NVIDIA GeForce GTX 1660 Ti vs NVIDIA GeForce GTX 650 Ti Comparison

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA GeForce GTX 650 Ti has an average benchmark score of 8053, while the NVIDIA GeForce GTX 1660 Ti has an average score of 7723. Despite the 1660 Ti being a much newer part, the 650 Ti edges it out by roughly 4.3% in the aggregate metric used here.

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

A: The GeForce GTX 1660 Ti scores 51029 in Geekbench OpenCL, which is 84.6% higher than the GTX 650 Ti’s score of 7877. This is the largest performance gap recorded in the head-to-head data, heavily favoring the 1660 Ti.

Q: Is there any benchmark where the older GTX 650 Ti wins?

A: Yes, in Geekbench Vulkan, the GTX 650 Ti scores 8229 versus the GTX 1660 Ti’s 5827, a 41.2% advantage for the 650 Ti. This single win offsets the OpenCL result in the head-to-head tally, leaving both cards with one win apiece.

Q: What are the respective memory configurations?

A: The GTX 650 Ti uses 1024 MB of GDDR5 on a 128-bit bus, yielding 86.40 GB/s of bandwidth. The GTX 1660 Ti uses 6 GB of GDDR6 on a 192-bit bus, yielding 288.0 GB/s of bandwidth. The 1660 Ti has six times the memory capacity and over three times the bandwidth.

Q: Which card has a higher pixel fill rate?

A: The GTX 1660 Ti achieves 84.96 GPixel/s, compared to the GTX 650 Ti’s 14.85 GPixel/s. The 1660 Ti is roughly 5.7 times faster in pixel throughput, reflecting its much larger ROP count (48 vs 16).

Q: What are the power connector requirements for each card?

A: The GTX 650 Ti requires a single 6-pin power connector, while the GTX 1660 Ti requires a single 8-pin power connector. Both cards share the same suggested PSU rating of 300 W.

Architecture Differences

The GeForce GTX 650 Ti is built on the GK106S chip using NVIDIA’s Kepler architecture, fabricated on a 28 nm process at TSMC. The GeForce GTX 1660 Ti uses the TU116 chip with the Turing architecture, also from TSMC but on a 12 nm node. The process shrink is significant: the 650 Ti packs 2,540 million transistors into a 221 mm² die, while the 1660 Ti fits 6,600 million transistors into a 284 mm² die. Transistor density jumps from 11.5M per mm² on Kepler to 23.2M per mm² on Turing, nearly doubling the packing efficiency.

Shader resources differ substantially. The GTX 650 Ti has 768 shading units, 64 texture mapping units, and 16 ROPs. The GTX 1660 Ti doubles the shader count to 1536, increases TMUs to 96, and triples ROPs to 48. Neither card has dedicated ray tracing or tensor cores, so the architectural gains come from raw throughput scaling rather than specialized hardware.

Clock behavior also differs. The GTX 650 Ti lists no base or boost clock in the data, only a memory clock of 1350 MHz (5.4 Gbps effective). The GTX 1660 Ti has a base clock of 1500 MHz and a boost clock of 1770 MHz, with memory at 1500 MHz (12 Gbps effective). The 1660 Ti’s memory clock is effectively more than double, and the wider 192-bit bus further multiplies bandwidth.

API support diverges in DirectX version: the 650 Ti supports DirectX 12 (11_0), while the 1660 Ti supports DirectX 12 (12_1). Both cards support OpenGL 4.6, but Vulkan support differs — the 650 Ti is rated for Vulkan 1.2.175, while the 1660 Ti supports Vulkan 1.4. The 1660 Ti also provides half-precision FP16 throughput of 10.87 TFLOPS (2:1 ratio), whereas the 650 Ti has no listed FP16 capability.

Where Each One Wins

The GeForce GTX 1660 Ti wins decisively in compute-heavy workloads that scale with shader count, memory bandwidth, and modern driver optimization. Its Geekbench OpenCL score of 51029 dwarfs the 650 Ti’s 7877, indicating a massive advantage in general-purpose GPU compute tasks such as rendering, physics simulation, or data-parallel workloads. The 1660 Ti also has a texture rate of 169.9 GTexel/s versus 59.39 GTexel/s on the 650 Ti, making it far better suited for texture-intensive 3D scenes and higher-resolution gaming.

The GTX 650 Ti wins in the Geekbench Vulkan test, scoring 8229 against the 1660 Ti’s 5827. This 41.2% advantage suggests that in specific low-level graphics API workloads, the older Kepler architecture can outperform the newer Turing part — possibly due to driver maturity for Vulkan on that architecture or the specific workload characteristics of the test. The 650 Ti is also a single-slot card at 145 mm length, making it far easier to fit into compact or legacy systems, whereas the 1660 Ti is a dual-slot card at 229 mm length.

For gaming, the 1660 Ti’s higher pixel rate (84.96 GPixel/s vs 14.85 GPixel/s) and ROP count (48 vs 16) point to clear advantages at higher resolutions and with anti-aliasing enabled. The 650 Ti’s 1024 MB memory is a hard constraint for modern titles, while the 1660 Ti’s 6 GB provides ample headroom for larger textures and higher detail settings. However, the 650 Ti’s Vulkan win shows it is not universally obsolete — in certain API-bound scenarios, it holds its own.

Specification Differences

The two cards differ across nearly every specification category. Process node: 28 nm for the 650 Ti versus 12 nm for the 1660 Ti. Transistor count: 2,540 million versus 6,600 million. Die size: 221 mm² versus 284 mm². Transistor density: 11.5M / mm² versus 23.2M / mm².

