NVIDIA CMP 90HX vs NVIDIA RTX A3000 Mobile Comparison

The NVIDIA RTX A3000 Mobile and NVIDIA CMP 90HX are both Ampere-based GPUs, but they are engineered for radically different purposes. The A3000 Mobile is a professional laptop solution, while the CMP 90HX is a desktop mining card. Benchmark data reveals a single head-to-head comparison, with the mobile part taking a decisive victory.

Head-to-Head Benchmarks

The only available direct comparison between these two cards is in the Geekbench OpenCL compute test. In this benchmark, the NVIDIA RTX A3000 Mobile scores 79,091 points, while the NVIDIA CMP 90HX scores 69,000 points. This results in a 14.6% advantage for the A3000 Mobile, marking it as the definitive winner in the single head-to-head test recorded. The A3000 Mobile claims 1 win, while the CMP 90HX has 0 wins.

This outcome is surprising given the raw specifications of each card. The CMP 90HX features a much larger GA102 chip with 28,300 million transistors and 6,400 shading units, compared to the A3000 Mobile’s GA104 chip with 17,400 million transistors and 4,096 shading units. The CMP 90HX also has a substantially higher FP32 throughput of 21.89 TFLOPS versus the A3000 Mobile’s 10.08 TFLOPS. Despite this theoretical compute advantage of over 2x, the CMP 90HX falls behind in the actual OpenCL workload measured here.

When placed in the broader context of their respective peer groups, both cards perform similarly. The A3000 Mobile’s average benchmark score is 70,140, placing it in the 91st percentile of all GPUs. Its nearest rival is the NVIDIA Quadro P6000, which scores 69,986, a mere 0.2% difference. The CMP 90HX, with an average score of 69,000, sits in the 90th percentile. Its closest competitor is the Intel Arc A770 at 68,809, just 0.3% behind. Interestingly, the two cards are direct rivals to each other, with the A3000 Mobile being 1.7% ahead of the CMP 90HX in the average score rankings. The data suggests that while the CMP 90HX has a higher theoretical peak, its real-world compute performance in this specific test is hampered, likely by its PCIe 1.0 x4 interface and lack of display outputs, which may limit its applicability in general-purpose compute tasks.

FAQ

Q: Which GPU is faster in the Geekbench OpenCL benchmark?

A: The NVIDIA RTX A3000 Mobile is faster, scoring 79,091 points compared to the NVIDIA CMP 90HX’s 69,000 points. This represents a 14.6% performance lead for the A3000 Mobile.

Q: How does the CMP 90HX compare to its nearest rival, the Intel Arc A770?

A: The CMP 90HX has an average benchmark score of 69,000, which is 0.3% higher than the Intel Arc A770’s average score of 68,809. This makes the CMP 90HX marginally faster on average.

Q: What is the performance difference between the A3000 Mobile and the Quadro P6000?

A: The A3000 Mobile has an average score of 70,140, which is 0.2% higher than the NVIDIA Quadro P6000’s average score of 69,986. The performance of these two cards is nearly identical.

Q: Does the CMP 90HX support display outputs?

A: No. The CMP 90HX is designed for mining and has no display outputs. In contrast, the RTX A3000 Mobile's display outputs are described as "Portable Device Dependent," meaning they rely on the laptop's integrated display connections.

Q: Which card has a higher memory bandwidth?

A: The NVIDIA CMP 90HX has significantly higher memory bandwidth at 760.3 GB/s, utilizing 10 GB of GDDR6X memory on a 320-bit bus. The RTX A3000 Mobile has 264.0 GB/s bandwidth with 6 GB of GDDR6 memory on a 192-bit bus.

Q: What is the transistor density difference between the two chips?

A: The CMP 90HX’s GA102 chip has a slightly higher transistor density of 45.1M / mm², compared to the A3000 Mobile’s GA104 chip at 44.4M / mm². Despite this, the GA102 is a much larger die at 628 mm² versus 392 mm².

