NVIDIA CMP 40HX vs NVIDIA GeForce RTX 4090 Comparison

Head-to-Head Benchmarks

The benchmark data available in the database shows a complete victory for the NVIDIA GeForce RTX 4090 in the two recorded head-to-head tests. The margin is substantial in both cases, with the RTX 4090 outperforming the CMP 40HX by a wide margin in compute and graphics API workloads.

In the Geekbench OpenCL test, the RTX 4090 scores 255416, while the CMP 40HX scores 93395. This represents a delta of -63.4% for the CMP 40HX, meaning the RTX 4090 is roughly 2.7 times faster in this compute-oriented benchmark. The OpenCL test exercises general-purpose GPU compute, and the results indicate that the Ada Lovelace architecture has a massive advantage in raw parallel processing throughput.

The Geekbench Vulkan test shows an even larger gap. The RTX 4090 scores 271631, compared to 77879 for the CMP 40HX, a delta of -71.3%. Vulkan is a low-level graphics and compute API that scales well with hardware resources, and the RTX 4090's advantage here suggests that its superior shading unit count, texture units, and memory bandwidth translate directly into higher performance in modern API workloads.

The CMP 40HX records zero wins in the head-to-head comparison, while the RTX 4090 wins both tests. The average benchmark score in the database reinforces this divide: the CMP 40HX has an average score of 85637, while the RTX 4090 averages 60347. This apparent contradiction, where the CMP 40HX has a higher average score than the RTX 4090, is explained by the fact that the two cards are benchmarked across different test suites. The RTX 4090's average is dragged down by several Passmark sub-tests with very low scores, including Passmark DirectX 9 at 397, DirectX 12 at 150, and DirectX 10 at 224. These are legacy API tests that do not reflect the card's true capability in modern workloads. The CMP 40HX, by contrast, only has Geekbench results recorded in the database.

When compared to their respective nearest rivals, the two cards occupy very different positions in the performance hierarchy. The CMP 40HX sits at the 93rd percentile of all GPUs, with its nearest rival being the AMD Radeon PRO W7600, which scores 87108, a delta of -1.7% relative to the CMP 40HX. This means the CMP 40HX is competitive with modern workstation-class GPUs, despite being an older mining-focused product. The RTX 4090, however, sits at the 88th percentile, with its nearest rival being the Intel Arc Pro A60, which scores 60326, a delta of 0%. This percentile ranking reflects the database's averaging methodology, which includes a broader range of tests for the RTX 4090, including several legacy DirectX benchmarks where it scores very low.

The raw specification data supports the benchmark results. The RTX 4090 has 16,384 shading units, 512 texture mapping units, and 176 ROPs, compared to 2,304 shading units, 144 TMUs, and 64 ROPs for the CMP 40HX. The RTX 4090 also has 128 RT cores and 512 tensor cores, versus 36 RT cores and 288 tensor cores for the CMP 40HX. Memory bandwidth is another decisive factor: the RTX 4090 delivers 1.01 TB/s over a 384-bit bus, while the CMP 40HX provides 448.0 GB/s over a 256-bit bus.

Architecture Differences

The two cards are separated by two full generations of NVIDIA GPU architecture. The CMP 40HX is built on the Turing architecture, specifically the TU106 chip, while the RTX 4090 uses the Ada Lovelace architecture with the AD102 chip. This architectural gap explains most of the performance difference observed in the benchmarks.

The manufacturing process differs substantially. The CMP 40HX uses a 12 nm TSMC process, while the RTX 4090 uses a 5 nm TSMC process. Transistor density scales accordingly: the CMP 40HX has 10,800 million transistors on a 445 mm² die, giving a density of 24.3 million transistors per square millimeter. The RTX 4090 packs 76,300 million transistors onto a 609 mm² die, achieving a density of 125.3 million per square millimeter. This is a five-fold increase in transistor density, which allows the RTX 4090 to fit far more compute resources into a comparable physical footprint.

