NVIDIA GeForce GTX 1660 Ti vs NVIDIA GeForce GTX 660 Comparison

The Verdict

The NVIDIA GeForce GTX 660 and the NVIDIA GeForce GTX 1660 Ti represent two very different eras of GPU design, and the benchmark data reflects that. The GTX 660 is a Kepler-era card from the GeForce 600 generation, built on a 28 nm process, while the GTX 1660 Ti is a Turing-based card from the GeForce 16 generation, built on a 12 nm process. Looking strictly at the recorded measurements, the GTX 1660 Ti is the clear winner for modern compute-heavy workloads, but the GTX 660 has a surprising edge in one specific API test.

The GTX 1660 Ti should be the pick for anyone running OpenCL-accelerated applications, as its Geekbench OpenCL score of 51029 absolutely crushes the GTX 660's 11347, a delta of 77.8% in favor of the newer card. This is not a marginal improvement; the newer card is roughly 4.5 times faster in that specific workload. The GTX 660, however, wins the Geekbench Vulkan test with a score of 11415 versus the GTX 1660 Ti's 5827, a 95.9% advantage for the older card. This is a peculiar result, and it suggests that the GTX 660 has better driver optimization or architectural efficiency for that particular Vulkan workload, despite being far older.

For general use, the average benchmark scores tell a more nuanced story. The GTX 660 has an average benchmark score of 9022, which places it at the 45th percentile of all GPUs in the database. The GTX 1660 Ti has an average score of 7723, placing it at the 41st percentile. This is counterintuitive given the OpenCL result, but it is important to note that the benchmark suites are different. The GTX 660's average is pulled up by its strong Vulkan score, while the GTX 1660 Ti's average is dragged down by its weak Vulkan result and includes several Passmark tests with varying scores.

The practical verdict is straightforward: if you are working with OpenCL-based compute, rendering, or physics simulations, the GTX 1660 Ti is the only sensible choice. If you are dealing with a legacy application that relies on Vulkan and cannot be updated, the GTX 660 might be worth considering, but that is a very narrow use case. For anything else, the GTX 1660 Ti's higher memory capacity, faster memory bandwidth, and superior compute throughput make it the better long-term investment, even if its average score is lower due to the Vulkan anomaly.

FAQ

Q: Which card has a higher average benchmark score?

A: The GTX 660 has an average benchmark score of 9022, which is higher than the GTX 1660 Ti's 7723. However, this is influenced by the GTX 660's strong Vulkan performance and the GTX 1660 Ti's weak Vulkan performance, so the average should not be read as overall superiority.

Q: How do the two cards compare in OpenCL performance?

A: The GTX 1660 Ti scores 51029 in Geekbench OpenCL, which is 77.8% higher than the GTX 660's 11347. This is a massive gap, indicating the newer card is far more capable for compute-heavy tasks.

Q: Is the GTX 660 better at any benchmark?

A: Yes, the GTX 660 wins the Geekbench Vulkan test with a score of 11415, which is 95.9% higher than the GTX 1660 Ti's 5827. This is a notable win for the older card.

Q: What is the memory configuration difference?

A: The GTX 660 has 2 GB of GDDR5 memory on a 192-bit bus with a bandwidth of 144.2 GB/s. The GTX 1660 Ti has 6 GB of GDDR6 memory on a 192-bit bus with a bandwidth of 288.0 GB/s. The newer card has three times the capacity and double the bandwidth.

Q: Which card has a higher transistor count?

A: The GTX 1660 Ti has 6,600 million transistors on a 284 mm² die, while the GTX 660 has 2,540 million transistors on a 221 mm² die. The newer card is significantly more complex.

Q: Are the API support levels different?

A: Yes, the GTX 660 supports DirectX 12 (11_0), while the GTX 1660 Ti supports DirectX 12 (12_1). Both support OpenGL 4.6, but the GTX 660 supports Vulkan 1.2.175, while the GTX 1660 Ti supports Vulkan 1.4.

Architecture Differences

The two GPUs are built on fundamentally different architectures. The GTX 660 uses the Kepler architecture with the GK106 chip, while the GTX 1660 Ti uses the Turing architecture with the TU116 chip. This is a generational leap, not just a clock speed bump.

The manufacturing process differs significantly. The GTX 660 is fabricated on a 28 nm process at TSMC, while the GTX 1660 Ti uses a 12 nm process, also at TSMC. The transistor density highlights this difference: the GTX 660 has 11.5 million transistors per square millimeter, while the GTX 1660 Ti has 23.2 million per square millimeter. The newer chip packs more than double the transistors per area.

The raw transistor counts are stark. The GTX 660 has 2,540 million transistors on a 221 mm² die, while the GTX 1660 Ti has 6,600 million transistors on a 284 mm² die. This is a 2.6 times increase in transistor count, enabling the newer card to have more shading units, texture mapping units, and render output units.

The execution resources differ as well. The GTX 660 has 960 shading units, 80 TMUs, and 24 ROPs. The GTX 1660 Ti has 1536 shading units, 96 TMUs, and 48 ROPs. The newer card has 60% more shading units, 20% more TMUs, and double the ROPs. This translates to higher pixel and texture rates: the GTX 660 achieves 20.64 GPixel/s and 82.56 GTexel/s, while the GTX 1660 Ti achieves 84.96 GPixel/s and 169.9 GTexel/s.

The FP32 compute throughput also reflects the architectural gap. The GTX 660 delivers 1.981 TFLOPS, while the GTX 1660 Ti delivers 5.437 TFLOPS, which is 2.7 times higher. The GTX 1660 Ti also has FP16 capability at 10.87 TFLOPS (2:1 rate), which the GTX 660 does not have at all.

