Intel Arc A770 vs NVIDIA CMP 90HX Comparison

Head-to-Head Benchmarks

The only direct benchmark comparison in the database is Geekbench OpenCL, and the result is a decisive win for the Intel Arc A770. The Arc A770 scores 109,175, while the NVIDIA CMP 90HX trails at 69,000. That translates to a 36.8% deficit for the NVIDIA card—a substantial margin that puts the Arc A770 in a different performance class for compute-heavy OpenCL workloads.

Looking at the aggregate benchmark scores, the two cards are nearly inseparable. The CMP 90HX averages 69,000 across all tests, while the Arc A770 averages 68,809. That is a difference of only 0.3% in favor of the NVIDIA part. In practical terms, the average score says these are equivalent performers overall, but the head-to-head OpenCL result reveals a specific workload where Intel holds a clear advantage.

The nearest rival data reinforces the picture. The CMP 90HX sits 0.3% above the Arc A770 in average score, but the Arc A770 is 0.4% above the AMD Radeon Instinct MI25 and 1.5% below the AMD Radeon Pro WX 8200. The CMP 90HX, by comparison, is 0.6% above the MI25 and 1.2% below the Pro WX 8200. Both cards occupy the same performance band—roughly the 90th percentile among all GPUs—but the Arc A770's OpenCL advantage is the standout data point.

The Arc A770 also brings additional benchmark results to the table: a 3DMark Steel Nomad DX12 score of 2,969 and a Geekbench Vulkan score of 94,284. The CMP 90HX has no corresponding entries for those tests, so no direct comparison is possible. However, the Vulkan score suggests the Intel card handles modern graphics APIs well, and the DX12 result indicates solid rasterization performance for a card in this class.

FAQ

Q: Which card wins in Geekbench OpenCL?

A: The Intel Arc A770 wins decisively, scoring 109,175 versus the NVIDIA CMP 90HX's 69,000. That is a 36.8% advantage for Intel in this specific test.

Q: How do the overall average benchmark scores compare?

A: They are nearly identical. The CMP 90HX averages 69,000, and the Arc A770 averages 68,809—a gap of just 0.3% in favor of NVIDIA.

Q: What is the percentile ranking for these GPUs?

A: Both cards sit at the 90th percentile among all GPUs in the database, meaning they outperform roughly nine out of ten competing products.

Q: Does the Arc A770 support modern graphics APIs?

A: Yes. The Arc A770 supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The CMP 90HX supports the same API set.

Q: What memory configurations do these cards have?

A: The NVIDIA CMP 90HX has 10 GB of GDDR6X on a 320-bit bus, delivering 760.3 GB/s of bandwidth. The Intel Arc A770 has 16 GB of GDDR6 on a 256-bit bus, delivering 512.0 GB/s.

Q: Which card has more shading units?

A: The CMP 90HX has 6,400 shading units, while the Arc A770 has 4,096. However, the Arc A770 has more texture mapping units (256 versus 200) and more ROPs (128 versus 80).

Where Each One Wins

The Intel Arc A770 is the clear winner in OpenCL compute workloads. The 36.8% lead in Geekbench OpenCL is not a marginal difference—it suggests the Xe-HPG architecture is better optimized for this type of parallel compute. This makes the Arc A770 the more sensible choice for tasks that rely heavily on OpenCL, such as certain scientific simulations, video encoding pipelines, or machine learning inference workloads that use this API.

The Arc A770 also has the advantage of being a fully functional graphics card. It has display outputs (1x HDMI 2.1 and 3x DisplayPort 2.0), while the CMP 90HX has no display outputs at all. For any use case that requires connecting a monitor—gaming, content creation, or general desktop use—the Intel card is the only viable option.

The NVIDIA CMP 90HX wins on memory bandwidth. With 760.3 GB/s of bandwidth versus the Arc A770's 512.0 GB/s, the NVIDIA card has a 48.5% advantage in raw memory throughput. This could matter in workloads that are bandwidth-limited rather than compute-limited, such as certain data processing tasks or high-resolution texture streaming. The CMP 90HX also has a higher FP32 compute rating at 21.89 TFLOPS versus the Arc A770's 19.66 TFLOPS, though the Intel card counters with a much higher FP16 rating of 39.32 TFLOPS (versus NVIDIA's 21.89 TFLOPS).

