Intel Arc A370M vs NVIDIA A2 Comparison

Head-to-Head Benchmarks

The recorded benchmark data shows a consistent, if not overwhelming, advantage for the NVIDIA A2 across both tested workloads. In the Geekbench OpenCL test, the NVIDIA A2 scores 35,357 points, while the Intel Arc A370M trails at 29,676 points. This translates to a 19.1% lead for the NVIDIA part. The Vulkan results tell a similar story: the A2 posts 34,023 points against the Arc A370M's 28,673, a margin of 18.7%. In both cases, the NVIDIA A2 is the clear winner, and the database records two wins for the A2 and zero for the Intel part.

The margin is notable, but it is not a blowout. A 19% performance gap in synthetic compute workloads places these two GPUs in adjacent performance tiers. The Intel Arc A370M's average benchmark score across all recorded tests is 29,175, which puts it just 0.1% behind the AMD Radeon RX Vega M GH and 0.1% behind the AMD FirePro W8000. The NVIDIA A2, by contrast, averages 34,690, sitting 0.4% ahead of the NVIDIA T1000 8 GB and 1.4% ahead of the NVIDIA RTX A1000. So while the A2 is clearly faster in these head-to-head tests, the gap between the two is comparable to the gap between either card and its nearest rivals.

Interestingly, the Intel Arc A370M's Vulkan score of 28,673 is closer to its OpenCL score of 29,676 than the A2's Vulkan score is to its OpenCL score. The A2 drops from 35,357 in OpenCL to 34,023 in Vulkan, a reduction of roughly 3.8%. The Arc A370M drops from 29,676 to 28,673, a reduction of about 3.4%. This suggests both architectures handle the two APIs with similar efficiency, though the A2's absolute advantage persists regardless of API.

The percentile rankings reinforce this picture. The NVIDIA A2 sits at the 79th percentile among all GPUs, while the Intel Arc A370M sits at the 74th percentile. That five-point gap in percentile terms is consistent with the 18-19% delta in raw scores. It is worth remembering the A2's nearest rivals are all within 1.4% of its average score, meaning the A2 is not a dominant outlier in its tier. Similarly, the Arc A370M's nearest rivals are within 1% of its average score, with the AMD Radeon RX 6800M trailing by 1% and the AMD Radeon RX 470 trailing by 0.6%.

Architecture Differences

The two GPUs come from different foundries and process nodes. The NVIDIA A2 uses the GA107 chip built on an 8 nm process at Samsung, while the Intel Arc A370M uses the DG2-128 chip on a 6 nm process at TSMC. The smaller process node gives Intel a transistor density advantage: the A2 packs 8,700 million transistors into a 200 mm² die, yielding 43.5 million transistors per square millimeter. The Arc A370M packs 7,200 million transistors into a 157 mm² die, yielding 45.9 million transistors per square millimeter. Despite having fewer total transistors, the Intel chip is more densely packed.

The memory configurations differ significantly. The NVIDIA A2 comes with 16 GB of GDDR6 memory on a 128-bit bus, delivering 200.1 GB/s of bandwidth. The Intel Arc A370M has 4 GB of GDDR6 on a 64-bit bus, delivering 112.0 GB/s. That is a 44% bandwidth advantage for the A2, and the capacity difference is fourfold. The A2's memory runs at 1563 MHz (12.5 Gbps effective), while the Arc A370M's memory runs at 1750 MHz (14 Gbps effective). The Intel part has faster memory clocks but a narrower bus, which explains the lower total bandwidth.

Compute resources tell a more nuanced story. The NVIDIA A2 has 1,280 shading units, 40 texture mapping units, and 32 ROPs. It also includes 10 ray tracing cores and 40 tensor cores. The Intel Arc A370M has 1,024 shading units, 64 TMUs, and 32 ROPs, plus 8 ray tracing cores and no tensor cores. The A2 has more shading units and ray tracing cores, but the Arc A370M has more TMUs. This shows in the texture rate: the Arc A370M achieves 131.2 GTexel/s versus the A2's 70.80 GTexel/s, a near-doubling. The pixel rate is close, with the Arc A370M at 65.60 GPixel/s and the A2 at 56.64 GPixel/s.

Clock speeds also differ. The Intel part has a base clock of 1550 MHz and a boost clock of 2050 MHz, while the NVIDIA part has a base of 1440 MHz and a boost of 1770 MHz. The Intel chip boosts higher, but the A2's higher shading unit count compensates. FP32 compute is nearly identical: the A2 delivers 4.531 TFLOPS, and the Arc A370M delivers 4.198 TFLOPS. In FP16, the Intel part doubles its throughput to 8.397 TFLOPS via a 2:1 ratio, while the A2 maintains a 1:1 ratio at 4.531 TFLOPS.

