AMD Radeon Instinct MI60 vs NVIDIA CMP 30HX Comparison

Head-to-Head Benchmarks

The recorded data shows a decisive performance gap between the AMD Radeon Instinct MI60 and the NVIDIA CMP 30HX across both benchmark tests. In Geekbench OpenCL, the MI60 scores 92,488 against the CMP 30HX’s 65,199, a 41.9% advantage. The Vulkan test tells a similar story: the MI60 posts 92,444 versus 62,484, widening the lead to 47.9%. The MI60 wins both head-to-head matchups, with a 2-0 record in the database.

These results place the MI60 in a different performance tier entirely. Its average benchmark score of 92,466 sits at the 93rd percentile of all GPUs, while the CMP 30HX averages 63,842, landing at the 89th percentile. The gap is not marginal; it reflects a fundamental difference in compute capability. The MI60’s nearest rivals include the AMD Radeon Pro VII at 97,131 (4.8% higher) and the AMD Radeon RX 7900M at 97,487 (5.2% higher), meaning the MI60 trails only slightly behind those larger workstation and mobile flagships. On the other side, the CMP 30HX sits within a tight cluster: the AMD Radeon RX 9060 XT LP matches it exactly at 63,830, the AMD Radeon RX 7600M is 0.1% behind at 63,775, and the AMD Radeon Pro Vega 56 trails by 0.2% at 63,693. The only rival ahead of the CMP 30HX in its recorded field is the AMD Radeon Pro WX 9100 at 64,212, a 0.6% gap.

The 41.9% OpenCL delta is substantial enough that the MI60 effectively doubles the CMP 30HX’s output in compute workloads. The Vulkan delta of 47.9% is even more pronounced, suggesting the MI60’s architecture scales better in graphics-adjacent compute tasks. For context, the MI60’s score of 92,488 is roughly 44% higher than the CMP 30HX’s OpenCL figure, and the Vulkan score difference is even larger in relative terms. Benchmark results indicate that any workload relying on OpenCL or Vulkan compute will favor the MI60 by a wide margin.

Architecture Differences

The two cards come from different architectural generations and target entirely different use cases. The AMD Radeon Instinct MI60 uses the Vega 20 chip built on GCN 5.1 architecture, manufactured on TSMC’s 7 nm process. It packs 13,230 million transistors into a 331 mm² die, yielding a transistor density of 40.0M per mm². The NVIDIA CMP 30HX uses the TU116 chip on Turing architecture, also from TSMC but on a 12 nm process. It contains 6,600 million transistors on a 284 mm² die, with a density of 23.2M per mm². The MI60’s process advantage is clear: nearly double the transistor density, which contributes directly to its compute lead.

Memory configurations diverge sharply. The MI60 carries 32 GB of HBM2 with a 4096-bit bus, delivering 1.02 TB/s of bandwidth. The CMP 30HX has 6 GB of GDDR6 on a 192-bit bus, providing 336.0 GB/s. That is a 3x difference in raw bandwidth, and the MI60’s memory capacity is over five times larger. Clock speeds tell a more nuanced story: the CMP 30HX runs at a base of 1530 MHz and boosts to 1785 MHz, while the MI60 starts at 1200 MHz and boosts to 1800 MHz. The CMP 30HX has a higher base clock, but the MI60’s boost clock edges slightly ahead. Memory clocks differ as well: the MI60 runs at 1000 MHz (2 Gbps effective), while the CMP 30HX runs at 1750 MHz (14 Gbps effective), but the MI60’s enormous bus width overwhelms the frequency disadvantage.

Compute resources are heavily skewed toward the MI60. It has 4096 shading units, 256 texture mapping units, and 64 render output units. The CMP 30HX has 1408 shading units, 88 TMUs, and 48 ROPs. The MI60’s pixel rate is 115.2 GPixel/s versus 85.68 GPixel/s for the CMP 30HX, and its texture rate is 460.8 GTexel/s versus 157.1 GTexel/s. Floating-point performance shows the largest gap: the MI60 delivers 14.75 TFLOPS FP32 and 29.49 TFLOPS FP16 (2:1), while the CMP 30HX manages 5.027 TFLOPS FP32 and 10.05 TFLOPS FP16 (2:1). Neither card features dedicated ray tracing or tensor cores, so both rely on traditional shader compute.

Power and physical characteristics also differ. The MI60 has a 300 W TDP with a suggested 700 W power supply and requires 1x 6-pin plus 1x 8-pin connectors. The CMP 30HX draws 125 W with a 300 W suggested PSU and a single 8-pin connector. The MI60 is longer at 267 mm (10.5 inches) versus 229 mm (9 inches), but both are dual-slot cards with the same 111 mm height. The CMP 30HX has a 35 mm width, while the MI60’s width is not recorded. Both use PCIe slots, but the MI60 uses PCIe 4.0 x16 while the CMP 30HX uses PCIe 1.0 x4, a significant interface difference that limits the CMP 30HX’s data transfer capabilities. The MI60 has one mini-DisplayPort 1.4a output, while the CMP 30HX has no display outputs at all, reflecting its mining-oriented design. API support is similar for DirectX 12 (12_1) and OpenGL 4.6, but the CMP 30HX supports Vulkan 1.4 versus the MI60’s Vulkan 1.3.

Release timing and status differ as well. The MI60 launched in November 2018 as part of the Radeon Instinct MIx generation, succeeding FirePro Data Center, and is now end-of-life. The CMP 30HX launched in February 2021 under the Mining GPUs generation with no predecessor and is also end-of-life. The CMP 30HX has a recorded launch MSRP of 799 USD, while the MI60 has no recorded launch MSRP.

