NVIDIA GeForce GTX 1660 SUPER vs NVIDIA GeForce GTX 670 Comparison

NVIDIA GeForce GTX 1660 SUPER and NVIDIA GeForce GTX 670 represent two distinct eras of GPU design, with the former built on Turing architecture and the latter on Kepler. The benchmark data indicates a decisive generational shift, as the GTX 1660 SUPER outperforms the GTX 670 by a substantial margin in every shared test, though the GTX 670 remains competitive in overall percentile ranking.

Head-to-Head Benchmarks

The direct comparison between these two cards is limited to two compute-oriented tests, both of which show overwhelming victories for the GTX 1660 SUPER. In Geekbench OpenCL, the GTX 1660 SUPER scores 52,490, while the GTX 670 manages only 15,341. This represents a 242.2% advantage for the newer card. The margin is even larger in Geekbench Vulkan, where the GTX 1660 SUPER achieves 57,102 against the GTX 670's 15,553, a 267.1% lead.

These results are not subtle. The GTX 1660 SUPER delivers more than three times the raw compute throughput in both API tests. The Vulkan delta is particularly telling, as it reflects the newer card's superior architecture-level support for modern graphics APIs. The GTX 670's Vulkan score of 15,553 places it far behind, and its OpenCL result of 15,341 is similarly dwarfed.

The overall win count reflects this dominance: the GTX 1660 SUPER wins 2 head-to-head benchmarks, while the GTX 670 wins 0. However, it is notably the average benchmark scores tell a slightly different story. The GTX 1660 SUPER has an average score of 12,986 across its full benchmark suite, while the GTX 670 averages 12,773. This puts the GTX 670 within 1.7% of the newer card in overall performance, despite the massive gap in the two shared tests. This discrepancy arises because the GTX 670's average is derived from a different set of benchmarks, including a Metal score of 7,424, which is not present in the GTX 1660 SUPER's results.

Where Each One Wins

The GTX 1660 SUPER wins decisively in compute-heavy workloads. Its OpenCL and Vulkan scores indicate it is far better suited for tasks like GPU-accelerated rendering, machine learning inference, and general-purpose compute. The 267.1% lead in Vulkan specifically suggests strong performance in modern games that leverage this API for lower overhead and better multi-threading.

The GTX 670, despite losing both head-to-head tests, holds its own in the broader benchmark landscape. Its percentile rank of 52 places it just one point below the GTX 1660 SUPER's 53rd percentile. This means that in the full spectrum of GPUs, the GTX 670 is not far behind. Its average score of 12,773 is within striking distance of the GTX 1660 SUPER's 12,986, and its nearest rivals include the NVIDIA GeForce GTX 590 (deltaPct -0.4) and the AMD Radeon Pro 455 (deltaPct -0.4), indicating it sits in a competitive mid-range segment.

For legacy DirectX workloads, the GTX 670 may still be relevant. While the FACT PACK does not include DirectX benchmark scores for the GTX 670, its Kepler architecture was designed for that era's APIs. The GTX 1660 SUPER, by contrast, shows strong results in PassMark DirectX 9 (189) and DirectX 11 (104) tests, though these are not compared directly against the GTX 670.

The GTX 1660 SUPER also wins on power efficiency in practical terms. Its TDP is 125 W versus the GTX 670's 170 W, and it requires a 300 W suggested PSU versus 450 W. This means the newer card delivers dramatically higher compute performance while consuming less power, a clear win for system builders and users with smaller power supplies.

FAQ

Q: Which card has the higher Geekbench OpenCL score?

A: The NVIDIA GeForce GTX 1660 SUPER scores 52,490, which is 242.2% higher than the GTX 670's 15,341.

Q: How much faster is the GTX 1660 SUPER in Vulkan performance?

A: The GTX 1660 SUPER achieves 57,102 in Geekbench Vulkan, a 267.1% improvement over the GTX 670's 15,553.

Q: Do the cards have similar overall performance?

A: The average benchmark scores are close: the GTX 1660 SUPER averages 12,986, while the GTX 670 averages 12,773. The GTX 670 sits at the 52nd percentile of all GPUs, just one point below the GTX 1660 SUPER's 53rd percentile.

Q: What is the memory configuration difference?

A: The GTX 1660 SUPER has 6 GB of GDDR6 memory with a 192-bit bus and 336.0 GB/s bandwidth. The GTX 670 has 2 GB of GDDR5 memory with a 256-bit bus and 192.3 GB/s bandwidth.

