NVIDIA GeForce GTX 1070 vs NVIDIA GeForce GTX 1660 Comparison

FAQ

Q: Which GPU wins more benchmark tests?

A: The NVIDIA GeForce GTX 1070 wins 7 of the 10 recorded head-to-head benchmarks, while the GTX 1660 wins 3.

Q: How much faster is the GTX 1660 in Vulkan?

A: The GTX 1660 scores 50137 in Geekbench Vulkan versus 22121 for the GTX 1070, a 126.6% advantage.

Q: What is the memory capacity difference?

A: The GTX 1070 has 8 GB of GDDR5 on a 256-bit bus, while the GTX 1660 has 6 GB of GDDR5 on a 192-bit bus.

Q: Which card has a higher average benchmark score?

A: The GTX 1660 has an average benchmark score of 11680, compared to 9780 for the GTX 1070.

Q: What are the launch MSRPs?

A: The GTX 1660 launched with an MSRP of 219 USD, and the GTX 1070 launched at 379 USD.

Q: How do their percentiles compare?

A: The GTX 1660 sits at the 51st percentile of all GPUs, while the GTX 1070 is at the 47th percentile.

Architecture Differences

The GTX 1660 uses the TU116 chip built on Turing architecture at TSMC's 12 nm process, while the GTX 1070 uses the GP104 chip on Pascal architecture at 16 nm. The Turing card packs 6,600 million transistors into a 284 mm² die, yielding a density of 23.2M per mm². The Pascal card has 7,200 million transistors across a 314 mm² die, for a density of 22.9M per mm².

The GTX 1660 features 1408 shading units, 88 texture mapping units, and 48 raster output pipelines. The GTX 1070 is more generously equipped with 1920 shading units, 120 TMUs, and 64 ROPs. Neither card includes ray tracing cores or tensor cores, so both rely on traditional rasterization.

Clock speeds differ meaningfully. The GTX 1660 runs at a base of 1530 MHz and boosts to 1785 MHz. The GTX 1070 has a lower base of 1506 MHz and a boost of 1683 MHz. Memory clocks are effectively identical at 8 Gbps, but the GTX 1070's wider 256-bit bus gives it substantially more bandwidth.

The GTX 1070's FP32 throughput is 6.463 TFLOPS versus 5.027 TFLOPS for the GTX 1660. In FP16, the GTX 1660 achieves 10.05 TFLOPS with a 2:1 ratio, while the GTX 1070 manages only 101.0 GFLOPS at a 1:64 ratio, a massive difference in half-precision compute.

The GTX 1070 is physically larger at 267 mm long, 112 mm tall, and 40 mm wide. The GTX 1660 measures 229 mm by 111 mm by 35 mm. Both are dual-slot cards with a single 8-pin power connector, but the GTX 1070 requires a 450 W suggested power supply versus 300 W for the GTX 1660. The GTX 1070 also offers three DisplayPort outputs plus one HDMI and one DVI, while the GTX 1660 has one DisplayPort, one HDMI, and one DVI.

Head-to-Head Benchmarks

The dominant pattern in the recorded data is the GTX 1070's strength across legacy DirectX tests. In Passmark DirectX 10, the GTX 1070 scores 82 versus 61 for the GTX 1660, a 25.6% lead. DirectX 11 shows a similar gap: 100 versus 79, a 21% advantage. Even DirectX 9 favors the older card, with 197 versus 177, a 10.2% margin.

The GTX 1070 also wins in general compute and 3D workloads. Passmark G3D shows 13498 for the GTX 1070 against 11646 for the GTX 1660, a 13.7% gap. Passmark GPU Compute gives 6102 versus 4963, an 18.7% lead. The GTX 1070 also takes Passmark G2D at 846 versus 776, an 8.3% edge.

The GTX 1660's wins are concentrated in modern APIs and compute interfaces. Geekbench Vulkan is the standout: 50137 versus 22121, a 126.6% advantage that dwarfs every other delta in the comparison. Geekbench OpenCL also goes to the GTX 1660, with 47850 versus 44700, a 7% lead. In Passmark DirectX 12, the GTX 1660 wins narrowly at 49 versus 48, a 2.1% margin.