Memory technology also differs. The CMP 40HX uses 8 GB of GDDR6 memory, while the RTX 4090 uses 24 GB of GDDR6X. The memory clock for the CMP 40HX is listed as 1750 MHz with 14 Gbps effective data rate, while the RTX 4090's memory runs at 1313 MHz with 21 Gbps effective. The GDDR6X standard used by the RTX 4090 delivers higher bandwidth per pin, and combined with the wider 384-bit bus, the total bandwidth advantage is decisive.

The RTX 4090 also has a significant advantage in the feature set. It supports PCIe 4.0 x16, while the CMP 40HX uses PCIe 1.0 x4, a legacy interface that severely limits data transfer between the GPU and the host system. The RTX 4090 has display outputs (1x HDMI 2.1, 3x DisplayPort 1.4a), while the CMP 40HX has no display outputs at all, reflecting its mining-oriented design. The RTX 4090 includes ray tracing and tensor core hardware in much greater abundance, with 128 RT cores and 512 tensor cores compared to 36 and 288 respectively. The CMP 40HX does support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, matching the RTX 4090's API support, but the underlying hardware is far less capable.

Clock speeds also differ significantly. The CMP 40HX has a base clock of 1470 MHz and a boost clock of 1650 MHz. The RTX 4090 runs at 2235 MHz base and 2520 MHz boost. The combination of higher clocks, more cores, and more memory bandwidth gives the RTX 4090 a massive theoretical throughput advantage: 82.58 TFLOPS FP32 versus 7.603 TFLOPS for the CMP 40HX.

FAQ

Q: Why does the RTX 4090 have a lower average benchmark score than the CMP 40HX if it wins every head-to-head test?

A: The average benchmark score is calculated across different test suites for each card. The CMP 40HX has only Geekbench OpenCL and Vulkan results recorded, while the RTX 4090 has a broader set of tests including several Passmark sub-tests (DirectX 9, 10, 11, 12, G2D, G3D, GPU Compute). Some of these legacy DirectX tests return very low scores for the RTX 4090, such as 150 for DirectX 12 and 224 for DirectX 10, which pulls its average down to 60347. The CMP 40HX averages 85637 because its two recorded scores are both relatively strong.

Q: Which card has better ray tracing performance?

A: The RTX 4090 is equipped with 128 RT cores, while the CMP 40HX has 36 RT cores. The RTX 4090 also uses the newer Ada Lovelace architecture, which includes more advanced ray tracing hardware. The database does not include a dedicated ray tracing benchmark, but the core count difference and the Vulkan benchmark result, where the RTX 4090 leads by 71.3%, indicate a substantial advantage for the RTX 4090 in any workload that uses RT cores.

Q: Can the CMP 40HX be used for gaming or display output?

A: No. The CMP 40HX has no display outputs, meaning it cannot drive a monitor. It is designed for compute and mining workloads. The RTX 4090, by contrast, includes 1x HDMI 2.1 and 3x DisplayPort 1.4a outputs, making it a full-featured graphics card.

Q: What is the memory bandwidth difference between the two cards?

A: The RTX 4090 offers 1.01 TB/s of memory bandwidth, while the CMP 40HX provides 448.0 GB/s. The RTX 4090 uses a 384-bit bus with GDDR6X memory running at 21 Gbps effective, while the CMP 40HX uses a 256-bit bus with GDDR6 memory at 14 Gbps effective. This difference is critical for memory-intensive workloads.

Q: Which card has a higher transistor density?

A: The RTX 4090 has a transistor density of 125.3 million per square millimeter, while the CMP 40HX has 24.3 million per square millimeter. This is a direct result of the process node difference: the RTX 4090 uses a 5 nm TSMC process, while the CMP 40HX uses 12 nm TSMC.

Q: What are the power requirements for each card?

A: The CMP 40HX has a TDP of 185 W and requires a 450 W power supply, using a single 8-pin connector. The RTX 4090 has a TDP of 450 W and requires an 850 W power supply, using a single 16-pin connector. The RTX 4090 is also a triple-slot card, while the CMP 40HX is dual-slot.