Q: Which card requires more power from the PSU?

A: The GTX 670 has a TDP of 170 W and requires a 450 W suggested PSU with two 6-pin connectors. The GTX 1660 SUPER has a TDP of 125 W, needs a 300 W suggested PSU, and uses a single 8-pin connector.

Q: Does the GTX 670 support modern APIs?

A: The GTX 670 supports DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. The GTX 1660 SUPER supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4.

Specification Differences

The two cards differ across nearly every core specification. The GTX 1660 SUPER uses a 12 nm process node, while the GTX 670 uses 28 nm, both fabricated by TSMC. The transistor count is 6,600 million for the newer card versus 3,540 million for the older one, though the die size is similar: 284 mm² for the GTX 1660 SUPER and 294 mm² for the GTX 670. This yields a transistor density of 23.2 million per mm² for the GTX 1660 SUPER versus 12.0 million per mm² for the GTX 670.

Clock speeds show a significant shift: the GTX 1660 SUPER runs at 1530 MHz base and 1785 MHz boost, while the GTX 670 runs at 915 MHz base and 980 MHz boost. Memory speed is also much higher on the newer card, with 14 Gbps effective versus 6 Gbps effective. Memory capacity is 6 GB versus 2 GB, and the bus width is 192-bit versus 256-bit, yet the GTX 1660 SUPER still achieves higher bandwidth at 336.0 GB/s compared to 192.3 GB/s.

The shading units are close: 1408 for the GTX 1660 SUPER versus 1344 for the GTX 670. However, TMUs and ROPs differ: the GTX 1660 SUPER has 88 TMUs and 48 ROPs, while the GTX 670 has 112 TMUs and 32 ROPs. Pixel rate is 85.68 GPixel/s for the newer card versus 27.44 GPixel/s for the older one, and texture rate is 157.1 GTexel/s versus 109.8 GTexel/s.

FP32 performance is 5.027 TFLOPS for the GTX 1660 SUPER versus 2.634 TFLOPS for the GTX 670. The newer card also supports FP16 at 10.05 TFLOPS (2:1), while the GTX 670 has no FP16 capability listed. Physical dimensions differ slightly: the GTX 1660 SUPER is 229 mm long, 111 mm high, and 35 mm wide, while the GTX 670 is 241 mm long, 111 mm high, and 38 mm wide. Display outputs also differ, with the GTX 1660 SUPER offering 1x DVI, 1x HDMI 2.0, and 1x DisplayPort 1.4a, versus the GTX 670's 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2.

Architecture Differences

The architectural divide is fundamental. The GTX 1660 SUPER is built on the Turing architecture with the TU116 chip, part of the GeForce 16 generation. The GTX 670 uses the Kepler architecture with the GK104 chip, from the GeForce 600 generation. Neither card has ray tracing cores or tensor cores, but the Turing design brings a modern feature set that Kepler lacks.

The process technology alone explains much of the performance gap: 12 nm versus 28 nm allows the GTX 1660 SUPER to pack more than double the transistors into a similar die area. This enables higher clock speeds (1785 MHz boost versus 980 MHz) and significantly higher compute throughput. The GTX 1660 SUPER's FP32 performance of 5.027 TFLOPS is nearly double the GTX 670's 2.634 TFLOPS.

Memory architecture is another major divergence. The GTX 1660 SUPER uses GDDR6 with a 192-bit bus, achieving 336.0 GB/s bandwidth. The GTX 670 uses GDDR5 with a wider 256-bit bus but only reaches 192.3 GB/s. The newer memory type and higher effective speed (14 Gbps versus 6 Gbps) more than compensate for the narrower bus.

API support reflects the newer architecture's capabilities. The GTX 1660 SUPER supports DirectX 12 (12_1) and Vulkan 1.4, while the GTX 670 supports DirectX 12 (11_0) and Vulkan 1.2.175. This means the newer card can take advantage of more advanced rendering features in modern games and applications.

Power characteristics also differ due to architectural efficiency. The GTX 1660 SUPER has a TDP of 125 W, while the GTX 670 draws 170 W. Despite the lower power draw, the newer card delivers far higher performance, indicating a substantial improvement in performance-per-watt. The GTX 1660 SUPER also has a smaller physical footprint (229 mm versus 241 mm length) and uses a single 8-pin power connector versus the GTX 670's two 6-pin connectors, making it easier to install in compact systems.