Intel Arc A310 vs NVIDIA GeForce GTX 660 Comparison

NVIDIA GeForce GTX 660 and Intel Arc A310 occupy very different positions in the GPU landscape, separated by a decade of architectural evolution. The data shows a clear overall winner in compute performance, but the GTX 660 still holds its own in specific legacy scenarios. In the recorded head-to-head benchmarks, the Intel Arc A310 wins both tests decisively, yet the NVIDIA card’s average benchmark score and percentile ranking tell a more nuanced story. This analysis relies strictly on database measurements to compare the two.

Head-to-Head Benchmarks

The database records two direct comparisons between these cards, and Intel wins both. In Geekbench OpenCL, the Arc A310 scores 30,607 against the GTX 660’s 11,347. That is a 62.9% advantage for Intel. The margin is substantial, nearly triple the score. For Vulkan, the Arc A310 scores 28,964, while the GTX 660 manages 11,415. Intel leads by 60.6% in that test. These are not close calls; the Arc A310 delivers roughly two and a half times the raw compute throughput in both APIs.

However, the average benchmark score across all recorded tests flips the expectation. The GTX 660 has an average benchmark score of 9,022, while the Arc A310 averages 7,550. This discrepancy exists because the database includes a broader set of tests for the Arc A310, including several Passmark tests where the card scores very low. For example, the Arc A310 records Passmark DirectX 10 at 31, DirectX 11 at 33, DirectX 12 at 29, and DirectX 9 at 69. These legacy API scores drag down its average. The GTX 660, with its three Geekbench scores only, sits at a higher average. This is a critical distinction: on modern compute APIs, Intel wins; on legacy rasterization APIs, the data suggests the GTX 660 is far more consistent.

The percentile rankings reflect this split. The GTX 660 sits at the 45th percentile among all GPUs, while the Arc A310 sits at the 40th. So although the Arc A310 wins the direct head-to-head compute tests, its overall standing in the database is lower. The nearest rivals for the GTX 660 include the NVIDIA TITAN V CEO Edition (average score 9,037, only 0.2% higher) and the GeForce GTX 560 (9,058, 0.4% higher). The Arc A310’s nearest rivals include the AMD Radeon R7 250 (7,557, 0.1% higher) and the Radeon Pro WX 3100 (7,580, 0.4% higher). This means the GTX 660 competes in a higher performance tier on average, despite losing the two compute tests.

Where Each One Wins

The Arc A310 wins in compute-heavy workloads. The OpenCL and Vulkan results are the clearest evidence. OpenCL is used for general-purpose GPU computing, such as video encoding, physics simulations, and some machine learning tasks. Vulkan is a modern graphics API that also handles compute workloads. The Arc A310’s scores in both, 30,607 and 28,964 respectively, indicate it is the better choice for applications built on these modern interfaces. The GTX 660’s scores, 11,347 and 11,415, are less than half of that. If a user’s software relies on Vulkan or OpenCL, the Arc A310 is clearly superior.

The GTX 660 wins in average benchmark score, which reflects consistency across a range of tests. Its 9,022 average is 19.5% higher than the Arc A310’s 7,550. This is partly due to the PassMark DirectX tests. The Arc A310’s PassMark DirectX 9 score is 69, while its DirectX 12 score is 29. These numbers are extremely low compared to its Geekbench results. The GTX 660 simply has no such recorded low scores. Therefore, for any workload measured by PassMark’s legacy DirectX tests, the GTX 660 is the safer bet. The PassMark G3D score for the Arc A310 is 5,433, and its G2D score is 625, which are moderate, but the DirectX 9-12 scores are anomalously low.

The Arc A310 also wins in pixel throughput. Its pixel rate is 28.00 GPixel/s, compared to the GTX 660’s 20.64 GPixel/s. That is a 35.7% advantage. Texture rate goes the other way: the GTX 660 has 82.56 GTexel/s versus the Arc A310’s 56.00 GTexel/s, a 47.4% advantage for NVIDIA. So the GTX 660 is stronger at texturing, while the Arc A310 is stronger at fill rate.

