NVIDIA GeForce GTX 660 vs NVIDIA GeForce GTX 980 Comparison

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA GeForce GTX 660 has a higher average benchmark score of 9022, compared to 8167 for the NVIDIA GeForce GTX 980. However, the GTX 980 wins every head-to-head benchmark test recorded in the database.

Q: How large is the performance gap in the Geekbench Metal test?

A: The GTX 980 scores 15163 in Geekbench Metal, while the GTX 660 scores 4305. That is a 71.6% advantage for the GTX 980.

Q: What is the transistor count difference between the two cards?

A: The GTX 660 uses the GK106 chip with 2,540 million transistors, while the GTX 980 uses the GM204 chip with 5,200 million transistors. The GTX 980 has roughly double the transistor count.

Q: Do both cards use the same manufacturing process?

A: Yes, both are built on a 28 nm process at TSMC. The GTX 660 is based on the Kepler architecture, while the GTX 980 uses Maxwell 2.0.

Q: Which card has more memory and bandwidth?

A: The GTX 980 has 4 GB of GDDR5 memory on a 256 bit bus with 224.4 GB/s bandwidth. The GTX 660 has 2 GB of GDDR5 memory on a 192 bit bus with 144.2 GB/s bandwidth.

Q: What is the DirectX support difference?

A: The GTX 660 supports DirectX 12 (11_0), while the GTX 980 supports DirectX 12 (12_1). The GTX 980 also supports Vulkan 1.4 compared to Vulkan 1.2.175 on the GTX 660.

Architecture Differences

The two cards come from different NVIDIA generations and architectures. The GTX 660 is a Kepler architecture part based on the GK106 chip, while the GTX 980 is a Maxwell 2.0 architecture part based on the GM204 chip. Both are fabricated by TSMC on a 28 nm process, but the GTX 980 packs far more hardware: 5,200 million transistors across a 398 mm² die, compared to 2,540 million transistors across a 221 mm² die for the GTX 660.

The compute resources differ dramatically. The GTX 660 has 960 shading units, 80 texture mapping units, and 24 ROPs. The GTX 980 has 2048 shading units, 128 TMUs, and 64 ROPs. That is more than double the shading units and ROPs, and 60% more TMUs. The transistor density also improves modestly, from 11.5M per mm² on the GTX 660 to 13.1M per mm² on the GTX 980.

Clock speeds are higher on the GTX 980 as well. The GTX 660 runs at a base clock of 980 MHz with a boost of 1032 MHz, while the GTX 980 runs at 1127 MHz base and 1216 MHz boost. Memory clocks also differ: the GTX 660 uses 1502 MHz memory with 6 Gbps effective, and the GTX 980 uses 1753 MHz memory with 7 Gbps effective.

Feature support also diverges. The GTX 660 supports DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. The GTX 980 supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The GTX 980 also has a newer display output set, including HDMI 2.0 and three DisplayPort 1.2 outputs, whereas the GTX 660 has HDMI 1.4a and a single DisplayPort 1.2.

Neither card has ray tracing cores or tensor cores. Both are dual-slot cards with PCIe 3.0 x16 interfaces, but the GTX 980 requires two 6-pin power connectors and a 450 W suggested PSU, while the GTX 660 requires one 6-pin connector and a 300 W suggested PSU. The GTX 980 is also physically larger at 267 mm (10.5 inches) long, 111 mm (4.4 inches) high, and 40 mm (1.6 inches) wide; the GTX 660 is 241 mm (9.5 inches) long with no recorded height or width.

Head-to-Head Benchmarks

The database records three head-to-head benchmark tests between these two cards, and the GTX 980 wins all three. The most lopsided result is in Geekbench Metal, where the GTX 980 scores 15163 against 4305 for the GTX 660. That is a 71.6% deficit for the older card, and it reflects the massive gap in raw compute resources and architecture efficiency.

In Geekbench OpenCL, the GTX 980 scores 34676 versus 11347 for the GTX 660, a 67.3% advantage. This test stresses general-purpose compute throughput, and the GTX 980's 4.981 TFLOPS of FP32 performance dwarfs the GTX 660's 1.981 TFLOPS. The GTX 980 also has more than double the texture rate (155.6 GTexel/s versus 82.56 GTexel/s) and nearly four times the pixel rate (77.82 GPixel/s versus 20.64 GPixel/s).

The closest head-to-head result is Geekbench Vulkan, where the GTX 980 scores 22543 versus 11415 for the GTX 660. The delta is 49.4%, still a decisive win, but the gap narrows compared to Metal and OpenCL. In absolute terms, the GTX 980 delivers roughly twice the Vulkan performance of the GTX 660.

There are no benchmark tests in the database where the GTX 660 wins. The head-to-head record is 3 wins for the GTX 980 and 0 for the GTX 660. The GTX 980 also has a much wider benchmark portfolio in the database, including Passmark DirectX 9, 10, 11, and 12 tests, Passmark G2D, G3D, and GPU compute tests, and 3DMark Steel Nomad DX12, while the GTX 660 only has Geekbench entries.