Specification Differences

The two cards differ across nearly every major specification. The CMP 90HX uses a GA102 chip built on Samsung's 8 nm process, while the Arc A770 uses a DG2-512 chip on TSMC's 6 nm process. The NVIDIA card has a larger die at 628 mm² versus 406 mm², but the Intel card has a higher transistor density at 53.4 million per mm² versus 45.1 million per mm². Total transistor counts are 28,300 million for NVIDIA and 21,700 million for Intel.

Clock speeds favor Intel significantly. The Arc A770 has a base clock of 2100 MHz and a boost clock of 2400 MHz, while the CMP 90HX runs at 1500 MHz base and 1710 MHz boost. Memory clocks also differ: the NVIDIA card runs at 1188 MHz (19 Gbps effective), while the Intel card runs at 2000 MHz (16 Gbps effective). The CMP 90HX has a wider 320-bit memory bus versus Intel's 256-bit bus, which explains NVIDIA's bandwidth advantage despite the lower clock.

Power requirements differ meaningfully. The CMP 90HX has a 320 W TDP and requires a 700 W power supply with two 8-pin connectors. The Arc A770 has a 225 W TDP and needs a 550 W power supply with one 6-pin and one 8-pin connector. The bus interface also differs: the CMP 90HX uses PCIe 1.0 x4 (an odd choice for a modern card), while the Arc A770 uses PCIe 4.0 x16.

Architecture Differences

The two cards represent fundamentally different architectures. The NVIDIA CMP 90HX is built on Ampere, which is a mature architecture designed primarily for gaming and professional workloads. It includes 50 RT cores and 200 tensor cores, which are dedicated to ray tracing and AI acceleration respectively. The Intel Arc A770 is built on Xe-HPG, a first-generation gaming architecture from Intel that includes 32 RT cores but no tensor cores.

The compute characteristics differ in important ways. The CMP 90HX delivers 21.89 TFLOPS in both FP32 and FP16, indicating a 1:1 ratio. The Arc A770 delivers 19.66 TFLOPS in FP32 but 39.32 TFLOPS in FP16, a 2:1 ratio that doubles throughput for half-precision workloads. This makes the Intel card more capable in applications that leverage FP16, such as certain AI inference tasks or graphics effects that use half-precision math.

The memory subsystems are also architecturally distinct. The CMP 90HX uses GDDR6X, a faster memory type with higher bandwidth per pin, while the Arc A770 uses standard GDDR6. The NVIDIA card's 320-bit bus combined with GDDR6X yields 760.3 GB/s, while Intel's 256-bit bus with GDDR6 yields 512.0 GB/s. The Intel card compensates with a higher pixel rate (307.2 GPixel/s versus 136.8 GPixel/s) and texture rate (614.4 GTexel/s versus 342.0 GTexel/s), suggesting better efficiency in rasterization throughput despite the lower memory bandwidth.

The Verdict

The data presents a clear split: the Intel Arc A770 is the better all-around graphics card, while the NVIDIA CMP 90HX has specific strengths that matter in narrow use cases.

For anyone building a system that needs display output, the Arc A770 is the only choice. The CMP 90HX has no display outputs, making it unsuitable for gaming or any desktop use. The Intel card also wins decisively in OpenCL compute, with a 36.8% advantage that cannot be ignored. Add in the higher clock speeds, lower power draw (225 W versus 320 W), smaller power supply requirement (550 W versus 700 W), and the significantly larger 16 GB memory pool, and the Arc A770 is the more versatile and practical card.

The NVIDIA CMP 90HX makes sense only for specialized compute workloads where its 760.3 GB/s memory bandwidth and higher FP32 throughput (21.89 TFLOPS) provide an advantage. Its 10 GB of GDDR6X on a 320-bit bus is a niche asset, but the lack of display outputs and PCIe 1.0 x4 interface severely limit its utility. The PCIe 1.0 x4 bus is a particularly concerning bottleneck—it could throttle data transfer in ways that negate the bandwidth advantage in real-world workloads.

The average benchmark scores are nearly tied, but the head-to-head result is not. The Arc A770's 36.8% OpenCL lead is the single most important performance data point, and it aligns with the card's other strengths. The CMP 90HX is an end-of-life mining card with a narrow purpose, while the Arc A770 is also end-of-life but offers a full feature set, modern API support, and competitive performance across multiple benchmarks.

Choose the Intel Arc A770 unless your specific workload demands the CMP 90HX's unique memory configuration and higher FP32 compute. For everything else—gaming, content creation, general compute, or any task that needs a display—the Arc A770 is the data-backed recommendation.