NVIDIA GeForce GTX 560 vs NVIDIA GeForce GTX 660 Comparison

NVIDIA GeForce GTX 560 and NVIDIA GeForce GTX 660 are two end-of-life graphics cards from consecutive generations, separated by roughly 16 months of silicon progress. The GTX 660 wins the only shared benchmark in the data, but the GTX 560 holds a slight edge in average benchmark score across all GPUs. This analysis breaks down where each card actually pulls ahead, based strictly on the available specifications and benchmark results.

Where Each One Wins

The GTX 660 has a clear win in the only direct head-to-head benchmark available: Geekbench OpenCL. It scores 11,347 points versus the GTX 560's 9,058, a 20.2% advantage. This is a decisive margin that points to the GTX 660 being the stronger compute performer in raw OpenCL workloads. The GTX 660 also demonstrates versatility with additional benchmark scores in Geekbench Metal (4,305) and Geekbench Vulkan (11,415), which the GTX 560 lacks entirely. This means the GTX 660 is the only one of the two with measurable Vulkan performance, giving it a functional edge in modern API-based workloads.

The GTX 560's win is narrower and more statistical. Its average benchmark score across all GPUs is 9,058, which is 0.4% higher than the GTX 660's average of 9,022. In the nearest rivals data, the GTX 560 sits 0.2% above the NVIDIA TITAN V CEO Edition (9,037) and 1.6% above the AMD Radeon 550X (8,918), while the GTX 660 is 0.2% below the TITAN V CEO Edition and 1.2% above the Radeon 550X. This suggests that in mixed or aggregated workloads, the GTX 560 slightly outperforms its successor. However, this gap is small enough that it could be within run-to-run variance, and the GTX 660's dedicated OpenCL win carries more weight for actual compute tasks.

For gaming and texture-heavy workloads, the GTX 660's higher texture rate (82.56 GTexel/s vs 45.36 GTexel/s) and pixel rate (20.64 GPixel/s vs 11.34 GPixel/s) indicate it should handle resolution scaling and texture filtering better. The GTX 560 counters with more ROPs (32 vs 24), which historically helps with fill-rate-bound scenarios at lower resolutions. In practical terms, the GTX 660 is the better all-rounder for modern workloads, while the GTX 560's edge is confined to aggregate benchmark averages.

Architecture Differences

The two cards come from different architectural generations. The GTX 560 uses the GF114 chip built on Fermi 2.0 architecture, fabricated on a 40 nm process at TSMC. The GTX 660 uses the GK106 chip with Kepler architecture, fabricated on a 28 nm process, also at TSMC. This node shrink is significant: the GTX 660 packs 2,540 million transistors into a 221 mm² die, achieving a transistor density of 11.5M per mm². The GTX 560 has 1,950 million transistors on a much larger 332 mm² die, with a density of only 5.9M per mm². The GTX 660 is therefore more than twice as dense, which explains how it delivers higher performance at a lower thermal design power (140 W vs 150 W).

The Kepler architecture in the GTX 660 brings a major shader count increase. The GTX 660 has 960 shading units and 80 texture mapping units (TMUs), compared to 336 shading units and 56 TMUs on the GTX 560. This nearly triples the shader count and increases TMUs by roughly 43%. However, the GTX 560 has more ROPs (32 vs 24), which is a counterintuitive advantage for a last-gen card. The GTX 660's memory subsystem is also different: it uses a 192-bit bus with 2 GB of GDDR5, while the GTX 560 uses a wider 256-bit bus but only 1 GB of GDDR5. Despite the narrower bus, the GTX 660's memory runs at a higher effective speed (6 Gbps vs 4 Gbps), resulting in higher bandwidth (144.2 GB/s vs 128.0 GB/s).