Clock speeds are higher on the newer card. The GTX 660 has a base clock of 980 MHz and a boost clock of 1032 MHz. The GTX 1660 Ti has a base clock of 1500 MHz and a boost clock of 1770 MHz. These higher clocks, combined with the architectural improvements, are the primary drivers of the performance difference.

Specification Differences

The two cards differ in almost every measurable specification except for a few shared traits. Both use a PCIe 3.0 x16 bus interface, both are dual-slot cards, and both have a suggested power supply of 300 W. The GTX 660 has a TDP of 140 W, while the GTX 1660 Ti has a lower TDP of 120 W, which is notable given the newer card's higher performance.

The memory subsystems are completely different. The GTX 660 has 2 GB of GDDR5 memory running at 1502 MHz (6 Gbps effective) with a 192-bit bus and 144.2 GB/s bandwidth. The GTX 1660 Ti has 6 GB of GDDR6 memory running at 1500 MHz (12 Gbps effective) with a 192-bit bus and 288.0 GB/s bandwidth. The newer card has three times the capacity and exactly double the bandwidth.

The power connectors differ. The GTX 660 uses a single 6-pin connector, while the GTX 1660 Ti uses a single 8-pin connector. This is a practical consideration for older power supplies.

The display outputs are different. The GTX 660 has 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2. The GTX 1660 Ti has 1x DVI, 1x HDMI 2.0, and 1x DisplayPort 1.4a. The newer card supports higher display standards.

Physical dimensions vary. The GTX 660 is 241 mm (9.5 inches) long. The GTX 1660 Ti is 229 mm (9 inches) long, 111 mm (4.4 inches) tall, and 35 mm (1.4 inches) wide. The newer card is shorter but has more defined dimensions.

The release dates are far apart. The GTX 660 was released on 2012-09-05, while the GTX 1660 Ti was released on 2019-02-21. The GTX 660 has a launch MSRP of 229 USD, and the GTX 1660 Ti has a launch MSRP of 279 USD.

Head-to-Head Benchmarks

The head-to-head benchmark data includes only two tests, and the results are polarizing. In the Geekbench OpenCL test, the GTX 1660 Ti scores 51029, while the GTX 660 scores 11347. The newer card wins by a delta of 77.8%. This is the most significant performance gap in the comparison, and it underscores the newer card's dominance in compute-heavy tasks.

In the Geekbench Vulkan test, the results are reversed. The GTX 660 scores 11415, while the GTX 1660 Ti scores 5827. The older card wins by a delta of 95.9%. This is a bizarre outcome, as the GTX 660 is not only older but also has significantly lower raw specifications. The Vulkan result suggests that either the GTX 660's drivers are exceptionally well-optimized for this specific workload, or the GTX 1660 Ti's Vulkan implementation is lacking in this particular benchmark scenario.

The wins are split evenly: one win for the GTX 660 and one win for the GTX 1660 Ti. However, the magnitude of the wins is not equal. The GTX 1660 Ti's OpenCL win is a 77.8% advantage, while the GTX 660's Vulkan win is a 95.9% advantage. In absolute score differences, the GTX 1660 Ti is ahead by 39682 points in OpenCL, while the GTX 660 is ahead by 5588 points in Vulkan. The newer card's win is far more impactful in raw score terms.

The percentile rankings reflect the overall database context. The GTX 660 sits at the 45th percentile, just above its nearest rival, the NVIDIA TITAN V CEO Edition, which has an average score of 9037 and a delta of -0.2%. The GTX 1660 Ti sits at the 41st percentile, just above the AMD Radeon 540, which has an average score of 7673 and a delta of 0.7%. The GTX 660's average score of 9022 is higher than the GTX 1660 Ti's 7723, but this is driven by the Vulkan result, which is not representative of typical gaming or compute workloads.

Where Each One Wins

The GTX 1660 Ti is the clear winner for any OpenCL-based application. This includes many scientific computing tools, machine learning frameworks, and rendering engines that leverage OpenCL for general-purpose GPU computing. The 77.8% lead in Geekbench OpenCL is decisive, and the newer card's 5.437 TFLOPS of FP32 compute and 10.87 TFLOPS of FP16 compute make it a far more capable compute device. Its 6 GB of GDDR6 memory with 288.0 GB/s bandwidth also allows it to handle larger datasets without running out of memory.

The GTX 1660 Ti also wins on raw throughput metrics. Its pixel rate of 84.96 GPixel/s is four times higher than the GTX 660's 20.64 GPixel/s, and its texture rate of 169.9 GTexel/s is more than double the GTX 660's 82.56 GTexel/s. For modern games and applications that rely on high fill rates, the newer card is vastly superior.

The GTX 660 wins in the specific case of the Geekbench Vulkan benchmark, where it scores 95.9% higher than the GTX 1660 Ti. This is a narrow but real victory, and it suggests that for legacy Vulkan applications or workloads that are sensitive to the GTX 660's specific driver implementation, the older card can hold its own. The GTX 660 also has a higher average benchmark score (9022 vs 7723) and a higher percentile ranking (45th vs 41st), but this is skewed by the Vulkan anomaly.

For gaming, the data does not include a direct gaming benchmark, but the specification differences point to the GTX 1660 Ti being the better choice. Its higher shading unit count, double ROPs, and faster memory bandwidth all contribute to better frame rates in modern titles. The GTX 660's only real advantage is its Vulkan score, which is unlikely to translate to a broad gaming advantage.

In summary, the GTX 1660 Ti wins for compute, modern API workloads, and overall throughput. The GTX 660 wins in the narrow case of the Vulkan benchmark and has a higher average score in the database, but this is not enough to recommend it for any practical modern use case. The data shows a generational shift: the GTX 1660 Ti is the card to pick for almost any workload, with the GTX 660 serving as a curiosity for Vulkan-specific legacy scenarios.