Power and physical characteristics diverge sharply. The NVIDIA A2 is a single-slot card with no power connectors and a 60 W TDP, requiring a 250 W suggested PSU. The Intel Arc A370M is an integrated GPU (IGP) with a 35 W TDP and no separate power connector. The A2 has no display outputs, while the Arc A370M's outputs are listed as portable device dependent. Both use PCIe 4.0 x8 interfaces, and both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The A2 was released on November 9, 2021, while the Arc A370M followed on March 29, 2022. Both are now end-of-life products.

Where Each One Wins

The NVIDIA A2 wins in compute throughput and memory capacity. Its 16 GB frame buffer is four times larger than the Arc A370M's 4 GB, which matters for workloads that need to hold large datasets or models in VRAM. The A2's 200.1 GB/s bandwidth is also 79% higher than the Arc A370M's 112.0 GB/s. For OpenCL and Vulkan workloads, the A2's 19.1% and 18.7% leads respectively mean it is the stronger choice for general-purpose GPU compute, particularly in tasks that are memory-bandwidth-bound.

The Intel Arc A370M wins in texture throughput and FP16 compute. Its 131.2 GTexel/s texture rate is 85% higher than the A2's 70.80 GTexel/s. This gives it an edge in texture-heavy workloads, such as certain graphics effects or image processing pipelines. Its FP16 performance of 8.397 TFLOPS is 85% higher than the A2's 4.531 TFLOPS, which could benefit applications that leverage reduced-precision arithmetic. The Arc A370M also has a higher boost clock (2050 MHz versus 1770 MHz) and a higher pixel rate (65.60 GPixel/s versus 56.64 GPixel/s), though the pixel rate margin is modest at 16%.

The power envelope favors the Intel part. At 35 W TDP, the Arc A370M uses 42% less power than the A2's 60 W TDP. For mobile or power-constrained deployments, the Intel part is the more efficient option, though efficiency is not directly measured in the benchmark data. The A2's single-slot form factor and lack of power connectors make it suitable for dense server installations, while the Arc A370M's IGP status means it is designed to be integrated into a portable device.

In the benchmark data, the A2 wins both recorded tests. There is no recorded test where the Arc A370M comes out ahead. The nearest the Intel part gets is in the texture rate and FP16 metrics, which are not part of the head-to-head benchmark suite. For any user whose primary metric is the Geekbench OpenCL or Vulkan scores, the A2 is the clear winner.

The Verdict

The data points to the NVIDIA A2 as the stronger performer in standardized compute benchmarks. It leads by roughly 19% in both OpenCL and Vulkan, and it offers 16 GB of memory versus 4 GB, with more than 78% higher bandwidth. The A2 also has more shading units (1,280 versus 1,024) and more ray tracing cores (10 versus 8). For tasks that rely on raw compute throughput, memory capacity, or ray tracing, the A2 is the safer choice.

The Intel Arc A370M is not without merits. It has a higher texture rate (131.2 GTexel/s versus 70.80 GTexel/s) and higher FP16 throughput (8.397 TFLOPS versus 4.531 TFLOPS). It consumes less power (35 W versus 60 W) and is built on a more advanced 6 nm process with higher transistor density. For texture-heavy or FP16-optimized workloads, the Arc A370M could be the better fit. Its 4 GB memory, however, is a hard limitation for large workloads.

The database records two wins for the A2 and zero for the Arc A370M. The A2's average benchmark score of 34,690 places it at the 79th percentile, while the Arc A370M's 29,175 sits at the 74th percentile. These are not distant tiers, but the A2 is consistently ahead.

For users who need maximum compute performance, memory capacity, and ray tracing capability, the NVIDIA A2 is the recommended pick. For users who prioritize texture throughput, FP16 compute, or lower power consumption in a portable device, the Intel Arc A370M is the better fit, provided the 4 GB memory limit is acceptable. The Arc A370M's 8.397 TFLOPS FP16 is a significant advantage for applications that can use it, and its 131.2 GTexel/s texture rate is nearly double the A2's. Neither card dominates across all metrics, but the A2 wins where it matters most in the standard benchmark suite.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA A2 has an average score of 34,690, while the Intel Arc A370M has an average score of 29,175.

Q: How large is the performance gap in the head-to-head tests?

A: The NVIDIA A2 leads by 19.1% in Geekbench OpenCL and 18.7% in Geekbench Vulkan.

Q: What is the memory capacity difference?

A: The NVIDIA A2 has 16 GB of GDDR6 memory, while the Intel Arc A370M has 4 GB of GDDR6 memory.

Q: Which GPU has higher FP16 compute performance?

A: The Intel Arc A370M delivers 8.397 TFLOPS FP16, while the NVIDIA A2 delivers 4.531 TFLOPS FP16.

Q: What are the TDP values for each GPU?

A: The NVIDIA A2 has a TDP of 60 W, and the Intel Arc A370M has a TDP of 35 W.

Q: Are both GPUs still in production?

A: No, both are listed as end-of-life products. The NVIDIA A2 was released in November 2021, and the Intel Arc A370M was released in March 2022.