Where Each One Wins

The NVIDIA RTX A3000 Mobile is the clear winner in compute performance as measured by the Geekbench OpenCL test. Its 14.6% lead over the CMP 90HX, combined with its portable form factor and display output capabilities, makes it the superior choice for any mobile workstation task requiring GPU acceleration. Its lower TDP of 70 W also suggests it is far more power-efficient, though the specific thermal implications are not detailed in the benchmark data. The A3000 Mobile is positioned for professional mobile workloads, with its "Ampere-MW" generation designation and predecessor in the Quadro Turing-M line indicating a direct lineage in the mobile workstation market.

The NVIDIA CMP 90HX, on the other hand, has no wins in the head-to-head comparison. Its strengths lie in its raw specification sheet, which is not reflected in the benchmark results. It has a larger memory pool of 10 GB and a much wider 320-bit memory bus, offering 760.3 GB/s of bandwidth. This is designed for memory-intensive operations, but the benchmark data shows it does not translate to a win in OpenCL compute. Its designation as a "Mining GPU" and lack of display outputs indicate its intended use case is cryptocurrency mining, where its high FP32 rate and memory bandwidth could be leveraged, but this specific data point does not demonstrate that advantage. The CMP 90HX also has a higher pixel rate (136.8 GPixel/s) and texture rate (342.0 GTexel/s) compared to the A3000 Mobile (78.72 GPixel/s and 157.4 GTexel/s).

Specification Differences

The most striking difference is the intended form factor and power profile. The RTX A3000 Mobile has a TDP of 70 W and requires no power connectors, reflecting its mobile design. The CMP 90HX is a dual-slot desktop card with a 320 W TDP, requiring 2x 8-pin power connectors and a suggested power supply of 700 W. The card’s dimensions are 285 mm in length and 112 mm in height.

Memory configurations differ significantly. The A3000 Mobile uses 6 GB of GDDR6 on a 192-bit bus, delivering 264.0 GB/s of bandwidth. The CMP 90HX features 10 GB of GDDR6X on a 320-bit bus, providing 760.3 GB/s. Clock speeds are also vastly different, with the CMP 90HX running at a base of 1500 MHz and boost of 1710 MHz, while the A3000 Mobile runs at a low 600 MHz base and 1230 MHz boost. The CMP 90HX's memory runs at 19 Gbps effective, whereas the A3000 Mobile's memory is at 11 Gbps effective.

The bus interface is another point of divergence. The A3000 Mobile uses a modern PCIe 4.0 x16 connection, while the CMP 90HX is limited to a legacy PCIe 1.0 x4 interface, which is a significant bottleneck for data transfer. The CMP 90HX has no display outputs, while the A3000 Mobile's outputs are dependent on the host device. Both cards have the same DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4 API support.

Architecture Differences

While both GPUs are based on the Ampere architecture and manufactured on Samsung's 8 nm process, their underlying chips are fundamentally different. The A3000 Mobile uses the GA104 chip, which is a smaller, more power-efficient design. The CMP 90HX uses the GA102 chip, which is NVIDIA's flagship Ampere die for desktop. This is reflected in the transistor counts: the GA102 has 28,300 million transistors on a 628 mm² die, while the GA104 has 17,400 million on a 392 mm² die. The transistor density is nearly identical, at 45.1M / mm² for the GA102 and 44.4M / mm² for the GA104.

The execution resources scale accordingly. The CMP 90HX has 6,400 shading units, 200 TMUs, 80 ROPs, 50 RT cores, and 200 Tensor Cores. The A3000 Mobile is configured with 4,096 shading units, 128 TMUs, 64 ROPs, 32 RT cores, and 128 Tensor Cores. This gives the CMP 90HX a theoretical FP32 performance of 21.89 TFLOPS, more than double the A3000 Mobile's 10.08 TFLOPS. Both offer 1:1 FP16 performance, meaning their FP16 throughput is identical to their FP32 rates.

The generation naming also highlights their different paths. The A3000 Mobile is part of the "Ampere-MW (Ax000)" generation, succeeding the Quadro Turing-M line. The CMP 90HX is listed under "Mining GPUs," a separate category with no predecessor or successor. The release dates differ as well, with the A3000 Mobile launching on 2021-04-11 and the CMP 90HX later on 2021-07-27. Both are now end-of-life products.