FAQ

Q: Which card wins in Geekbench OpenCL performance?

A: The AMD Radeon Instinct MI60 scores 92,488 versus the NVIDIA CMP 30HX’s 65,199, a 41.9% advantage.

Q: How do the two cards compare in Vulkan benchmarks?

A: The MI60 scores 92,444 while the CMP 30HX scores 62,484, giving the MI60 a 47.9% lead.

Q: What are the memory capacity and bandwidth differences?

A: The MI60 has 32 GB of HBM2 with 1.02 TB/s bandwidth on a 4096-bit bus. The CMP 30HX has 6 GB of GDDR6 with 336.0 GB/s bandwidth on a 192-bit bus.

Q: Which card has higher FP32 compute throughput?

A: The MI60 delivers 14.75 TFLOPS FP32, while the CMP 30HX delivers 5.027 TFLOPS FP32. The MI60 also leads in FP16 at 29.49 TFLOPS versus 10.05 TFLOPS.

Q: Do either of these cards support ray tracing or tensor cores?

A: No. Neither the MI60 nor the CMP 30HX has dedicated ray tracing cores or tensor cores in the recorded specifications.

Q: What is the power consumption difference?

A: The MI60 has a 300 W TDP with a suggested 700 W power supply, while the CMP 30HX has a 125 W TDP with a suggested 300 W power supply.

The Verdict

The data points to a clear choice depending on workload. The AMD Radeon Instinct MI60 is the superior compute card by every measurable metric in the database. Its 41.9% OpenCL and 47.9% Vulkan leads over the CMP 30HX are decisive, and its 93rd percentile ranking versus the CMP 30HX’s 89th confirms the performance gap. The MI60’s 32 GB HBM2 memory with 1.02 TB/s bandwidth makes it suitable for large datasets and memory-intensive workloads, and its 14.75 TFLOPS FP32 throughput places it in a class well above the CMP 30HX’s 5.027 TFLOPS. The MI60 also offers a display output, PCIe 4.0 x16 connectivity, and a 7 nm process with higher transistor density.

The NVIDIA CMP 30HX, by contrast, is a mining-focused card with no display outputs and a PCIe 1.0 x4 interface, which severely limits its utility outside of dedicated compute farms. Its 125 W TDP and 300 W suggested PSU make it far more power-efficient in absolute terms, and its smaller 229 mm length fits in tighter chassis. However, its 6 GB memory capacity and 336.0 GB/s bandwidth are modest, and its compute throughput is less than half of the MI60’s in FP32. The CMP 30HX’s nearest rivals include cards like the AMD Radeon RX 9060 XT LP at 63,830 and the AMD Radeon RX 7600M at 63,775, showing it competes with midrange and mobile parts, not workstation flagships.

For users prioritizing raw compute, memory bandwidth, and general-purpose GPU workloads, the MI60 is the only rational choice. For users constrained by power draw or physical size and focused purely on mining tasks, the CMP 30HX offers a lower-power alternative, but the benchmark data shows it delivers significantly less performance per score. The MI60’s 300 W TDP is more than double the CMP 30HX’s 125 W, but the performance return justifies the power draw for compute-heavy applications. The CMP 30HX’s launch MSRP of 799 USD provides a reference point, but the MI60’s absence of a recorded MSRP makes direct cost comparison impossible from the available data.

Specification Differences

| Specification | AMD Radeon Instinct MI60 | NVIDIA CMP 30HX |

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

| Architecture | GCN 5.1 | Turing |

| Chip | Vega 20 | TU116 |

| Process Node | 7 nm | 12 nm |

| Transistors | 13,230 million | 6,600 million |

| Die Size | 331 mm² | 284 mm² |

| Transistor Density | 40.0M / mm² | 23.2M / mm² |

| Base Clock | 1200 MHz | 1530 MHz |

| Boost Clock | 1800 MHz | 1785 MHz |

| Memory Clock | 1000 MHz (2 Gbps effective) | 1750 MHz (14 Gbps effective) |

| Memory Size | 32 GB | 6 GB |

| Memory Type | HBM2 | GDDR6 |

| Memory Bus Width | 4096 bit | 192 bit |

| Memory Bandwidth | 1.02 TB/s | 336.0 GB/s |

| Shading Units | 4096 | 1408 |

| TMUs | 256 | 88 |

| ROPs | 64 | 48 |

| Pixel Rate | 115.2 GPixel/s | 85.68 GPixel/s |

| Texture Rate | 460.8 GTexel/s | 157.1 GTexel/s |

| FP32 | 14.75 TFLOPS | 5.027 TFLOPS |

| FP16 | 29.49 TFLOPS (2:1) | 10.05 TFLOPS (2:1) |

| TDP | 300 W | 125 W |

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

| Suggested PSU | 700 W | 300 W |

| Bus Interface | PCIe 4.0 x16 | PCIe 1.0 x4 |

| Display Outputs | 1x mini-DisplayPort 1.4a | No outputs |

| Vulkan | 1.3 | 1.4 |

| Length | 267 mm (10.5 inches) | 229 mm (9 inches) |

| Width | Not recorded | 35 mm (1.4 inches) |

| Release Date | 2018-11-17 | 2021-02-24 |

| Generation | Radeon Instinct (MIx) | Mining GPUs |

| Launch MSRP | Not recorded | 799 USD |