NVIDIA GeForce GTX 1080 vs NVIDIA GeForce GTX 1660 SUPER Comparison

The GeForce GTX 1660 SUPER and the GeForce GTX 1080 represent two distinct philosophies from NVIDIA, separated by a generation but both anchored in the GeForce 10-series lineage. The data shows a clear split: the older GTX 1080 dominates in raw rasterization and compute, while the newer GTX 1660 SUPER posts surprising wins in modern API workloads like Vulkan. The benchmark results indicate that architectural age matters less than specific workload optimization, making this a nuanced comparison rather than a simple generational upgrade.

Head-to-Head Benchmarks

The GTX 1080 takes the overall win count, securing 8 of the 10 head-to-head tests. Its largest margin comes in Passmark GPU Compute, where it scores 7614 against the 1660 SUPER's 5076, a 33.3% advantage. This is not a small gap; it reflects a fundamental difference in compute throughput that shows up in any workload leveraging raw shader processing. The GTX 1080 also leads decisively in Passmark DirectX 10 (93 vs 65, a 30.1% lead) and Passmark G3D (15586 vs 12699, an 18.5% lead), the latter being a strong proxy for overall gaming performance in legacy titles.

The 3DMark Steel Nomad DX12 test reinforces the GTX 1080's lead, with a score of 1560 versus 1278, an 18.1% difference. This is a forward-looking benchmark, yet the older card still wins by a wide margin, suggesting its larger memory bus and higher pixel throughput matter even in modern DirectX 12 workloads. The GTX 1080 also edges ahead in Passmark DirectX 11 (124 vs 104, a 16.1% lead), DirectX 9 (211 vs 189, a 10.4% lead), DirectX 12 (55 vs 50, a 9.1% lead), and the 2D test (888 vs 805, a 9.3% lead).

The GTX 1660 SUPER's two wins are striking, however. In Geekbench Vulkan, it achieves 57102 against the GTX 1080's 30398, a massive 87.8% advantage. This is not a marginal victory; it is a near-doubling of performance, indicating that the Turing architecture's Vulkan driver implementation is vastly superior to Pascal's. The 1660 SUPER also wins Geekbench OpenCL, scoring 52490 versus 51204, a 2.5% lead, showing that in compute API abstraction layers, the newer card can hold its own.

The Verdict

From the data, the GTX 1080 is the superior card for pure gaming and compute performance. Its wins in Passmark G3D, DirectX 10/11/12, and GPU Compute are consistent and often by double-digit percentages. The 33.3% lead in GPU Compute makes it the clear choice for any workload that is not explicitly optimized for Vulkan. The GTX 1080's 18.1% lead in the 3DMark Steel Nomad test is particularly telling, as this is a demanding DX12 benchmark that should favor newer architectures.

The GTX 1660 SUPER is the better choice for Vulkan-focused applications, where its 87.8% lead is impossible to ignore. The data also shows it has a slight edge in OpenCL, making it a viable option for compute tasks that utilize that API. The 1660 SUPER's average benchmark score of 12986 is also higher than the GTX 1080's 11960, although this is skewed by the Vulkan outlier. The GTX 1080's percentile rank of 51 versus the 1660 SUPER's 53 suggests they sit at similar overall performance tiers, but the distribution of strengths is very different.

Architecture Differences

The two GPUs are built on different architectures and process nodes. The GTX 1660 SUPER uses the TU116 chip on the Turing architecture, fabricated on a 12 nm process at TSMC. The GTX 1080 uses the GP104 chip on the older Pascal architecture, fabricated on a 16 nm process, also at TSMC. This process difference is significant, but the transistor counts tell a more complex story: the GTX 1080 has 7,200 million transistors on a 314 mm² die, while the 1660 SUPER has 6,600 million on a smaller 284 mm² die. The transistor density is nearly identical (23.2M/mm² for the 1660 SUPER versus 22.9M/mm² for the 1080), meaning the architectural efficiency gains in Turing are not from density but from design changes.

The GTX 1080's Pascal architecture uses a 1:64 FP16 ratio, as indicated by its 138.6 GFLOPS FP16 performance, while the 1660 SUPER's Turing architecture offers a 2:1 FP16 ratio, delivering 10.05 TFLOPS. This makes the 1660 SUPER vastly more capable in FP16 compute, a feature that explains its Vulkan and OpenCL advantages. The GTX 1080 compensates with a much higher FP32 throughput, delivering 8.873 TFLOPS versus the 1660 SUPER's 5.027 TFLOPS, which directly translates to its compute and DirectX wins.