Connectivity and API support differ meaningfully. The GTX 660 uses PCIe 3.0 x16, while the GTX 560 uses PCIe 2.0 x16. The GTX 660 also has Vulkan support (version 1.2.175), which the GTX 560 lacks entirely. Both support DirectX 12 (11_0) and OpenGL 4.6. Display outputs differ as well: the GTX 560 offers 2x DVI and 1x mini-HDMI 1.3a, while the GTX 660 provides 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2. Power requirements are lower on the GTX 660, needing a single 6-pin connector and a 300 W suggested PSU, versus the GTX 560's dual 6-pin connectors and 450 W suggested PSU. The GTX 660 is also longer at 241 mm (9.5 inches) versus the GTX 560's 210 mm (8.3 inches).

FAQ

Q: Which card is faster in OpenCL compute?

A: The GTX 660 is significantly faster, scoring 11,347 in Geekbench OpenCL compared to the GTX 560's 9,058. This represents a 20.2% advantage for the GTX 660 in this specific test.

Q: Does the GTX 560 have any performance advantage over the GTX 660?

A: In aggregate benchmark averages, the GTX 560 scores 9,058 versus the GTX 660's 9,022, a 0.4% difference. The GTX 560 also has more ROPs (32 vs 24), which can help in certain fill-rate-bound scenarios.

Q: Can the GTX 560 run Vulkan applications?

A: No. The GTX 560's API support includes DirectX 12 (11_0) and OpenGL 4.6, but Vulkan is not listed. The GTX 660 supports Vulkan version 1.2.175 and has a Geekbench Vulkan score of 11,415.

Q: Which card has better memory bandwidth?

A: The GTX 660, with 144.2 GB/s of bandwidth, outperforms the GTX 560's 128.0 GB/s. This is achieved through a 192-bit bus running at 6 Gbps effective, versus the GTX 560's 256-bit bus at 4 Gbps effective.

Q: Is the GTX 660 more power-efficient than the GTX 560?

A: Yes. The GTX 660 has a 140 W TDP and requires a single 6-pin power connector with a 300 W suggested PSU. The GTX 560 has a 150 W TDP, dual 6-pin connectors, and a 450 W suggested PSU.

Q: Which card has more shading units?

A: The GTX 660 has 960 shading units, which is nearly triple the GTX 560's 336 shading units. This is a major architectural advantage for the Kepler-based card.

Specification Differences

The two cards differ in nearly every major specification category:

• Chip and Architecture: GF114 (Fermi 2.0) on the GTX 560 vs GK106 (Kepler) on the GTX 660.
• Process Node: 40 nm on the GTX 560 vs 28 nm on the GTX 660.
• Transistors: 1,950 million on the GTX 560 vs 2,540 million on the GTX 660.
• Die Size: 332 mm² on the GTX 560 vs 221 mm² on the GTX 660.
• Transistor Density: 5.9M / mm² on the GTX 560 vs 11.5M / mm² on the GTX 660.
• Memory Speed: 1000 MHz / 4 Gbps effective on the GTX 560 vs 1502 MHz / 6 Gbps effective on the GTX 660.
• Memory Size: 1024 MB on the GTX 560 vs 2 GB on the GTX 660.
• Memory Bus Width: 256 bit on the GTX 560 vs 192 bit on the GTX 660.
• Memory Bandwidth: 128.0 GB/s on the GTX 560 vs 144.2 GB/s on the GTX 660.
• Shading Units: 336 on the GTX 560 vs 960 on the GTX 660.
• TMUs: 56 on the GTX 560 vs 80 on the GTX 660.
• ROPs: 32 on the GTX 560 vs 24 on the GTX 660.
• Pixel Rate: 11.34 GPixel/s on the GTX 560 vs 20.64 GPixel/s on the GTX 660.
• Texture Rate: 45.36 GTexel/s on the GTX 560 vs 82.56 GTexel/s on the GTX 660.
• FP32 Performance: 1,088.6 GFLOPS on the GTX 560 vs 1.981 TFLOPS on the GTX 660.
• TDP: 150 W on the GTX 560 vs 140 W on the GTX 660.
• Power Connectors: 2x 6-pin on the GTX 560 vs 1x 6-pin on the GTX 660.
• Suggested PSU: 450 W on the GTX 560 vs 300 W on the GTX 660.
• Bus Interface: PCIe 2.0 x16 on the GTX 560 vs PCIe 3.0 x16 on the GTX 660.
• Display Outputs: 2x DVI, 1x mini-HDMI 1.3a on the GTX 560 vs 2x DVI, 1x HDMI 1.4a, 1x DisplayPort 1.2 on the GTX 660.
• Vulkan Support: Not listed on the GTX 560 vs version 1.2.175 on the GTX 660.
• Card Length: 210 mm (8.3 inches) on the GTX 560 vs 241 mm (9.5 inches) on the GTX 660.
• Launch MSRP: 199 USD on the GTX 560 vs 229 USD on the GTX 660.

