NVIDIA GeForce GTX 660 Ti vs NVIDIA GeForce RTX 3070 Comparison

The Verdict

The data in this comparison is unambiguous. The NVIDIA GeForce RTX 3070 is the definitive choice for any modern workload, outperforming the GTX 660 Ti by a staggering margin in every recorded benchmark. The RTX 3070 delivers a 646.5% higher score in Geekbench OpenCL and a 40% higher score in Geekbench Vulkan compared to the GTX 660 Ti. The GTX 660 Ti, a Kepler-generation card from 2012, simply cannot compete with the Ampere architecture in any measurable way.

Who should pick which? The RTX 3070 is for anyone running contemporary applications, especially those that leverage Vulkan or OpenCL APIs. Its average benchmark score of 17,208 places it in the 61st percentile of all GPUs, a full 2,145 points ahead of the GTX 660 Ti’s average of 15,063. The GTX 660 Ti, sitting in the 57th percentile, is only suitable for legacy systems or very light, older workloads where its 2 GB memory buffer and 192-bit bus are sufficient. For any serious gaming or compute task, the RTX 3070 is the only rational option.

FAQ

Q: How much faster is the RTX 3070 than the GTX 660 Ti in compute benchmarks?

A: The RTX 3070 is dramatically faster. In Geekbench OpenCL, it scores 112,821 versus 15,113 for the GTX 660 Ti, a 646.5% advantage. In Geekbench Vulkan, the RTX 3070 scores 21,022 versus 15,012, a 40% lead.

Q: Which card has better API support for modern games?

A: The RTX 3070 supports DirectX 12 Ultimate (12_2), Vulkan 1.4, and OpenGL 4.6. The GTX 660 Ti supports DirectX 12 (11_0), Vulkan 1.2.175, and OpenGL 4.6. The RTX 3070’s DirectX 12 Ultimate feature set is a clear step forward.

Q: Is the GTX 660 Ti competitive in any benchmark category?

A: No. In the two head-to-head tests recorded (Geekbench OpenCL and Geekbench Vulkan), the RTX 3070 wins both. The GTX 660 Ti does not win any recorded test.

Q: What is the memory configuration difference?

A: The RTX 3070 has 8 GB of GDDR6 memory on a 256-bit bus, delivering 448.0 GB/s of bandwidth. The GTX 660 Ti has 2 GB of GDDR5 memory on a 192-bit bus, delivering 144.2 GB/s. The RTX 3070 has over three times the bandwidth.

Q: How do the two cards compare in terms of raw shading power?

A: The RTX 3070 has 5,888 shading units and delivers 20.31 TFLOPS of FP32 performance. The GTX 660 Ti has 1,344 shading units and delivers 2.634 TFLOPS. The RTX 3070 provides roughly 7.7 times the FP32 throughput.

Q: What are the average benchmark scores for each card?

A: The RTX 3070 has an average benchmark score of 17,208 across all recorded tests. The GTX 660 Ti has an average benchmark score of 15,063. The RTX 3070’s average is about 14% higher.

Architecture Differences

The architectural gap between these two GPUs is enormous, spanning nearly a decade of development. The RTX 3070 is built on the Ampere architecture using the GA104 chip, manufactured on an 8 nm process at Samsung. It integrates 17,400 million transistors onto a 392 mm² die, yielding a transistor density of 44.4 million per mm². The GTX 660 Ti uses the Kepler architecture with the GK104 chip, built on a 28 nm process at TSMC. It has 3,540 million transistors on a 294 mm² die, with a density of 12.0 million per mm². The RTX 3070’s process advantage allows for 4.9 times the transistor count in a die that is only 33% larger.

The RTX 3070 introduces hardware features that the Kepler architecture simply lacks. It includes 46 ray tracing cores and 184 tensor cores, enabling hardware-accelerated ray tracing and AI-based features like DLSS. The GTX 660 Ti has neither. The RTX 3070 also supports PCIe 4.0 x16, while the GTX 660 Ti is limited to PCIe 3.0 x16. The RTX 3070’s memory subsystem is entirely modern: 8 GB of GDDR6 on a 256-bit bus, while the GTX 660 Ti uses 2 GB of GDDR5 on a 192-bit bus.

