NVIDIA GeForce GTX 1660 vs NVIDIA GeForce GTX 680 Comparison

FAQ

Q: Which GPU has the higher average benchmark score in the database?

A: The NVIDIA GeForce GTX 680 has a higher average benchmark score of 14,150 compared to the GTX 1660's 11,680. The GTX 680 also holds a higher percentile ranking at 55 versus the GTX 1660's 51.

Q: What are the two direct head-to-head benchmark results?

A: The database records two head-to-head tests: Geekbench OpenCL and Geekbench Vulkan. The GTX 1660 wins both, scoring 47,850 in OpenCL and 50,137 in Vulkan, while the GTX 680 scores 16,906 and 17,646 respectively.

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

A: The GTX 1660 leads by 64.7% in Geekbench OpenCL and 64.8% in Geekbench Vulkan. These deltas are nearly identical, indicating a consistent advantage across both compute APIs.

Q: Which card has a higher transistor count?

A: The GTX 1660's TU116 chip contains 6,600 million transistors, nearly double the GTX 680's 3,540 million. However, the GTX 680's die size is slightly larger at 294 mm² versus 284 mm² for the GTX 1660.

Q: What is the memory configuration difference?

A: The GTX 680 has 2 GB of GDDR5 on a 256-bit bus with 192.3 GB/s bandwidth. The GTX 1660 has 6 GB of GDDR5 on a 192-bit bus with 192.1 GB/s bandwidth, making the bandwidth nearly identical despite different bus widths.

Q: Which card supports newer DirectX features?

A: The GTX 1660 supports DirectX 12 (12_1), while the GTX 680 only supports DirectX 12 (11_0). Both cards support OpenGL 4.6, but the GTX 1660 has a newer Vulkan version at 1.4 versus 1.2.175.

Architecture Differences

The GTX 680 is built on the Kepler architecture with the GK104 chip, manufactured on a 28 nm process at TSMC. The GTX 1660 uses the Turing architecture with the TU116 chip, built on a 12 nm process, also at TSMC. This process change is significant: the GTX 1660 packs 6,600 million transistors into a 284 mm² die, achieving a transistor density of 23.2M per mm², while the GTX 680 has 3,540 million transistors across a 294 mm² die at 12.0M per mm².

The shading unit counts differ, with the GTX 680 featuring 1,536 shading units, 128 texture mapping units, and 32 ROPs. The GTX 1660 has fewer shading units at 1,408, fewer TMUs at 88, but more ROPs at 48. This configuration suggests the GTX 1660 prioritizes pixel throughput over texture work. The pixel rate confirms this: the GTX 1660 delivers 85.68 GPixel/s versus 33.86 GPixel/s for the GTX 680, a 2.5x advantage in raw pixel fill.

Clock speeds show a generational jump. The GTX 680 runs at 1006 MHz base and 1058 MHz boost, while the GTX 1660 operates at 1530 MHz base and 1785 MHz boost. The memory clocks also differ: the GTX 680's memory runs at 1502 MHz (6 Gbps effective), while the GTX 1660's memory runs at 2001 MHz (8 Gbps effective). However, the wider 256-bit bus on the GTX 680 compensates for its lower memory clock, resulting in nearly identical bandwidth: 192.3 GB/s versus 192.1 GB/s.

The FP32 compute output is substantially higher on the GTX 1660 at 5.027 TFLOPS, compared to 3.250 TFLOPS on the GTX 680. The GTX 1660 also supports FP16 computation at 10.05 TFLOPS with a 2:1 ratio, while the GTX 680 has no FP16 capability listed. Both cards lack ray tracing and tensor cores. The GTX 680 uses two 6-pin power connectors with a 195 W TDP and 450 W suggested PSU, while the GTX 1660 uses a single 8-pin connector with a 120 W TDP and 300 W suggested PSU.

Where Each One Wins

The benchmark data shows the GTX 1660 winning both head-to-head compute tests decisively, but the broader database picture is more nuanced. The GTX 680's average benchmark score of 14,150 places it in a higher percentile (55) than the GTX 1660's 11,680 (percentile 51). This indicates that across the full range of recorded tests, the GTX 680 performs better relative to the entire GPU landscape.

The GTX 680's nearest rivals include the NVIDIA Tesla K10 (only 0.9% behind), the GeForce GTX 1070 Ti (0.9% ahead), and the Intel Iris Xe MAX Graphics (1.2% ahead). This clustering shows the GTX 680 sits in a competitive band where small score differences separate cards. The GTX 1660's nearest rivals include the AMD Radeon RX 7800 XT (0.5% behind), the Radeon Pro 5500M (1.3% ahead), and the Radeon RX 6500 XT (1.4% behind), placing it in a similar tight grouping.

For compute-heavy workloads like OpenCL and Vulkan, the GTX 1660 is the clear choice, with roughly 2.8x higher scores in both APIs. However, for overall benchmark performance as measured by the database's average, the GTX 680 holds an advantage. The GTX 680 also offers a higher texture rate at 135.4 GTexel/s versus 157.1 GTexel/s for the GTX 1660, though the GTX 1660's pixel rate is far superior. Users prioritizing pixel fill and compute throughput should choose the GTX 1660, while those looking at the aggregate benchmark profile may find the GTX 680 more appealing.