Specification Differences

| Specification | NVIDIA CMP 40HX | NVIDIA GeForce RTX 4090 |

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

| Architecture | Turing | Ada Lovelace |

| Process Node | 12 nm | 5 nm |

| Transistors | 10,800 million | 76,300 million |

| Die Size | 445 mm² | 609 mm² |

| Transistor Density | 24.3M / mm² | 125.3M / mm² |

| Base Clock | 1470 MHz | 2235 MHz |

| Boost Clock | 1650 MHz | 2520 MHz |

| Memory Size | 8 GB | 24 GB |

| Memory Type | GDDR6 | GDDR6X |

| Memory Bus | 256 bit | 384 bit |

| Memory Bandwidth | 448.0 GB/s | 1.01 TB/s |

| Effective Memory Clock | 14 Gbps | 21 Gbps |

| Shading Units | 2304 | 16384 |

| TMUs | 144 | 512 |

| ROPs | 64 | 176 |

| RT Cores | 36 | 128 |

| Tensor Cores | 288 | 512 |

| Pixel Rate | 105.6 GPixel/s | 443.5 GPixel/s |

| Texture Rate | 237.6 GTexel/s | 1,290.2 GTexel/s |

| FP32 Performance | 7.603 TFLOPS | 82.58 TFLOPS |

| FP16 Performance | 15.21 TFLOPS (2:1) | 82.58 TFLOPS (1:1) |

| TDP | 185 W | 450 W |

| Slot Width | Dual-slot | Triple-slot |

| Power Connectors | 1x 8-pin | 1x 16-pin |

| Suggested PSU | 450 W | 850 W |

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

| Display Outputs | No outputs | 1x HDMI 2.1, 3x DisplayPort 1.4a |

| Length | 229 mm | 304 mm |

| Height | 111 mm | 137 mm |

| Width | 35 mm | 61 mm |

| Release Date | 2021-02-24 | 2022-09-19 |

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

Where Each One Wins

The RTX 4090 wins every recorded benchmark comparison. In Geekbench OpenCL, it leads by 63.4%, and in Geekbench Vulkan, it leads by 71.3%. The specification data shows that the RTX 4090 has roughly seven times the shading units, more than three times the texture units, and more than double the memory bandwidth. It also has a 5 nm process advantage, higher clocks, and far more RT and tensor cores.

The CMP 40HX does hold one comparative advantage: its average benchmark score of 85637 is higher than the RTX 4090's 60347. This is due to the different test suites recorded for each card, as discussed above. The CMP 40HX also has a much lower TDP of 185 W versus 450 W, making it more power-efficient per watt in pure compute tasks. Its dual-slot form factor and 229 mm length make it physically smaller, which could be relevant in dense compute environments. It also uses a single 8-pin power connector, which simplifies installation in systems with limited power delivery.

For mining or compute workloads where display output is not needed, the CMP 40HX could be a viable option due to its lower power draw and smaller physical footprint. However, the raw performance difference is so large that any workload that can utilize the RTX 4090's resources will complete significantly faster on the newer card.

The Verdict

The data is unambiguous: the NVIDIA GeForce RTX 4090 is the superior GPU by a massive margin. It wins both head-to-head benchmark tests, with deltas of -63.4% in OpenCL and -71.3% in Vulkan relative to the CMP 40HX. The architectural advantages of Ada Lovelace over Turing, combined with the 5 nm process, 24 GB of GDDR6X memory, and 16,384 shading units, make this a one-sided comparison.

The CMP 40HX is an end-of-life mining product with no display outputs, a legacy PCIe 1.0 x4 interface, and a much smaller compute footprint. It does have a higher average benchmark score in the database, but this is an artifact of the test suite differences, not a reflection of real-world capability. Its 93rd percentile ranking places it above the RTX 4090's 88th percentile in the database's overall standings, but this ranking must be interpreted with caution given the different benchmark distributions.

For any user who needs maximum compute performance, ray tracing capability, modern API support, or display output, the RTX 4090 is the clear choice. Its 82.58 TFLOPS FP32 performance is more than ten times the CMP 40HX's 7.603 TFLOPS. For users with strict power or physical space constraints who do not need display output, the CMP 40HX offers a lower-power alternative, but the performance sacrifice is enormous. The verdict is straightforward: the RTX 4090 wins on every meaningful performance metric recorded in the database.