The RTX 3070’s API support is also more advanced. It supports DirectX 12 Ultimate (12_2), which includes features like mesh shaders and variable rate shading. The GTX 660 Ti only supports DirectX 12 (11_0), a more limited feature set. Vulkan support also differs, with the RTX 3070 running Vulkan 1.4 versus the GTX 660 Ti’s Vulkan 1.2.175. Both cards support OpenGL 4.6, but that is where the similarities end.

Specification Differences

The specification table shows stark contrasts across every major component. The RTX 3070 operates at a base clock of 1500 MHz and boosts to 1725 MHz, while the GTX 660 Ti runs at 915 MHz base and 980 MHz boost. Memory clocks differ similarly: the RTX 3070 runs at 1750 MHz (14 Gbps effective) versus 1502 MHz (6 Gbps effective) for the GTX 660 Ti.

The compute units tell the story of generational progress. The RTX 3070 has 5,888 shading units, 184 texture mapping units, and 96 ROPs. The GTX 660 Ti has 1,344 shading units, 112 TMUs, and 24 ROPs. Pixel fill rate is 165.6 GPixel/s for the RTX 3070 versus 27.44 GPixel/s for the GTX 660 Ti. Texture fill rate is 317.4 GTexel/s versus 109.8 GTexel/s. FP32 performance is 20.31 TFLOPS versus 2.634 TFLOPS.

Power requirements also diverge significantly. The RTX 3070 has a TDP of 220 W and requires a 550 W power supply, using a single 12-pin connector. The GTX 660 Ti has a TDP of 150 W, requires a 450 W power supply, and uses two 6-pin connectors. Both cards are dual-slot designs, but the RTX 3070 has a longer PCB at 242 mm versus 241 mm for the GTX 660 Ti.

Display outputs differ as well. The RTX 3070 offers 1x HDMI 2.1 and 3x DisplayPort 1.4a. The GTX 660 Ti offers 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2. The RTX 3070’s modern outputs support higher resolutions and refresh rates.

Head-to-Head Benchmarks

The recorded head-to-head data shows a complete dominance by the RTX 3070. In Geekbench OpenCL, the RTX 3070 scores 112,821 against the GTX 660 Ti’s 15,113. This is a 646.5% delta, meaning the RTX 3070 is more than seven times faster in OpenCL compute workloads. This massive gap reflects not just the newer architecture, but also the 8 GB GDDR6 memory versus 2 GB GDDR5, along with the vastly higher shader count and clock speeds.

In Geekbench Vulkan, the RTX 3070 scores 21,022 versus 15,012 for the GTX 660 Ti, a 40% advantage. The smaller percentage gap in Vulkan is notable. It suggests that the GTX 660 Ti’s Kepler architecture, despite its age, handles Vulkan’s low-level API relatively efficiently. However, the RTX 3070 still wins decisively, and its Vulkan support is more modern with version 1.4 versus 1.2.175.

The overall benchmark averages reinforce this picture. The RTX 3070’s average score across its full benchmark suite (including Passmark tests, 3DMark Steel Nomad, and Geekbench) is 17,208. The GTX 660 Ti’s average, based only on its two Geekbench results, is 15,063. The RTX 3070 also holds a percentile rank of 61 against all GPUs, while the GTX 660 Ti sits at 57. The RTX 3070’s nearest rivals include the AMD Radeon RX 7600 XT (0.7% higher average score) and the NVIDIA Tesla K40c (1.5% higher), while the GTX 660 Ti’s nearest rivals are clustered close, such as the AMD Radeon Pro 560X (0.1% higher) and the AMD Radeon RX 7600 (0.7% higher).

The RTX 3070 wins 2 out of 2 head-to-head tests. The GTX 660 Ti wins none. This is not a close contest by any measure; the recorded data shows a generational chasm that no amount of driver optimization or overclocking could bridge.