Specification Differences

The specification sheets reveal several key differences beyond architecture. The GTX 1080 has a larger memory configuration: 8 GB of GDDR5X on a 256-bit bus, versus 6 GB of GDDR6 on a 192-bit bus for the 1660 SUPER. Interestingly, the 1660 SUPER has higher memory bandwidth at 336.0 GB/s against the 320.3 GB/s of the GTX 1080, due to the faster 14 Gbps effective GDDR6 speed versus 10 Gbps effective GDDR5X. The GTX 1080 has more shading units (2560 vs 1408), more texture mapping units (160 vs 88), and more ROPs (64 vs 48), which explains its higher pixel rate of 110.9 GPixel/s versus 85.68 GPixel/s.

The GTX 1080 also has a higher base clock (1607 MHz vs 1530 MHz) and a slightly lower boost clock (1733 MHz vs 1785 MHz). Its texture rate of 277.3 GTexel/s far exceeds the 157.1 GTexel/s of the 1660 SUPER. The power draw is also different: the GTX 1080 has a 180 W TDP with a 450 W suggested PSU, while the 1660 SUPER has a 125 W TDP with a 300 W suggested PSU. Physically, the GTX 1080 is longer (267 mm vs 229 mm) and wider (40 mm vs 35 mm), but both are dual-slot cards with a single 8-pin power connector. The GTX 1080 offers an additional DisplayPort output (3x DisplayPort 1.4a vs 1x on the 1660 SUPER).

FAQ

Q: Which card is faster in the Passmark G3D benchmark?

A: The NVIDIA GeForce GTX 1080 scores 15586, while the GTX 1660 SUPER scores 12699, giving the GTX 1080 an 18.5% lead in this overall 3D graphics test.

Q: How significant is the Vulkan performance gap?

A: The GTX 1660 SUPER leads by 87.8% in Geekbench Vulkan, scoring 57102 versus the GTX 1080's 30398, indicating a massive architectural advantage for Turing in this API.

Q: Does the GTX 1080 have more memory bandwidth?

A: No. The GTX 1660 SUPER has 336.0 GB/s of bandwidth, while the GTX 1080 has 320.3 GB/s, despite the 1080 having a wider 256-bit bus. The GDDR6 memory on the 1660 SUPER operates at a higher effective speed of 14 Gbps versus 10 Gbps.

Q: Which card has a higher transistor count?

A: The GTX 1080 has 7,200 million transistors, compared to 6,600 million on the GTX 1660 SUPER. The 1080 also has a larger die at 314 mm² versus 284 mm².

Q: What is the difference in FP32 compute performance?

A: The GTX 1080 delivers 8.873 TFLOPS of FP32 performance, which is significantly higher than the 5.027 TFLOPS of the GTX 1660 SUPER, explaining its lead in compute-heavy benchmarks.

Q: Which card is better for DirectX 12 gaming?

A: The GTX 1080 wins in the Passmark DirectX 12 test with a score of 55 versus 50, a 9.1% lead. It also wins the 3DMark Steel Nomad DX12 test by 18.1%.

Where Each One Wins

The GTX 1080 is the clear winner for traditional rasterization workloads. Its victories span the entire DirectX spectrum, from DirectX 9 through DirectX 12, with margins ranging from 9.1% to 30.1%. The 3DMark Steel Nomad DX12 result shows this lead persists in modern benchmark suites, making it the safer choice for a broad library of games. Its 33.3% lead in Passmark GPU Compute makes it the preferred option for GPGPU tasks like video encoding or physics simulations, provided they rely on FP32 math. The GTX 1080's higher pixel rate (110.9 GPixel/s) and texture rate (277.3 GTexel/s) give it a fundamental throughput advantage that is hard to overcome.

The GTX 1660 SUPER wins are concentrated in two areas: Vulkan and OpenCL. The 87.8% Vulkan lead is a generational leap in driver and hardware support, making it the go-to card for Vulkan-native titles or emulators that heavily leverage this API. Its 2.5% OpenCL win suggests it is also capable in OpenCL compute workloads, despite having fewer shading units. The 1660 SUPER's lower TDP of 125 W versus 180 W means it achieves these results with a significantly lower power draw, and its smaller physical footprint (229 mm vs 267 mm) makes it easier to fit in compact cases. For a user primarily playing Vulkan-based games or running FP16 compute tasks, the 1660 SUPER is the data-backed pick. For everything else, the GTX 1080's specifications and benchmark wins make it the dominant performer.