Intel Arc A570M vs NVIDIA CMP 90HX Comparison

Head-to-Head Benchmarks

The database records a single head-to-head comparison between the NVIDIA CMP 90HX and the Intel Arc A570M, and it is decisive. In the Geekbench OpenCL test, the NVIDIA CMP 90HX scores 69,000 points, while the Intel Arc A570M scores 58,239 points. That represents an 18.5% advantage for the NVIDIA part, a substantial margin in synthetic compute workloads. The CMP 90HX wins the only recorded head-to-head benchmark, giving it a clean 1-0 record in this comparison.

Looking at the broader context, the NVIDIA CMP 90HX sits at the 90th percentile among all GPUs in the database, while the Intel Arc A570M sits at the 88th percentile. On the surface, a two-percentile gap seems modest, but the raw score difference tells a more compelling story. The CMP 90HX's closest rivals in the database include the Intel Arc A770 at 68,809 points (a 0.3% difference), the AMD Radeon Instinct MI25 at 68,562 points (0.6% behind), the AMD Radeon Pro WX 8200 at 69,870 points (1.2% ahead), and the NVIDIA Quadro P6000 at 69,986 points (1.4% ahead). This clustering shows the CMP 90HX is operating in a tightly contested performance band, with its nearest competitors all within roughly 1.5% of its score.

The Intel Arc A570M, by contrast, finds itself among different company. Its nearest rivals include the AMD Radeon RX 6950 XT at 58,392 points (0.3% ahead), the AMD Radeon RX 5600 OEM at 58,085 points (0.3% behind), the NVIDIA P102-100 at 58,528 points (0.5% ahead), and the AMD Radeon PRO V710 at 58,657 points (0.7% ahead). The A570M's position here is more ambiguous: it is nearly tied with its closest competitors, with deltas under 1% in either direction. This suggests the A570M is not a clear winner in its own tier, whereas the CMP 90HX, despite its narrow margins, holds a slight edge over its immediate rivals.

The delta between the two cards in this comparison, 18.5%, is far larger than the deltas separating each card from its own nearest rivals. This indicates that the CMP 90HX and the Arc A570M are not really competing in the same performance class, despite the database placing them in a head-to-head context. The CMP 90HX is closer to the Arc A770 than it is to the A570M, while the A570M is closer to the RX 6950 XT than it is to the CMP 90HX. The 18.5% gap is a chasm in synthetic OpenCL performance.

Architecture Differences

The two GPUs come from fundamentally different design philosophies. The NVIDIA CMP 90HX is built on the GA102 chip using the Ampere architecture, fabricated on an 8 nm process at Samsung. It packs 28,300 million transistors into a die size of 628 mm², yielding a transistor density of 45.1 million transistors per square millimeter. The Intel Arc A570M uses the DG2-256 chip with the Xe-HPG architecture, built on a 6 nm process at TSMC. It contains 11,500 million transistors on a 269 mm² die, giving a density of 42.8 million transistors per square millimeter.

The transistor counts are starkly different: the NVIDIA chip has roughly 2.5 times as many transistors as the Intel chip. The die size difference is even more pronounced, with the GA102 being more than twice the physical area of the DG2-256. Interestingly, the transistor density is nearly identical, with the NVIDIA part at 45.1M per mm² and the Intel part at 42.8M per mm², a difference of only about 5%. This suggests both architectures are similarly efficient at packing transistors into silicon, but NVIDIA chose to deploy far more of them.

The memory subsystems diverge significantly. The CMP 90HX uses 10 GB of GDDR6X on a 320-bit bus, delivering 760.3 GB/s of bandwidth. The Arc A570M uses 8 GB of GDDR6 on a 128-bit bus, delivering 224.0 GB/s. The bandwidth difference is enormous: the NVIDIA part offers more than three times the memory bandwidth of the Intel part. This is a direct consequence of the wider bus and faster memory type, and it likely explains a large portion of the OpenCL performance gap.

Compute resources also differ by a wide margin. The CMP 90HX has 6,400 shading units, 200 texture mapping units, 80 render output units, 50 ray tracing cores, and 200 tensor cores. The Arc A570M has 2,048 shading units, 128 TMUs, 64 ROPs, and 16 ray tracing cores, with no tensor cores listed. The NVIDIA part has roughly three times the shading units and more than three times the ray tracing cores. The pixel rate for the CMP 90HX is 136.8 GPixel/s versus 83.20 GPixel/s for the A570M, and the texture rate is 342.0 GTexel/s versus 166.4 GTexel/s.

Floating-point performance tells a similar story. The CMP 90HX delivers 21.89 TFLOPS in both FP32 and FP16, operating at a 1:1 ratio. The Arc A570M delivers 5.325 TFLOPS in FP32 and 10.65 TFLOPS in FP16, at a 2:1 ratio. The NVIDIA card offers more than four times the FP32 throughput, and even its FP16 figure is roughly double the Intel card's FP16 output. The 1:1 ratio on the CMP 90HX means it does not accelerate FP16 relative to FP32, whereas the A570M's 2:1 ratio gives it a modest FP16 boost.

Clock speeds are also quite different. The CMP 90HX runs at a base clock of 1500 MHz and a boost clock of 1710 MHz. The Arc A570M runs at 900 MHz base and 1300 MHz boost. The NVIDIA part operates at substantially higher frequencies, which compounds its architectural advantages.