Memory: 1024 MB GDDR5 on a 128-bit bus with 86.40 GB/s bandwidth versus 6 GB GDDR6 on a 192-bit bus with 288.0 GB/s bandwidth. Shading units: 768 versus 1536. TMUs: 64 versus 96. ROPs: 16 versus 48. Pixel rate: 14.85 GPixel/s versus 84.96 GPixel/s. Texture rate: 59.39 GTexel/s versus 169.9 GTexel/s. FP32: 1,425.4 GFLOPS versus 5.437 TFLOPS. FP16: not listed versus 10.87 TFLOPS (2:1).

Power and physical design: TDP is 110 W for the 650 Ti versus 120 W for the 1660 Ti. Slot width is single-slot versus dual-slot. Power connectors are 1x 6-pin versus 1x 8-pin. Suggested PSU is 300 W for both. Dimensions: 145 mm (5.7 inches) length for the 650 Ti; the 1660 Ti is 229 mm (9 inches) long, 111 mm (4.4 inches) high, and 35 mm (1.4 inches) wide.

Display outputs: the 650 Ti has 2x DVI and 1x mini-HDMI 1.4a; the 1660 Ti has 1x DVI, 1x HDMI 2.0, and 1x DisplayPort 1.4a. DirectX support: 12 (11_0) versus 12 (12_1). Vulkan: 1.2.175 versus 1.4. Launch MSRP: 149 USD for the 650 Ti, 279 USD for the 1660 Ti. Release dates: 2012-10-08 versus 2019-02-21.

Head-to-Head Benchmarks

The head-to-head data contains only two benchmark comparisons, and they split sharply between the two cards.

The Geekbench OpenCL test is a landslide for the GTX 1660 Ti. It scores 51029 versus the GTX 650 Ti’s 7877, a delta of -84.6% from the 1660 Ti’s perspective. This means the 1660 Ti delivers roughly 6.5 times the OpenCL performance. The magnitude of this gap is consistent with the specification differences: the 1660 Ti has double the shading units, 3.8 times the FP32 throughput (5.437 TFLOPS vs 1,425.4 GFLOPS), and 3.3 times the memory bandwidth. In any compute task that is not bottlenecked by API overhead, the 1660 Ti should be overwhelmingly faster.

The Geekbench Vulkan test flips the result. The GTX 650 Ti scores 8229, while the GTX 1660 Ti scores 5827, giving the 650 Ti a 41.2% lead. This is a surprising outcome given the 1660 Ti’s raw spec advantages and newer architecture. The data suggests that in this specific Vulkan workload, the 650 Ti’s driver path or the workload’s characteristics favor Kepler over Turing. It is notably the 1660 Ti’s Vulkan score is lower than its own OpenCL score by a factor of roughly 8.8, while the 650 Ti’s Vulkan score is only slightly higher than its OpenCL score (8229 vs 7877). The 1660 Ti appears to underperform in Vulkan relative to its own capabilities, while the 650 Ti performs consistently across both APIs.

With one win each, the head-to-head tally is even. However, the magnitude of the OpenCL win (84.6% delta) far exceeds the magnitude of the Vulkan win (41.2% delta). In absolute score terms, the 1660 Ti’s OpenCL result (51029) is more than six times the 650 Ti’s Vulkan result (8229). The aggregate benchmark score — where the 650 Ti edges ahead at 8053 versus 7723 — appears to be driven by a broader set of tests that may not align with the head-to-head selections.

The Verdict

Based strictly on the data, the choice depends on the workload. For compute-heavy applications that leverage OpenCL — such as rendering, scientific simulation, or GPU-accelerated processing — the GeForce GTX 1660 Ti is the clear pick. Its 51029 OpenCL score represents an 84.6% advantage over the 650 Ti, and its 6 GB memory capacity and 288.0 GB/s bandwidth provide a far more capable platform for large datasets and modern workloads. The 1660 Ti also wins on raw throughput metrics: 5.437 TFLOPS FP32, 84.96 GPixel/s pixel rate, and 169.9 GTexel/s texture rate all point to superior performance in graphics-intensive scenarios.

For users targeting Vulkan-based applications specifically, the GTX 650 Ti’s 41.2% lead in Geekbench Vulkan is notable. However, this win comes with significant caveats: the 650 Ti has only 1024 MB of memory, a 128-bit bus, and 14.85 GPixel/s pixel rate, which will limit its real-world usefulness in modern titles regardless of the Vulkan benchmark result. The 650 Ti’s aggregate benchmark score of 8053 is higher than the 1660 Ti’s 7723, and its percentile rank of 42 versus 41 is marginally better, but this is a narrow margin that does not offset the 1660 Ti’s overwhelming OpenCL dominance.

The 1660 Ti is also the more future-proof option in terms of API support: DirectX 12 (12_1) versus 12 (11_0), and Vulkan 1.4 versus 1.2.175. Its dual-slot design and 8-pin power connector are standard for modern GPUs, whereas the 650 Ti’s single-slot form factor and 6-pin connector reflect its 2012-era design. Both cards are end-of-life, so longevity is not a differentiator.

In practical terms, the 1660 Ti is the card to choose for anyone running current software that can utilize its compute resources, memory capacity, and modern API features. The 650 Ti’s Vulkan win is an interesting data point but unlikely to translate into better real-world gaming performance given its hardware limitations. The verdict is straightforward: the GTX 1660 Ti is the superior card for the vast majority of use cases, with the only exception being specific Vulkan-bound workloads where the 650 Ti demonstrably leads.