The 3DMark Steel Nomad DX12 test is close, with the GTX 1070 edging out 1082 versus 1065, a 1.6% difference. This near-tie suggests that in modern DX12 gaming workloads, the two cards perform comparably despite their architectural differences.

Specification Differences

| Specification | GTX 1660 | GTX 1070 |

|---|---|---|

| Architecture | Turing | Pascal |

| Process Node | 12 nm | 16 nm |

| Transistors | 6,600 million | 7,200 million |

| Die Size | 284 mm² | 314 mm² |

| Base Clock | 1530 MHz | 1506 MHz |

| Boost Clock | 1785 MHz | 1683 MHz |

| Memory Size | 6 GB | 8 GB |

| Memory Bus Width | 192 bit | 256 bit |

| Memory Bandwidth | 192.1 GB/s | 256.3 GB/s |

| Shading Units | 1408 | 1920 |

| TMUs | 88 | 120 |

| ROPs | 48 | 64 |

| FP32 Performance | 5.027 TFLOPS | 6.463 TFLOPS |

| FP16 Performance | 10.05 TFLOPS (2:1) | 101.0 GFLOPS (1:64) |

| Pixel Rate | 85.68 GPixel/s | 107.7 GPixel/s |

| Texture Rate | 157.1 GTexel/s | 202.0 GTexel/s |

| TDP | 120 W | 150 W |

| Suggested PSU | 300 W | 450 W |

| Length | 229 mm | 267 mm |

| Width | 35 mm | 40 mm |

| Display Outputs | 1x DVI, 1x HDMI, 1x DP | 1x DVI, 1x HDMI, 3x DP |

| Release Date | 2019-03-13 | 2016-06-09 |

The Verdict

The data shows two cards with different strengths. The GTX 1070 wins the majority of benchmark comparisons, particularly in DirectX 9, 10, and 11 legacy tests, plus G3D and compute workloads. Its 8 GB memory buffer and 256-bit bus provide 256.3 GB/s of bandwidth, versus 192.1 GB/s for the GTX 1660. The GTX 1070 also offers higher raw pixel and texture rates, at 107.7 GPixel/s and 202.0 GTexel/s respectively.

The GTX 1660 counters with a decisive win in Vulkan and a solid margin in OpenCL. Its 10.05 TFLOPS of FP16 compute versus 101.0 GFLOPS for the GTX 1070 points to better half-precision capability. The GTX 1660 also draws less power, with a 120 W TDP versus 150 W, and is shorter at 229 mm versus 267 mm.

For users prioritizing older DirectX titles and general 3D performance, the GTX 1070 is the clear choice based on benchmark data. For those working with Vulkan-heavy applications or OpenCL compute, the GTX 1660 offers advantages that exceed its lower transistor count. The 3DMark DX12 result is nearly even, indicating that modern games may not show a decisive winner.

Where Each One Wins

The GTX 1070 wins in these scenarios:

• Legacy DirectX workloads: Passmark DX9, DX10, and DX11 all show double-digit leads for the GTX 1070.
• General 3D rendering: Passmark G3D favors the GTX 1070 by 13.7%.
• Compute workloads: Passmark GPU Compute gives the GTX 1070 an 18.7% edge.
• 2D tasks: Passmark G2D scores favor the GTX 1070 by 8.3%.
• Higher resolution or memory-heavy workloads: the 8 GB frame buffer and 256.3 GB/s bandwidth provide a substantial capacity and throughput advantage.

The GTX 1660 wins in these scenarios:

• Vulkan-based applications: the 126.6% lead in Geekbench Vulkan is the largest delta in the entire comparison.
• OpenCL compute: a 7% advantage over the GTX 1070.
• DirectX 12: a narrow 2.1% win in Passmark DX12.
• Power-constrained builds: the 120 W TDP and 300 W suggested PSU make it easier to integrate into smaller systems.
• Compact cases: at 229 mm long, it is 38 mm shorter than the GTX 1070.

The benchmark results indicate that the GTX 1070 remains a strong performer for traditional rasterization and compute, while the GTX 1660 demonstrates clear superiority in modern API efficiency, particularly Vulkan. The choice depends on which workload profile matches the user's priorities, as neither card dominates across all recorded tests.