Head-to-Head Benchmarks

The sole head-to-head benchmark is Geekbench OpenCL, and it is a decisive win for the GTX 660. The GTX 660 scores 11,347 points while the GTX 560 manages 9,058, yielding a 20.2% delta in favor of the newer card. This is not a marginal difference; it is a substantial performance gap that aligns with the GTX 660's architectural advantages. The GTX 660 has nearly three times the shading units (960 vs 336), roughly 82% more texture fill rate (82.56 GTexel/s vs 45.36 GTexel/s), and nearly double the pixel rate (20.64 GPixel/s vs 11.34 GPixel/s). Its FP32 throughput of 1.981 TFLOPS is also almost double the GTX 560's 1,088.6 GFLOPS. These raw compute advantages translate directly into the observed OpenCL result.

The GTX 560's aggregate average score of 9,058 slightly exceeds the GTX 660's 9,022, but this is a 0.4% difference that is unlikely to be meaningful in real workloads. The nearest rivals data shows both cards clustered closely around the 9,000-point mark, with the AMD Radeon 550X trailing at 8,918 and the AMD Radeon 890M leading at 9,210. The GTX 560's percentile rank is 45th among all GPUs, identical to the GTX 660's 45th percentile. This suggests that in the broader GPU landscape, neither card is a standout; they occupy similar mid-tier positions.

The GTX 660 also has additional benchmark data points that the GTX 560 lacks: a Geekbench Metal score of 4,305 and a Geekbench Vulkan score of 11,415. The Vulkan score is particularly notable because it is nearly identical to the OpenCL score, indicating consistent performance across different compute APIs. The GTX 560 has no Vulkan or Metal scores listed, meaning its compute capability is only measurable through OpenCL. For users running Vulkan-based applications, the GTX 660 is the only viable option between the two.

The Verdict

The data is unambiguous: the NVIDIA GeForce GTX 660 is the stronger card for compute-heavy and modern workloads. Its 20.2% OpenCL advantage over the GTX 560, combined with Vulkan support and a nearly triple shader count, makes it the clear pick for anyone running OpenCL or Vulkan applications. The GTX 660 also offers double the memory (2 GB vs 1 GB), higher bandwidth (144.2 GB/s vs 128.0 GB/s), and better power efficiency (140 W TDP with a single 6-pin connector versus 150 W with dual 6-pin). Its higher pixel and texture rates suggest better gaming performance in most scenarios.

The GTX 560's only statistical win is a 0.4% higher average benchmark score, which is negligible in practice. Its advantage in ROP count (32 vs 24) might help in specific fill-rate-bound situations, but this does not compensate for the GTX 660's overwhelming advantages in shader throughput, memory capacity, and API support. Users building a system today should choose the GTX 660 for its superior compute performance, modern connectivity (PCIe 3.0, DisplayPort 1.2), and lower power demands. The GTX 560 is only relevant if you specifically need a shorter card (210 mm vs 241 mm) or are constrained to a 450 W PSU budget, but even then, the GTX 660's lower 300 W PSU requirement makes it easier to integrate. For any general-purpose use, the GTX 660 wins decisively.