NVIDIA CMP 90HX vs NVIDIA RTX A2000 Comparison

NVIDIA offers two very different Ampere-based products in the CMP 90HX and the RTX A2000. The database shows one clear benchmark comparison between them, but the specification sheets reveal two devices built for entirely separate purposes. The CMP 90HX is a dedicated mining card with no display outputs, while the RTX A2000 is a compact workstation card with four mini-DisplayPort connectors. The recorded data shows a 1.9% performance gap in a single OpenCL test, but the architectural differences tell a much larger story.

Head-to-Head Benchmarks

The only direct benchmark comparison available in the database is the Geekbench OpenCL test. The NVIDIA CMP 90HX scores 69000 points, while the NVIDIA RTX A2000 scores 67695 points. That gives the CMP 90HX a 1.9% lead. This is a narrow margin, and it places both cards in the same performance neighborhood for raw compute workloads. The CMP 90HX wins the only head-to-head benchmark recorded, but the delta is small enough that real-world application differences could easily flip the result depending on the workload.

Looking at their positions in the broader database, the CMP 90HX sits at the 90th percentile of all GPUs, while the RTX A2000 sits at the 85th percentile. The CMP 90HX has an average benchmark score of 69000, which puts it 0.3% ahead of the Intel Arc A770 and 0.6% ahead of the AMD Radeon Instinct MI25. It trails the AMD Radeon Pro WX 8200 by 1.2% and the NVIDIA Quadro P6000 by 1.4%. These are tight groupings, suggesting the CMP 90HX performs in line with high-end workstation and enthusiast cards from previous generations.

The RTX A2000 presents a more complex picture. Its average benchmark score across all recorded tests is 46043, but that number is dragged down by its 3DMark Steel Nomad DX12 result of 1345. The Geekbench OpenCL score of 67695 and Geekbench Vulkan score of 69089 are much higher. The average places it just 0.2% behind the NVIDIA RTX 5880 Ada Generation and 1% ahead of the Intel Arc A730M. It sits 1.2% behind both the AMD Radeon RX 5600M and the Intel Arc A530M. The RTX A2000's average is heavily influenced by the 3DMark result, which is a DX12 gaming-oriented test that may not reflect its workstation strengths.

The interesting implication here is that the CMP 90HX and RTX A2000 are nearly identical in OpenCL compute performance, despite the CMP 90HX having nearly double the shading units, texture units, and memory bandwidth. The 1.9% gap in the head-to-head test suggests that the RTX A2000's lower clocks and smaller GPU are not as limiting in this particular workload as the raw specifications might suggest.

Architecture Differences

Both cards use the Ampere architecture and are built on Samsung's 8 nm process, but they are fundamentally different chips. The CMP 90HX uses the GA102 die, the large flagship-class silicon with 28,300 million transistors on a 628 mm² die. The RTX A2000 uses the GA106 die, a much smaller chip with 12,000 million transistors on a 276 mm² die. The transistor density is similar, 45.1M per mm² for the CMP 90HX versus 43.5M per mm² for the RTX A2000, which reflects the shared manufacturing process.

The compute resources differ dramatically. The CMP 90HX has 6400 shading units, 200 texture mapping units, 80 ROPs, 50 ray tracing cores, and 200 tensor cores. The RTX A2000 has 3328 shading units, 104 TMUs, 48 ROPs, 26 ray tracing cores, and 104 tensor cores. The CMP 90HX has roughly double the resources across every category. This shows up in the theoretical throughput: the CMP 90HX delivers 21.89 TFLOPS of FP32 and FP16 compute, while the RTX A2000 delivers 7.987 TFLOPS in both. The CMP 90HX also has higher pixel and texture rates at 136.8 GPixel/s and 342.0 GTexel/s, compared to 57.60 GPixel/s and 124.8 GTexel/s for the RTX A2000.

