Intel Arc A730M vs NVIDIA CMP 90HX Comparison

The NVIDIA CMP 90HX and Intel Arc A730M occupy very different corners of the GPU landscape. The CMP 90HX is a desktop mining-focused card built on NVIDIA’s Ampere architecture, while the Arc A730M is Intel’s mobile graphics solution from the Xe-HPG generation. The recorded benchmark data offers a direct comparison point in Geekbench OpenCL, where the Intel part edges out the NVIDIA card by a narrow margin. This analysis examines the architectural split, performance results, and the specification differences that define these two products.

FAQ

Q: Which GPU has the higher average benchmark score in the database?

A: The Intel Arc A730M has a higher average benchmark score of 45,592, while the NVIDIA CMP 90HX averages 69,000. However, the averages are computed from different test suites: the CMP 90HX has a single Geekbench OpenCL score, while the Arc A730M’s average includes three tests (3DMark Steel Nomad DX12, Geekbench OpenCL, and Geekbench Vulkan).

Q: How do the two GPUs compare in the Geekbench OpenCL test?

A: In the head-to-head Geekbench OpenCL benchmark, the Intel Arc A730M scores 70,352, which is 1.9% higher than the NVIDIA CMP 90HX’s score of 69,000. This makes the Arc A730M the winner in the only direct comparison available.

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

A: The Intel Arc A730M, built on TSMC’s 6 nm process, has a transistor density of 53.4 million transistors per square millimeter. The NVIDIA CMP 90HX, fabricated on Samsung’s 8 nm node, has a lower density of 45.1 million transistors per square millimeter.

Q: Which GPU has a higher boost clock speed?

A: The Intel Arc A730M boosts to 2050 MHz, while the NVIDIA CMP 90HX boosts to 1710 MHz. The base clocks are 1100 MHz for the Intel part and 1500 MHz for the NVIDIA part.

Q: What is the memory bandwidth for each GPU?

A: The NVIDIA CMP 90HX has a memory bandwidth of 760.3 GB/s, using 10 GB of GDDR6X on a 320-bit bus. The Intel Arc A730M has 336.0 GB/s bandwidth, using 12 GB of GDDR6 on a 192-bit bus.

Q: How does the NVIDIA CMP 90HX compare to its nearest rivals in percentile ranking?

A: The CMP 90HX sits at the 90th percentile among all GPUs. Its closest rival is the Intel Arc A770 with an average score of 68,809 (0.3% lower), followed by the AMD Radeon Instinct MI25 at 68,562 (0.6% lower). The AMD Radeon Pro WX 8200 scores 69,870 (1.2% higher) and the NVIDIA Quadro P6000 scores 69,986 (1.4% higher).

Architecture Differences

The NVIDIA CMP 90HX is built on the Ampere architecture with the GA102 chip, fabricated by Samsung on an 8 nm process. The die contains 28,300 million transistors across a 628 mm² area, yielding a transistor density of 45.1 million per square millimeter. This is a large, power-hungry desktop GPU, evidenced by its 320 W TDP and dual-slot cooling requirement. The CMP 90HX uses 2x 8-pin power connectors and recommends a 700 W power supply. Notably, it has no display outputs, reflecting its mining-oriented design. The bus interface is listed as PCIe 1.0 x4, which is unusual for a modern GPU and likely reflects a legacy or restricted interface for this specific product.

In contrast, the Intel Arc A730M is a mobile GPU from the Xe-HPG architecture, specifically the Alchemist generation (Arc 7 Mobile). It uses the DG2-512 chip fabricated by TSMC on a 6 nm process. The die is smaller at 406 mm², with 21,700 million transistors, resulting in a higher transistor density of 53.4 million per square millimeter. This is a more modern, denser process, which helps achieve a lower 80 W TDP despite a higher boost clock. The Arc A730M is an integrated-style package (IGP slot width) with no separate power connectors or suggested PSU, as it relies on the host laptop’s power delivery. It uses PCIe 4.0 x16 for connectivity and has display outputs that are portable device dependent, meaning it drives integrated laptop displays.

The compute resources differ significantly. The NVIDIA CMP 90HX features 6400 shading units, 200 texture mapping units, and 80 raster output units. It also includes 50 ray tracing cores and 200 tensor cores, supporting DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The Intel Arc A730M has fewer shading units at 3072, but a higher count of TMUs at 192 and ROPs at 96. It has 24 ray tracing cores and no tensor cores listed. Both support the same API set: DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The FP32 compute is 21.89 TFLOPS for the NVIDIA card, while the Intel part delivers 12.60 TFLOPS. However, the Intel GPU’s FP16 throughput is 25.19 TFLOPS (2:1 ratio), meaning it doubles FP32, whereas the NVIDIA card offers FP16 at 21.89 TFLOPS (1:1 ratio, same as FP32).