It is notably the GTX 660 has a higher average benchmark score (9022) than the GTX 980 (8167), but this is a function of the test mix. The GTX 660's average includes only Geekbench Metal, OpenCL, and Vulkan scores, all of which are relatively high for its class. The GTX 980's average includes several Passmark tests with lower absolute scores, which pulls its average down despite its clear performance superiority.

The Verdict

The data is unambiguous: the NVIDIA GeForce GTX 980 is the far stronger GPU in every recorded head-to-head benchmark. It wins Geekbench Metal by 71.6%, Geekbench OpenCL by 67.3%, and Geekbench Vulkan by 49.4%. The GTX 980 has more than double the shading units, more than double the ROPs, double the memory capacity, and 55.6% more memory bandwidth than the GTX 660.

The GTX 980 should be the choice for anyone who needs maximum compute throughput, higher resolution textures (4 GB versus 2 GB), or newer API support like DirectX 12_1 and Vulkan 1.4. It also has a higher pixel rate and texture rate, making it better suited to demanding rendering workloads.

The GTX 660, despite losing every head-to-head test, is not without merit. It has a lower TDP of 140 W versus 165 W, requires only a single 6-pin power connector and a 300 W suggested PSU versus 450 W, and is shorter at 241 mm versus 267 mm. Its average benchmark score of 9022 is higher than the GTX 980's 8167, and it sits at the 45th percentile of all GPUs in the database, compared to the 43rd percentile for the GTX 980. For users constrained by power supply capacity or physical space, the GTX 660 is the more accommodating card.

However, the GTX 660 is end-of-life hardware from 2012, while the GTX 980 is also end-of-life but from 2014. The GTX 980's successor is the GeForce 10 series, while the GTX 660's successor is the GeForce 700 series. In terms of pure recorded performance, the GTX 980 is the clear winner.

Specification Differences

| Specification | NVIDIA GeForce GTX 660 | NVIDIA GeForce GTX 980 |

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

| Chip | GK106 | GM204 |

| Architecture | Kepler | Maxwell 2.0 |

| Generation | GeForce 600 | GeForce 900 |

| Process Node | 28 nm | 28 nm |

| Transistors | 2,540 million | 5,200 million |

| Die Size | 221 mm² | 398 mm² |

| Transistor Density | 11.5M / mm² | 13.1M / mm² |

| Base Clock | 980 MHz | 1127 MHz |

| Boost Clock | 1032 MHz | 1216 MHz |

| Memory Clock | 1502 MHz (6 Gbps effective) | 1753 MHz (7 Gbps effective) |

| Memory Size | 2 GB | 4 GB |

| Memory Bus Width | 192 bit | 256 bit |

| Memory Bandwidth | 144.2 GB/s | 224.4 GB/s |

| Shading Units | 960 | 2048 |

| TMUs | 80 | 128 |

| ROPs | 24 | 64 |

| Pixel Rate | 20.64 GPixel/s | 77.82 GPixel/s |

| Texture Rate | 82.56 GTexel/s | 155.6 GTexel/s |

| FP32 Performance | 1.981 TFLOPS | 4.981 TFLOPS |

| TDP | 140 W | 165 W |

| Power Connectors | 1x 6-pin | 2x 6-pin |

| Suggested PSU | 300 W | 450 W |

| Display Outputs | 2x DVI, 1x HDMI 1.4a, 1x DisplayPort 1.2 | 1x DVI, 1x HDMI 2.0, 3x DisplayPort 1.2 |

| DirectX Support | 12 (11_0) | 12 (12_1) |

| Vulkan Support | 1.2.175 | 1.4 |

| Length | 241 mm (9.5 inches) | 267 mm (10.5 inches) |

| Height | Not recorded | 111 mm (4.4 inches) |

| Width | Not recorded | 40 mm (1.6 inches) |

| Launch MSRP | 229 USD | 549 USD |

| Release Date | 2012-09-05 | 2014-09-18 |

Where Each One Wins

The GTX 980 wins on raw performance across every benchmark category recorded in the head-to-head data. It is the choice for compute-heavy workloads, high-resolution texture rendering, and modern API features. Its 4 GB memory capacity and 224.4 GB/s bandwidth give it a clear edge in memory-intensive scenarios. The GTX 980 also has superior pixel and texture throughput, making it better for high-detail gaming and rendering tasks.

The GTX 660 wins on efficiency and system integration. It draws 140 W versus 165 W, needs only one 6-pin power connector instead of two, and requires a 300 W PSU instead of 450 W. It is also shorter, which matters in compact cases. The GTX 660's higher average benchmark score (9022 versus 8167) and higher percentile ranking (45th versus 43rd) in the database suggest that, in the specific tests where both have data, the GTX 660 holds its own in the broader GPU landscape. Still, in any direct comparison, the GTX 980 is the definitive performance winner, with the GTX 660 serving as the lower-power, shorter-card alternative.