NVIDIA GeForce GTX 1660 Ti

The NVIDIA GeForce GTX 1660 Ti, built on the Turing architecture and fabricated on TSMC's 12 nm process, occupies a specific niche in the benchmark database. With a 6,600 million transistor count on a 284 mm² die, this card was positioned as an end-of-life product, yet its data reveals a consistent performance profile. Its average benchmark score of 13,925 places it at the 55th percentile of all GPUs tracked, indicating a solidly mid-range standing. The following analysis breaks down the card's specifications and benchmark results, interpreting what the numbers mean for a user's real-world experience.

Memory Subsystem

The GTX 1660 Ti is equipped with 6 GB of GDDR6 memory, a configuration that directly influences its behavior at higher resolutions. The memory operates across a 192-bit bus interface, yielding a total bandwidth of 288.0 GB/s. This is a substantial pipeline for the card's class, and the data suggests it is well-matched to the GPU's compute capabilities. The effective memory speed of 12 Gbps ensures that the shading units are not starved for data in most scenarios.

For high-resolution gaming, the 6 GB capacity is the primary constraint. While the 288.0 GB/s bandwidth provides enough throughput for 1080p and 1440p workloads, the absolute memory size can become a limiting factor in texture-heavy titles at maximum settings. The architecture's 1536 shading units, 96 texture mapping units, and 48 ROPs are paired with this memory configuration to deliver a balanced pipeline. The pixel rate of 84.96 GPixel/s and texture rate of 169.9 GTexel/s are direct consequences of the core clock speeds and the memory subsystem's ability to feed the render back-end. At 4K, the bandwidth is adequate for maintaining frame pacing, but the 6 GB frame buffer often forces a reduction in texture quality to avoid stuttering, making this a card best suited for 1080p and entry-level 1440p gaming.

Ray Tracing and Feature Set

The GTX 1660 Ti is notably distinct from its higher-tier Turing siblings in one critical aspect: it lacks dedicated ray tracing and tensor cores. The fact pack lists both `rtCores` and `tensorCores` as null, meaning the hardware acceleration for real-time ray tracing and AI-driven features like DLSS is absent. This is a defining characteristic of the GeForce 16 series, which prioritized traditional rasterization performance over the new feature set introduced with the GeForce 20 series.

Despite this omission, the card supports a modern API feature set. It is compatible with DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. This ensures broad compatibility with current game engines and allows developers to leverage the latest graphics APIs for improved draw call efficiency and multi-threaded rendering. The absence of tensor cores means that any compute workloads relying on those specific instructions will not see hardware acceleration, but the general-purpose FP32 performance of 5.437 TFLOPS remains available for standard shader work. The FP16 performance of 10.87 TFLOPS (2:1) indicates that half-precision operations can be processed at a higher rate, though this is less relevant for gaming and more for specific compute tasks. In essence, this is a rasterization-focused card that forgoes the novel features of the Turing generation to deliver performance at a specific price point.

Power and Cooling

The GTX 1660 Ti presents a modest power profile, with a thermal design power (TDP) of 120 W. This figure is a key indicator of the card's efficiency and cooling requirements. The suggested power supply for a system housing this GPU is 300 W, which is a low threshold that allows for integration into a wide range of pre-built systems or upgrades without requiring a new PSU. The power delivery is handled by a single 8-pin connector, a standard and widely compatible interface.

The physical design is a dual-slot solution, measuring 229 mm in length, 111 mm in height, and 35 mm in width. This makes it a relatively compact card that should fit into most mid-tower cases without clearance issues. The 120 W TDP means that a capable air cooler with a dual-slot design is sufficient to manage thermals under load. The low power draw also contributes to lower overall system heat output, which can be a benefit in smaller form factor builds. The combination of a 300 W PSU recommendation and a single 8-pin connector simplifies installation, making this a straightforward component for users building a system with a moderate power supply.

How It Compares

The GTX 1660 Ti's position in the market is best understood by its proximity to a cluster of rival GPUs, all of which have average benchmark scores within a narrow band. The data shows a tight grouping, with performance differences of less than one percent separating these cards.

The closest rival is the AMD Radeon RX 570X, which posts an average score of 13,871. The GTX 1660 Ti is a mere 0.4% ahead, placing these two cards in a statistical dead heat. Users can expect interchangeable performance between these two in a majority of workloads.

Another direct competitor is the NVIDIA Tesla K10, a compute-focused card with an average score of 14,029. Here, the GTX 1660 Ti trails by 0.7%. The Tesla K10's marginal lead in this aggregate score is notable, though its driver support and feature set are geared toward professional compute, not gaming.

The NVIDIA RTX A2000 Mobile is also in this performance tier, with an average score of 13,821. The GTX 1660 Ti leads this mobile workstation part by 0.8%. This comparison highlights the efficiency of the desktop GTX 1660 Ti, which matches a newer mobile professional GPU in raw compute benchmarks.

Finally, the NVIDIA GeForce GTX 680, a previous-generation flagship, scores 13,812. The GTX 1660 Ti is 0.8% faster, demonstrating that the newer Turing-based card holds a slight edge over the older Kepler architecture while consuming significantly less power.

Benchmark Performance

Delving into the specific benchmark scores reveals the GTX 1660 Ti's strengths and weaknesses across different API workloads. The average score of 13,925 is the headline figure, but the individual tests provide a more granular view. In the Passmark G3D test, which is a general 3D gaming benchmark, the card achieves a score of 12,871. This is a strong result that aligns with its 55th percentile ranking, indicating it performs better than the majority of GPUs in the database.

The synthetic tests show a varied picture. In 3DMark Steel Nomad DX12, the card scores 1,061, which is a low number for this demanding next-generation test, suggesting that it struggles with the most modern DirectX 12 workloads. Conversely, the Geekbench OpenCL score of 60,863 is quite high, indicating robust compute performance for general-purpose tasks. The Vulkan score of 58,014 is similarly strong, showing good driver optimization and hardware support for this cross-platform API.

The Passmark suite offers a breakdown by DirectX version. The card performs best in legacy APIs, with a DirectX 9 score of 190 and a DirectX 10 score of 66. The DirectX 11 score is 103, and the DirectX 12 score drops to 51. This pattern suggests that while the hardware is capable, its driver performance degrades as the API becomes more modern and complex. However, the G2D score of 804 indicates solid 2D and desktop rendering performance, which is not a bottleneck.

The GPU Compute score of 5,226 is moderate, placing it below the top-tier compute cards but still ahead of many integrated solutions. When comparing to its nearest rivals, the aggregate deltaPct of 0.4% ahead of the RX 570X and 0.8% ahead of both the RTX A2000 Mobile and GTX 680 underscores how closely these GPUs are matched. The 0.7% deficit to the Tesla K10 shows that the GTX 1660 Ti's lead over the AMD card is effectively negligible. The benchmark results indicate that the GTX 1660 Ti is a well-rounded performer for its generation, with the data showing it as a capable 1080p gaming card that holds its own against a tightly packed field of competitors, but it lacks the hardware features to excel in the most demanding next-generation rendering workloads.