The memory subsystems also tell a story of different design goals. The CMP 90HX uses 10 GB of GDDR6X on a 320-bit bus, achieving 760.3 GB/s bandwidth. The Arc A730M has 12 GB of GDDR6 on a 192-bit bus, with 336.0 GB/s bandwidth. The NVIDIA card’s bandwidth is more than double, suited for memory-intensive mining workloads, while the Intel part’s larger capacity but narrower bus suits mobile gaming and general compute. The memory clocks are 1188 MHz (19 Gbps effective) for the NVIDIA card and 1750 MHz (14 Gbps effective) for the Intel part.

The pixel and texture rates further highlight the divergence. The Intel Arc A730M has a higher pixel rate at 196.8 GPixel/s and a higher texture rate at 393.6 GTexel/s. The NVIDIA CMP 90HX achieves 136.8 GPixel/s and 342.0 GTexel/s. This is surprising given the NVIDIA card’s higher FP32 throughput, but the Intel part’s higher ROP count (96 vs 80) and TMU count (192 vs 200, though slightly lower) with a much higher boost clock (2050 MHz vs 1710 MHz) contribute to these rasterization rates.

Head-to-Head Benchmarks

The database records one direct head-to-head benchmark between these two GPUs: Geekbench OpenCL. In this test, the Intel Arc A730M scores 70,352, while the NVIDIA CMP 90HX scores 69,000. The delta is 1.9% in favor of the Intel part, making it the sole winner in the head-to-head comparison. This is a narrow margin, indicating that despite the CMP 90HX’s higher FP32 compute (21.89 TFLOPS vs 12.60 TFLOPS), the Arc A730M manages to edge ahead in this particular OpenCL workload. The result suggests that the Arc A730M’s higher boost clock (2050 MHz vs 1710 MHz) and superior pixel/texture rates (196.8 GPixel/s and 393.6 GTexel/s vs 136.8 GPixel/s and 342.0 GTexel/s) may offset the raw compute advantage of the NVIDIA part in OpenCL.

Looking at the broader benchmark context, the NVIDIA CMP 90HX’s average benchmark score is 69,000, which places it at the 90th percentile among all GPUs. Its nearest rivals are close: the Intel Arc A770 scores 68,809 (0.3% lower), the AMD Radeon Instinct MI25 scores 68,562 (0.6% lower), the AMD Radeon Pro WX 8200 scores 69,870 (1.2% higher), and the NVIDIA Quadro P6000 scores 69,986 (1.4% higher). This tight cluster shows that the CMP 90HX is firmly in the mid-to-high performance tier, with single-digit percentage differences among its peers.

The Intel Arc A730M’s average benchmark score is 45,592, but this figure is skewed by the inclusion of the 3DMark Steel Nomad DX12 score of 1,732, which is a much lower-scoring test. Its Geekbench OpenCL score of 70,352 is the highest among its recorded benchmarks, followed by Geekbench Vulkan at 64,693. The Arc A730M sits at the 84th percentile among all GPUs, which is lower than the CMP 90HX’s 90th percentile. Its nearest rivals include the AMD Radeon Pro 5500 XT (45,384, 0.5% lower), the NVIDIA RTX 5880 Ada Generation (45,972, 0.8% higher), the NVIDIA GeForce RTX 5090 Mobile (45,152, 1% lower), and the NVIDIA RTX A2000 (46,043, 1% higher). These delta values are all within 1%, showing that the Arc A730M’s average score is highly competitive with these other GPUs, but the average is not representative of its OpenCL performance, which is significantly higher.

In the head-to-head Geekbench OpenCL, the Intel part’s win is notable because the CMP 90HX has a 73.7% higher FP32 compute (21.89 vs 12.60 TFLOPS) and more than double the memory bandwidth (760.3 vs 336.0 GB/s). Yet the Arc A730M still outscores it. This implies that the OpenCL workload may be more sensitive to clock speed, texture throughput, or architecture-specific optimizations in Intel’s Xe-HPG design. The Arc A730M’s 2:1 FP16 ratio (25.19 TFLOPS) could also play a role if the test uses FP16 operations, though the database does not specify precision usage. The CMP 90HX’s lack of display outputs and mining-focused design may also mean its drivers or firmware are not optimized for general compute workloads like OpenCL, which could explain the slightly lower score.