Clock speeds tell a different story. The CMP 90HX runs at a 1500 MHz base and 1710 MHz boost, while the RTX A2000 is much more conservative at 562 MHz base and 1200 MHz boost. The RTX A2000's lower clocks are likely a consequence of its 70 W power target, which is far below the CMP 90HX's 320 W. The RTX A2000 draws its power from the PCIe slot with no additional power connectors, while the CMP 90HX requires two 8-pin connectors and a 700 W power supply. The RTX A2000 only suggests a 250 W power supply.

Memory configurations are also starkly different. The CMP 90HX has 10 GB of GDDR6X on a 320-bit bus, delivering 760.3 GB/s of bandwidth. The RTX A2000 has 6 GB of GDDR6 on a 192-bit bus, delivering 288.0 GB/s. The CMP 90HX's memory runs at 1188 MHz with 19 Gbps effective speed, while the RTX A2000 runs at 1500 MHz with 12 Gbps effective. The CMP 90HX has more capacity, a wider bus, and faster memory, which should benefit bandwidth-hungry workloads.

The bus interface is another major difference. The CMP 90HX uses PCIe 1.0 x4, which is a severely limited connection, while the RTX A2000 uses PCIe 4.0 x16, the fastest standard available. This matters for any workload that requires transferring data between the GPU and the rest of the system. The CMP 90HX's mining-focused design clearly does not prioritize host communication, while the RTX A2000's full-bandwidth interface supports professional workflows that move data frequently.

Physical size and output capabilities reflect their intended roles. The CMP 90HX is 285 mm long and 112 mm tall, with no display outputs at all. The RTX A2000 is 167 mm long and 69 mm tall, a compact dual-slot card with 4x mini-DisplayPort 1.4a outputs. Both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, so API-level features are identical.

The Verdict

The data points to two different buyers. The CMP 90HX offers roughly double the raw compute resources of the RTX A2000, with a 21.89 TFLOPS FP32 throughput compared to 7.987 TFLOPS. It has more memory, much higher bandwidth, and a 90th percentile position in the database. However, it has no display outputs, uses a crippled PCIe 1.0 x4 interface, and consumes 320 W. This card was built for dedicated mining operations where display output and host bandwidth are irrelevant.

The RTX A2000 is the more versatile card despite its lower raw specifications. It has display outputs, a full PCIe 4.0 x16 interface, and a 70 W power draw that requires no external power connectors. Its Geekbench Vulkan score of 69089 actually exceeds its OpenCL score, and it sits at the 85th percentile of all GPUs. The RTX A2000's 3DMark Steel Nomad DX12 score of 1345 is the only recorded gaming-oriented benchmark, and it is low, but the card's workstation positioning suggests that test may not represent its intended workloads.

The launch MSRP of the RTX A2000 was 449 USD. The CMP 90HX has no recorded launch price. The RTX A2000 also has a successor in the Workstation Ada generation, while the CMP 90HX has no successor listed.

For compute-heavy workloads that do not require display output, the CMP 90HX is the stronger choice based on its 21.89 TFLOPS and 760.3 GB/s bandwidth. For any professional use case that requires a monitor, a standard PCIe connection, or low power consumption, the RTX A2000 is the only viable option between these two.

FAQ

Q: Which card is faster in the head-to-head OpenCL benchmark?

A: The NVIDIA CMP 90HX scores 69000 points versus 67695 for the RTX A2000, a 1.9% advantage for the CMP 90HX.

Q: How much memory does each card have?

A: The CMP 90HX has 10 GB of GDDR6X on a 320-bit bus with 760.3 GB/s bandwidth. The RTX A2000 has 6 GB of GDDR6 on a 192-bit bus with 288.0 GB/s bandwidth.

Q: Does the RTX A2000 support display output?

A: Yes, the RTX A2000 has 4x mini-DisplayPort 1.4a outputs. The CMP 90HX has no display outputs.

Q: What are the power requirements for each card?

A: The CMP 90HX has a 320 W TDP, requires 2x 8-pin power connectors, and suggests a 700 W power supply. The RTX A2000 has a 70 W TDP, requires no power connectors, and suggests a 250 W power supply.

Q: Which card has the higher FP32 compute throughput?