Architecture Differences

The two cards come from completely different eras. The GTX 660 uses the GK106 chip on the Kepler architecture, built on a 28 nm process at TSMC. It has 2,540 million transistors on a 221 mm² die. The Arc A310 uses the DG2-128 chip on the Xe-HPG architecture, built on a 6 nm process at TSMC, with 7,200 million transistors on a 157 mm² die. The Arc A310 is far more transistor-dense, at 45.9 million transistors per mm², versus the GTX 660’s 11.5 million. The node shrink and density difference explain how the Arc A310 achieves higher compute with fewer shading units.

The shading unit counts differ inversely to compute. The GTX 660 has 960 shading units, 80 texture mapping units, and 24 raster operation units. The Arc A310 has 768 shading units, 32 TMUs, and 16 ROPs. Despite fewer units, the Arc A310’s FP32 throughput is 2.688 TFLOPS, versus the GTX 660’s 1.981 TFLOPS. That is a 35.7% higher raw floating-point performance. The Arc A310 also supports FP16 at 5.376 TFLOPS, while the GTX 660 has no recorded FP16 capability. The Arc A310 includes 6 ray-tracing cores, which the GTX 660 lacks entirely. This makes the Arc A310 compatible with modern games that use ray-traced effects.

Memory systems are also different. The GTX 660 has 2 GB of GDDR5 memory on a 192-bit bus, with a bandwidth of 144.2 GB/s. The Arc A310 has 4 GB of GDDR6 memory on a 64-bit bus, with a bandwidth of 124.0 GB/s. The GTX 660 has a wider bus and higher bandwidth, but the Arc A310 has double the capacity. Clock speeds: the GTX 660 runs at 980 MHz base and 1032 MHz boost, while the Arc A310 runs at 1750 MHz base and boost. The memory clock for the GTX 660 is 1502 MHz (6 Gbps effective), versus 1937 MHz (15.5 Gbps effective) for the Arc A310.

The power profiles are drastically different. The GTX 660 has a TDP of 140 W and requires a single 6-pin power connector, with a suggested power supply of 300 W. The Arc A310 has a TDP of 30 W, requires no power connector, and has a suggested PSU of 200 W. The Arc A310 is a single-slot card, while the GTX 660 is dual-slot. The Arc A310 also uses PCIe 4.0 x8, whereas the GTX 660 uses PCIe 3.0 x16. Display outputs differ: the GTX 660 has 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2; the Arc A310 has 4x mini-DisplayPort 2.0. API support: the GTX 660 supports DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. The Arc A310 supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The Arc A310 is newer and more feature-complete.

FAQ

Q: Which card has a higher average benchmark score?

A: The NVIDIA GeForce GTX 660 has an average benchmark score of 9,022, while the Intel Arc A310 has an average of 7,550. The GTX 660 is 19.5% higher on average.

Q: Does the Intel Arc A310 support ray tracing?

A: Yes. The Arc A310 has 6 ray tracing cores. The GTX 660 has none.

Q: Which card performs better in the Geekbench Vulkan test?

A: The Intel Arc A310 scores 28,964, which is 60.6% higher than the GTX 660’s 11,415.

Q: What is the memory capacity difference?

A: The Arc A310 has 4 GB of GDDR6 memory, while the GTX 660 has 2 GB of GDDR5 memory. The GTX 660 has a wider 192-bit bus and higher bandwidth at 144.2 GB/s, versus the Arc A310’s 124.0 GB/s.

Q: Is the GTX 660 more power-efficient?

A: No. The GTX 660 has a TDP of 140 W, while the Arc A310 has a TDP of 30 W. The Arc A310 requires no power connector.

Q: Which card has a higher transistor density?

A: The Arc A310 has 45.9 million transistors per mm², while the GTX 660 has 11.5 million per mm².

Specification Differences

The two cards differ in almost every measurable specification.

