card is based on the TU116 chip manufactured on TSMC's 12 nm process, packing 6,600 million transistors into a 284 mm² die. It lacks any display outputs, making it exclusively a compute-oriented product, yet its raw throughput metrics — including 5.027 TFLOPS of FP32 performance — indicate it remains a formidable processing unit even in an end-of-life status.

How It Compares

The nearest rival to the CMP 30HX is the AMD Radeon VII, which posts an average score of 64,356. The NVIDIA card trails this competitor by a marginal 0.3%, a difference well within typical run-to-run variance. This effectively places the two cards at parity in raw compute benchmarks, though the Radeon VII is a consumer-oriented product with display outputs, while the CMP 30HX is a mining-only part.

Against the AMD Radeon Pro WX 9100, the CMP 30HX holds a slim 0.3% advantage, with the rival averaging 64,002. This professional workstation card is built for reliability and compute workloads, yet the data shows the NVIDIA mining GPU edges it out in average benchmark scores. The performance gap is negligible, meaning the two are interchangeable in raw throughput terms.

The CMP 30HX leads the AMD Radeon RX 7600M by a more noticeable 1.1%, with the mobile GPU scoring an average of 63,505. This margin, while not transformative, does indicate that the desktop-oriented CMP 30HX holds a consistent advantage over this laptop-class part. The gap likely stems from the CMP 30HX's higher sustained power delivery and dedicated cooling solution.

The largest delta among the listed rivals is against the AMD Radeon RX 7800M, where the CMP 30HX is 2.9% faster, with the rival averaging 62,360. This is the most decisive victory in the comparison set, though still a modest edge. The data suggests that across this cluster of AMD GPUs, the CMP 30HX occupies the top position in average performance, reinforcing its standing in the 91st percentile.

Ray Tracing and Feature Set

The CMP 30HX is built on the Turing architecture but does not carry dedicated ray tracing cores or tensor cores. This omission is notable, as it distinguishes the card from other Turing-based products that include such hardware. The absence of these specialized units means the card is not optimized for real-time ray tracing workloads or AI-accelerated tasks that rely on tensor core operations.

API support is robust for a compute-focused card. The GPU supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4, making it compatible with a wide range of modern compute frameworks. While the lack of display outputs prevents any graphical output, the API support ensures that the card can execute headless compute workloads across various software stacks. The Vulkan 1.4 support is particularly noteworthy, as it provides access to advanced compute features.

The card's compute capabilities are defined by its shading units, TMUs, and ROPs. It features 1,408 shading units, 88 texture mapping units, and 48 render output units. These translate to a pixel rate of 85.68 GPixel/s and a texture rate of 157.1 GTexel/s. The FP32 throughput of 5.027 TFLOPS is complemented by an FP16 rate of 10.05 TFLOPS (2:1), indicating that half-precision workloads can run at double the speed.

Benchmark Performance

Benchmark results from Geekbench show the CMP 30HX achieving an OpenCL score of 67,007 and a Vulkan score of 61,336. The OpenCL score is notably higher, reflecting the card's strong compute performance under that API. The average benchmark score of 64,172 sits between these two values, providing a balanced view of the card's capabilities across different workloads.

In comparison to the AMD Radeon VII, the CMP 30HX's average score of 64,172 is 0.3% lower than the rival's 64,356. This is a razor-thin margin that places the two cards in a statistical tie. However, the NVIDIA card's Vulkan score of 61,336 suggests it may perform differently across specific API workloads, potentially favoring OpenCL for peak throughput.

The 1.1% lead over the AMD Radeon RX 7600M (63,505) is consistent across the average benchmark score. This indicates that the CMP 30HX offers a tangible, if modest, performance advantage over this mobile part. The 2.9% edge over the AMD Radeon RX 7800M (62,360) is the clearest performance separation in the rival set, showing that the CMP 30HX can outpace a higher-tier mobile GPU by a meaningful margin.

The percentile ranking of 91 underscores the card's position in the broader GPU landscape. This means it outperforms 91% of all GPUs in the benchmark database, a strong showing for a mining-specific product. The data indicates that despite its niche purpose, the CMP 30HX delivers competitive raw compute performance that rivals mainstream and professional cards.