The Verdict

From the recorded data, the Intel Arc A730M is the winner in the only direct benchmark comparison, taking Geekbench OpenCL by 1.9%. This makes it the better choice for users or systems that rely on OpenCL-based compute tasks, despite the NVIDIA CMP 90HX’s higher FP32 throughput and memory bandwidth. The Arc A730M also offers a higher boost clock (2050 MHz vs 1710 MHz) and better pixel/texture rates, which may translate to better performance in certain rasterization-heavy workloads, even though the CMP 90HX has more shading units (6400 vs 3072) and ray tracing cores (50 vs 24).

However, the NVIDIA CMP 90HX has a higher percentile ranking (90th vs 84th) and a higher average benchmark score (69,000 vs 45,592), though the latter is misleading due to the different test mixes. The CMP 90HX’s average score is based solely on Geekbench OpenCL, while the Arc A730M’s average includes a low 3DMark DX12 score. If the comparison is limited to the head-to-head test, the Intel part wins. For a user with a mining-focused workload, the CMP 90HX’s 10 GB GDDR6X memory and 760.3 GB/s bandwidth are advantageous, but the card’s lack of display outputs and PCIe 1.0 x4 interface make it unsuitable for general desktop use. The Arc A730M, as a mobile GPU, is designed for laptops and has portable device dependent outputs, making it the only option for portable systems.

The data suggests that the Intel Arc A730M is the more versatile and, in the tested metric, faster GPU. The NVIDIA CMP 90HX remains relevant for niche applications where its compute density and memory bandwidth are prioritized, but the benchmark results show no advantage over the Intel part in OpenCL. The Arc A730M’s higher pixel rate (196.8 vs 136.8 GPixel/s) and texture rate (393.6 vs 342.0 GTexel/s) further support its case for graphics workloads, despite the CMP 90HX’s larger shader count. The verdict, strictly from the data, is that the Intel Arc A730M is the better performer in the direct comparison, while the NVIDIA CMP 90HX offers a higher overall percentile ranking that reflects its position among all GPUs, not necessarily its superiority over this specific rival.

Specification Differences

The two GPUs differ across nearly every specification field. The NVIDIA CMP 90HX uses the GA102 chip on Samsung’s 8 nm process, while the Intel Arc A730M uses the DG2-512 chip on TSMC’s 6 nm process. The transistor counts are 28,300 million for NVIDIA and 21,700 million for Intel, with die sizes of 628 mm² and 406 mm² respectively. Transistor density is higher for Intel at 53.4 million per square millimeter versus NVIDIA’s 45.1 million.

Clock speeds differ notably: the CMP 90HX has a base clock of 1500 MHz and a boost of 1710 MHz, while the Arc A730M has a base of 1100 MHz and a boost of 2050 MHz. Memory clocks are 1188 MHz (19 Gbps effective) for NVIDIA and 1750 MHz (14 Gbps effective) for Intel. Memory capacity is 10 GB GDDR6X for NVIDIA versus 12 GB GDDR6 for Intel. Bus widths are 320-bit for NVIDIA and 192-bit for Intel, leading to memory bandwidth of 760.3 GB/s versus 336.0 GB/s.

Shading units are 6400 for NVIDIA and 3072 for Intel. TMUs are 200 for NVIDIA and 192 for Intel. ROPs are 80 for NVIDIA and 96 for Intel. Ray tracing cores are 50 for NVIDIA and 24 for Intel. Tensor cores are 200 for NVIDIA, while Intel has none listed. Pixel rate is 136.8 GPixel/s for NVIDIA and 196.8 GPixel/s for Intel. Texture rate is 342.0 GTexel/s for NVIDIA and 393.6 GTexel/s for Intel. FP32 compute is 21.89 TFLOPS for NVIDIA and 12.60 TFLOPS for Intel. FP16 is 21.89 TFLOPS (1:1) for NVIDIA and 25.19 TFLOPS (2:1) for Intel.

Power consumption is 320 W for NVIDIA with a dual-slot form factor and 2x 8-pin connectors, plus a suggested 700 W PSU. The Intel part has an 80 W TDP, an IGP slot width, no power connectors, and no suggested PSU. The bus interface is PCIe 1.0 x4 for NVIDIA and PCIe 4.0 x16 for Intel. Display outputs are absent on the CMP 90HX, while the Arc A730M has portable device dependent outputs. Dimensions are 285 mm (11.2 inches) in length and 112 mm (4.4 inches) in height for NVIDIA, while Intel’s dimensions are null. The NVIDIA card has a release date of 2021-07-27, while Intel’s release date is null. Both cards are end-of-life in production status. Both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. No launch MSRP is recorded for either product.