• Process Node: GTX 660 uses 28 nm, Arc A310 uses 6 nm.
• Transistors: GTX 660 has 2,540 million, Arc A310 has 7,200 million.
• Die Size: GTX 660 is 221 mm², Arc A310 is 157 mm².
• Transistor Density: GTX 660 has 11.5M per mm², Arc A310 has 45.9M per mm².
• Base Clock: GTX 660 is 980 MHz, Arc A310 is 1750 MHz.
• Boost Clock: GTX 660 is 1032 MHz, Arc A310 is 1750 MHz.
• Memory Clock: GTX 660 is 1502 MHz (6 Gbps effective), Arc A310 is 1937 MHz (15.5 Gbps effective).
• Memory Size: GTX 660 is 2 GB, Arc A310 is 4 GB.
• Memory Type: GTX 660 uses GDDR5, Arc A310 uses GDDR6.
• Bus Width: GTX 660 is 192-bit, Arc A310 is 64-bit.
• Bandwidth: GTX 660 is 144.2 GB/s, Arc A310 is 124.0 GB/s.
• Shading Units: GTX 660 has 960, Arc A310 has 768.
• TMUs: GTX 660 has 80, Arc A310 has 32.
• ROPs: GTX 660 has 24, Arc A310 has 16.
• RT Cores: GTX 660 has none, Arc A310 has 6.
• Pixel Rate: GTX 660 is 20.64 GPixel/s, Arc A310 is 28.00 GPixel/s.
• Texture Rate: GTX 660 is 82.56 GTexel/s, Arc A310 is 56.00 GTexel/s.
• FP32: GTX 660 is 1.981 TFLOPS, Arc A310 is 2.688 TFLOPS.
• FP16: GTX 660 has none, Arc A310 has 5.376 TFLOPS.
• TDP: GTX 660 is 140 W, Arc A310 is 30 W.
• Slot Width: GTX 660 is dual-slot, Arc A310 is single-slot.
• Power Connectors: GTX 660 requires 1x 6-pin, Arc A310 requires none.
• Suggested PSU: GTX 660 is 300 W, Arc A310 is 200 W.
• Bus Interface: GTX 660 is PCIe 3.0 x16, Arc A310 is PCIe 4.0 x8.
• Display Outputs: GTX 660 has 2x DVI, 1x HDMI, 1x DisplayPort; Arc A310 has 4x mini-DisplayPort 2.0.
• DirectX Support: GTX 660 supports 12 (11_0), Arc A310 supports 12 Ultimate.
• Vulkan Version: GTX 660 supports 1.2.175, Arc A310 supports 1.4.
• Release Date: GTX 660 is 2012-09-05, Arc A310 is 2022-10-11.
• Launch MSRP: GTX 660 is 229 USD, Arc A310 has no recorded launch MSRP.

The Verdict

The data clearly favors the Intel Arc A310 for users who rely on modern compute APIs. Its OpenCL and Vulkan performance is superior, its FP32 throughput is 35.7% higher, and it offers ray tracing support, 4 GB of memory, and a much lower power draw. The GTX 660 is a decade older, and its architectural limitations show in the compute tests. The Arc A310’s 30 W TDP also makes it a far better fit for small form factor systems or low-power builds.

However, the GTX 660 is not without merit. Its average benchmark score is higher, and its texture rate is significantly better. For older software that relies on DirectX 9, 10, or 11, the GTX 660’s scores are not recorded, but the Arc A310’s PassMark DirectX scores are critically low, ranging from 29 to 69. The GTX 660 also has higher memory bandwidth, which can benefit certain texture-heavy workloads. The 45th percentile ranking of the GTX 660 versus the 40th for the Arc A310 suggests that, across the full database, the GTX 660 is slightly more competitive.

The verdict: choose the Intel Arc A310 for any modern graphics or compute task, especially those using Vulkan, OpenCL, or ray tracing. Choose the NVIDIA GeForce GTX 660 only if the workload relies on legacy DirectX APIs and the higher texture fill rate is essential. The GTX 660 also has the advantage of a known launch MSRP of 229 USD, whereas the Arc A310 has no recorded launch price. For most current applications, the Arc A310’s performance lead in the head-to-head tests is decisive. The GTX 660 is end-of-life for a reason: its compute power is a generation old. The Arc A310, despite its lower average score, is the more capable card for the future.