FAQ

Q: What is the average benchmark score of the NVIDIA CMP 30HX?

A: The average benchmark score is 64,172, with Geekbench OpenCL scoring 67,007 and Vulkan scoring 61,336.

Q: How does the CMP 30HX compare to the AMD Radeon VII?

A: The CMP 30HX trails the AMD Radeon VII by 0.3%, with average scores of 64,172 versus 64,356.

Q: Does the CMP 30HX support ray tracing?

A: No, the card does not include ray tracing cores or tensor cores, as it is built for mining workloads.

Q: What is the memory configuration of the CMP 30HX?

A: It features 6 GB of GDDR6 memory on a 192-bit bus, providing 336.0 GB/s of bandwidth.

Q: What API versions does the CMP 30HX support?

A: It supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4.

Q: Is the CMP 30HX still in production?

A: No, the production status is end-of-life, and it was released on 2021-02-24.

Memory Subsystem

The CMP 30HX is equipped with 6 GB of GDDR6 memory, a capacity that is modest by modern standards but sufficient for its intended mining workloads. The memory operates at 1750 MHz with an effective data rate of 14 Gbps, delivering a total bandwidth of 336.0 GB/s. This bandwidth figure is achieved through a 192-bit memory bus, which balances capacity and throughput for compute tasks.

For high-resolution compute workloads, the 6 GB capacity may prove limiting. Mining algorithms that require large datasets could exceed this allocation, though the card's architecture is tailored to specific hashing functions that fit within this footprint. The bandwidth of 336.0 GB/s is competitive, allowing the card to move data efficiently between the GPU cores and memory. This is critical for mining operations where memory throughput directly impacts hash rates.

The memory subsystem's efficiency is further supported by the card's pixel rate of 85.68 GPixel/s and texture rate of 157.1 GTexel/s. While these metrics are typically associated with graphical output, they also indicate the card's ability to process large blocks of data in parallel. The 192-bit bus width is a mid-range configuration, but the high effective memory speed partially compensates for the narrower bus compared to higher-end cards.

Who Should Consider It

The NVIDIA CMP 30HX is a purpose-built mining card, and its lack of display outputs makes it unsuitable for any graphical use case. However, benchmark data shows it fits a specific niche: users who require high compute throughput without the need for video output. Its 91st percentile ranking means it outperforms most GPUs in raw compute, making it a viable option for headless mining rigs.

At lower resolutions or compute loads, the card's 6 GB memory and 336.0 GB/s bandwidth are adequate. The Geekbench OpenCL score of 67,007 indicates strong performance in compute-heavy applications, suggesting it can handle algorithms that prioritize integer or floating-point operations. For users focused on mining cryptocurrencies that favor NVIDIA architectures, this card provides a competitive hash rate potential.

The card is not suitable for gaming, content creation, or any workload requiring a display connection. Its end-of-life status and mining-specific design mean it should only be considered by those with dedicated mining operations. The FP16 throughput of 10.05 TFLOPS (2:1) may also appeal to users experimenting with mixed-precision compute, though this is not a primary use case for a mining card.

Power and Cooling

The CMP 30HX has a thermal design power (TDP) of 125 W, a modest figure that reflects its mid-range positioning. This power draw is manageable for a dual-slot card, which measures 229 mm in length, 111 mm in height, and 35 mm in width. The physical dimensions indicate a compact design that can fit in most mining rigs without space constraints.

Power is delivered through a single 8-pin connector, a standard configuration for GPUs in this power class. NVIDIA recommends a 300 W power supply, which provides ample headroom for the card's 125 W TDP along with other system components. The dual-slot cooler is designed to dissipate heat effectively, though specific cooling performance is not detailed in the data.

The card's power efficiency is implied by its transistor density of 23.2 million per mm², a figure that balances performance and power consumption. The 12 nm process node from TSMC contributes to this efficiency, allowing the card to deliver 5.027 TFLOPS of FP32 performance within a 125 W envelope. This makes the CMP 30HX a relatively power-efficient option for mining operations, where electricity costs are a primary concern. The PCIe 1.0 x4 bus interface is sufficient for compute workloads, though it is a legacy standard that may limit data transfer speeds in certain scenarios.