A: The CMP 90HX delivers 21.89 TFLOPS of FP32 compute, while the RTX A2000 delivers 7.987 TFLOPS.

Q: How do their PCIe interfaces compare?

A: The CMP 90HX uses PCIe 1.0 x4, while the RTX A2000 uses PCIe 4.0 x16.

Where Each One Wins

The CMP 90HX wins in raw compute throughput. Its 21.89 TFLOPS FP32 and FP16 performance is nearly triple the RTX A2000's 7.987 TFLOPS. Its pixel rate of 136.8 GPixel/s and texture rate of 342.0 GTexel/s dwarf the RTX A2000's 57.60 GPixel/s and 124.8 GTexel/s. The memory subsystem is also a clear win for the CMP 90HX: 10 GB of GDDR6X at 760.3 GB/s versus 6 GB of GDDR6 at 288.0 GB/s. For workloads that saturate memory bandwidth or require massive parallel compute, the CMP 90HX is the data-backed choice.

The RTX A2000 wins in practical usability. It has display outputs, a modern PCIe 4.0 x16 interface, and a 70 W power draw that the CMP 90HX cannot match. The CMP 90HX's PCIe 1.0 x4 interface is a severe bottleneck for any workload that requires frequent host-GPU communication. The RTX A2000 also has a successor in the Workstation Ada line, while the CMP 90HX is a dead end with no successor. The RTX A2000's Geekbench Vulkan score of 69089 is slightly higher than its OpenCL score, suggesting it handles multiple API workloads well.

The 3DMark Steel Nomad DX12 result of 1345 for the RTX A2000 is the only gaming-oriented data point, and it is low. The CMP 90HX has no gaming benchmark data at all. Neither card is positioned as a gaming product, but the RTX A2000 at least has recorded DX12 performance.

Specification Differences

The two cards differ in nearly every major specification. The CMP 90HX uses the GA102 chip with 28,300 million transistors on a 628 mm² die, while the RTX A2000 uses the GA106 chip with 12,000 million transistors on a 276 mm² die. Transistor density is similar at 45.1M per mm² versus 43.5M per mm².

Clock speeds differ significantly: the CMP 90HX runs at 1500 MHz base and 1710 MHz boost, while the RTX A2000 runs at 562 MHz base and 1200 MHz boost. Memory clocks are 1188 MHz with 19 Gbps effective for the CMP 90HX versus 1500 MHz with 12 Gbps effective for the RTX A2000.

Compute resources: 6400 shading units, 200 TMUs, 80 ROPs, 50 RT cores, and 200 tensor cores for the CMP 90HX. The RTX A2000 has 3328 shading units, 104 TMUs, 48 ROPs, 26 RT cores, and 104 tensor cores. Pixel rate is 136.8 GPixel/s versus 57.60 GPixel/s. Texture rate is 342.0 GTexel/s versus 124.8 GTexel/s. FP32 and FP16 are both 21.89 TFLOPS versus 7.987 TFLOPS.

Power and physical specs: 320 W TDP with 2x 8-pin connectors and a 700 W suggested PSU for the CMP 90HX. The RTX A2000 has a 70 W TDP, no power connectors, and a 250 W suggested PSU. The CMP 90HX measures 285 mm by 112 mm, while the RTX A2000 measures 167 mm by 69 mm. Both are dual-slot cards.

The bus interface is PCIe 1.0 x4 for the CMP 90HX and PCIe 4.0 x16 for the RTX A2000. Display outputs: none for the CMP 90HX, 4x mini-DisplayPort 1.4a for the RTX A2000. Memory: 10 GB GDDR6X on a 320-bit bus versus 6 GB GDDR6 on a 192-bit bus. Bandwidth: 760.3 GB/s versus 288.0 GB/s. The RTX A2000 has a predecessor in Quadro Turing and a successor in Workstation Ada, while the CMP 90HX has neither. The RTX A2000 also has a launch MSRP of 449 USD, while the CMP 90HX has no recorded launch MSRP.