Specification Differences

The two cards differ across nearly every major specification category. The GTX 680 uses the Kepler architecture on a 28 nm process, while the GTX 1660 uses Turing on 12 nm. Transistor counts are 3,540 million versus 6,600 million, with die sizes of 294 mm² and 284 mm² respectively. The GTX 680 has more shading units (1,536 vs 1,408) and more TMUs (128 vs 88), but the GTX 1660 has more ROPs (48 vs 32).

Clock speeds favor the GTX 1660: base clocks are 1530 MHz versus 1006 MHz, and boost clocks are 1785 MHz versus 1058 MHz. Memory size differs significantly, with the GTX 680 offering 2 GB and the GTX 1660 offering 6 GB. Both use GDDR5, but the bus widths differ: 256-bit for the GTX 680, 192-bit for the GTX 1660. Bandwidth is nearly equal at 192.3 GB/s and 192.1 GB/s.

The FP32 compute output is 3.250 TFLOPS for the GTX 680 and 5.027 TFLOPS for the GTX 1660. The GTX 1660 also has FP16 capability at 10.05 TFLOPS, which the GTX 680 lacks. Pixel rates are 33.86 GPixel/s versus 85.68 GPixel/s, and texture rates are 135.4 GTexel/s versus 157.1 GTexel/s. Power requirements differ substantially: the GTX 680 draws 195 W TDP with two 6-pin connectors and a 450 W suggested PSU, while the GTX 1660 draws 120 W with one 8-pin connector and a 300 W suggested PSU.

The display outputs also differ. The GTX 680 offers 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2. The GTX 1660 offers 1x DVI, 1x HDMI 2.0, and 1x DisplayPort 1.4a. The GTX 680 is longer at 256 mm (10.1 inches) and wider at 38 mm (1.5 inches), while the GTX 1660 measures 229 mm (9 inches) and 35 mm (1.4 inches). Both are dual-slot cards with 111 mm height. The GTX 680 supports DirectX 12 (11_0) and Vulkan 1.2.175, while the GTX 1660 supports DirectX 12 (12_1) and Vulkan 1.4. Both use PCIe 3.0 x16 and support OpenGL 4.6.

Head-to-Head Benchmarks

The database records two direct head-to-head comparisons between these cards, both in compute-oriented workloads. In Geekbench OpenCL, the GTX 1660 scores 47,850 against the GTX 680's 16,906, a delta of 64.7% in favor of the GTX 1660. In Geekbench Vulkan, the GTX 1660 scores 50,137 versus 17,646, a delta of 64.8%. These results show the GTX 1660 delivers roughly 2.8 times the compute performance in both APIs, and the consistency across OpenCL and Vulkan suggests the advantage is architectural rather than workload-specific.

The GTX 680's individual benchmark scores are 7,897 in Geekbench Metal, 16,906 in OpenCL, and 17,646 in Vulkan. The GTX 1660 has a broader benchmark suite in the database, including 3DMark Steel Nomad DX12 at 1,065, PassMark G3D at 11,646, and PassMark GPU Compute at 4,963. The GTX 1660 also records PassMark scores for DirectX 9 (177), DirectX 10 (61), DirectX 11 (79), DirectX 12 (49), and G2D (776).

While the head-to-head results are lopsided, the average benchmark scores tell a different story. The GTX 680's average of 14,150 exceeds the GTX 1660's 11,680, and the GTX 680's percentile ranking of 55 is higher than the GTX 1660's 51. The GTX 1660's nearest rivals include the AMD Radeon RX 7800 XT, which scores within 0.5%, and the Tesla K20c, which is 1.8% behind. The GTX 680's rivals include the GTX 1070 Ti, which is only 0.9% ahead, and the Radeon Vega 11, which is 1.4% ahead.

The Verdict

The data supports a clear split decision. For compute workloads measured by OpenCL and Vulkan, the GTX 1660 is the dominant choice, outperforming the GTX 680 by nearly 65% in both APIs. Its higher FP32 throughput (5.027 TFLOPS versus 3.250 TFLOPS), higher pixel rate (85.68 GPixel/s versus 33.86 GPixel/s), and larger 6 GB memory capacity make it the stronger card for modern compute tasks and higher resolution textures. The GTX 1660 also runs cooler and more efficiently, with a 120 W TDP versus 195 W, and requires only a 300 W PSU versus 450 W.

However, the database's average benchmark score favors the GTX 680 at 14,150 versus 11,680, and its percentile rank of 55 exceeds the GTX 1660's 51. The GTX 680 also has a higher texture rate in some configurations, though the GTX 1660's texture rate of 157.1 GTexel/s actually exceeds the GTX 680's 135.4 GTexel/s. The GTX 680's nearest rival cluster includes the GTX 1070 Ti, a much newer card, which suggests the GTX 680's aggregate performance profile remains competitive in the database's broader test suite.

The GTX 1660 is the pick for users prioritizing compute performance, modern API support (DirectX 12_1, Vulkan 1.4), and efficiency. The GTX 680 retains appeal for those looking at the overall benchmark aggregate, where it scores higher and ranks better. Both cards are end-of-life products, but the GTX 1660's newer architecture and superior compute results make it the more future-proof option in head-to-head testing. The GTX 680's advantage in average score is real but narrower, and it comes with higher power demands and less memory. Based strictly on the recorded data, the GTX 1660 wins the direct comparisons, while the GTX 680 wins the aggregate average.