Where Each One Wins

The NVIDIA CMP 90HX wins in every measured benchmark category, but the nature of the product suggests specific use cases. The card has no display outputs, which means it is not intended for traditional gaming or workstation visual output. Its architecture, however, is fully capable of supporting DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, so it retains feature compatibility with modern graphics APIs even though it cannot drive a monitor. The card's massive bandwidth and compute throughput make it suited for raw compute tasks, and its 320 W TDP with dual-slot cooling and dual 8-pin power connectors indicates a desktop-bound, high-power design.

The Intel Arc A570M, by contrast, is a mobile part. Its slot width is listed as IGP, meaning integrated graphics package, and its display outputs are portable-device dependent. It has a 75 W TDP and no separate power connectors, reflecting its low-power, mobile-oriented design. It supports the same API set, including DirectX 12 Ultimate, OpenGL 4.6, and Vulkan 1.4, so it can handle modern workloads when paired with a host device's display. Its smaller memory footprint of 8 GB and narrower 128-bit bus are appropriate for its form factor, but they limit its bandwidth-intensive performance.

In practical terms, the CMP 90HX is the clear choice for compute-heavy, bandwidth-hungry workloads where display output is irrelevant. Its 760.3 GB/s bandwidth and 21.89 TFLOPS FP32 performance put it in a class that can handle large data sets and heavy parallel processing. The A570M, with its 224.0 GB/s bandwidth and 5.325 TFLOPS FP32, is better suited to mobile workloads where power efficiency and size matter more than raw throughput. Its 75 W power draw is a fraction of the CMP 90HX's 320 W, making it viable for laptops and compact systems.

The benchmark data confirms this split. The 18.5% OpenCL gap is consistent with the architectural disparity: more shading units, more memory bandwidth, higher clocks, and a larger transistor budget all favor the NVIDIA part. The A570M's advantage is not in performance but in power envelope and physical integration. It is an active product, while the CMP 90HX is end-of-life, so availability favors the Intel part in new systems.

Specification Differences

The two cards differ across nearly every specification category. The process node is 8 nm for the NVIDIA part versus 6 nm for the Intel part. The foundry is Samsung for NVIDIA and TSMC for Intel. Transistor counts are 28,300 million versus 11,500 million, and die sizes are 628 mm² versus 269 mm². Transistor density is 45.1M per mm² versus 42.8M per mm².

Clock speeds differ substantially: base clock 1500 MHz versus 900 MHz, boost clock 1710 MHz versus 1300 MHz, and memory clock 1188 MHz (19 Gbps effective) versus 1750 MHz (14 Gbps effective). Memory size is 10 GB versus 8 GB, memory type is GDDR6X versus GDDR6, bus width is 320 bit versus 128 bit, and bandwidth is 760.3 GB/s versus 224.0 GB/s.

Compute units differ: shading units 6,400 versus 2,048, TMUs 200 versus 128, ROPs 80 versus 64, RT cores 50 versus 16, and tensor cores 200 versus none listed. Pixel rate is 136.8 GPixel/s versus 83.20 GPixel/s, texture rate is 342.0 GTexel/s versus 166.4 GTexel/s, FP32 is 21.89 TFLOPS versus 5.325 TFLOPS, and FP16 is 21.89 TFLOPS (1:1) versus 10.65 TFLOPS (2:1).

Power and physical specifications also differ. TDP is 320 W versus 75 W. The CMP 90HX is dual-slot with 2x 8-pin power connectors and a suggested PSU of 700 W, while the A570M is IGP with no power connectors and no suggested PSU. Bus interface is PCIe 1.0 x4 for the NVIDIA part versus PCIe 4.0 x8 for the Intel part, an unusual pairing where the older architecture has a slower bus. Display outputs are none for the CMP 90HX versus portable-device dependent for the A570M. The CMP 90HX has dimensions of 285 mm length and 112 mm height, while the A570M has no listed dimensions.

Production status differs: end-of-life for the CMP 90HX versus active for the A570M. Release dates differ as well, with the CMP 90HX released in July 2021 and the A570M in July 2023. The API support is identical for both: DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

FAQ

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

A: The NVIDIA CMP 90HX scores 69,000 in Geekbench OpenCL, while the Intel Arc A570M scores 58,239. The CMP 90HX leads by 18.5%.

Q: How does the memory bandwidth compare between the two cards?

A: The NVIDIA CMP 90HX delivers 760.3 GB/s from 10 GB of GDDR6X on a 320-bit bus. The Intel Arc A570M delivers 224.0 GB/s from 8 GB of GDDR6 on a 128-bit bus.

Q: What are the power requirements for each GPU?

A: The NVIDIA CMP 90HX has a 320 W TDP, uses dual-slot cooling with 2x 8-pin power connectors, and suggests a 700 W power supply. The Intel Arc A570M has a 75 W TDP, is an IGP form factor, and has no separate power connectors or suggested PSU.

Q: Do both GPUs support the same graphics APIs?

A: Yes, both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Q: Which GPU is currently in production?

A: The Intel Arc A570M is listed as active, while the NVIDIA CMP 90HX is end-of-life. The CMP 90HX was released in July 2021, and the A570M was released in July 2023.

Q: How do the transistor counts compare?

A: The NVIDIA CMP 90HX contains 28,300 million transistors on a 628 mm² die, while the Intel Arc A570M contains 11,500 million transistors on a 269 mm² die. Transistor density is 45.1M per mm² for NVIDIA and 42.8M per mm² for Intel.