NVIDIA GeForce GTX 660 vs NVIDIA GeForce GTX 760 Comparison

The NVIDIA GeForce GTX 760 and NVIDIA GeForce GTX 660 are two Kepler-based graphics cards from consecutive generations, and the benchmark data reveals a clear but not absolute hierarchy between them. The GTX 760 secures a victory in two of the three tested workloads, with its most commanding lead appearing in Vulkan performance, while the GTX 660 manages to edge out its newer counterpart in a single OpenCL test. Both cards share identical core clock speeds, yet the GTX 760 leverages a larger chip and wider memory bus to pull ahead in the majority of scenarios, making it the more consistent performer despite the GTX 660's surprising strength in one compute workload.

Head-to-Head Benchmarks

The most decisive result in this comparison comes from the Geekbench Vulkan test, where the GTX 760 scores 12,551 against the GTX 660's 11,415. This represents a 10% advantage for the GTX 760, the largest delta between the two cards across any benchmark. The Vulkan API, which relies heavily on parallel execution and memory throughput, clearly favors the GTX 760's superior hardware configuration. This is not a marginal win; it is a substantial gap that suggests the GTX 760 will handle modern, low-level API workloads with notably more headroom.

In the Geekbench Metal test, the GTX 760 posts a score of 4,524, while the GTX 660 trails at 4,305. The 5.1% delta in favor of the GTX 760 is more modest than its Vulkan lead, but it still represents a consistent pattern of superiority. Metal benchmarks often reflect raw shading throughput and memory bandwidth, and the GTX 760's higher pixel rate of 24.77 GPixel/s compared to 20.64 GPixel/s for the GTX 660 likely contributes to this edge. The GTX 760 also boasts a texture rate of 99.07 GTexel/s, well ahead of the GTX 660's 82.56 GTexel/s, which further explains its better showing in this graphics-heavy test.

The one reversal comes in Geekbench OpenCL, where the GTX 660 scores 11,347 versus the GTX 760's 11,299. The delta is a razor-thin -0.4% in favor of the GTX 660, a margin so small that it falls within typical run-to-run variance. This result is curious because the GTX 760 has a higher theoretical FP32 throughput of 2.378 TFLOPS compared to 1.981 TFLOPS for the GTX 660. However, OpenCL workloads can be sensitive to driver scheduling and memory access patterns, and the GTX 660's narrower 192-bit bus, which delivers 144.2 GB/s versus the GTX 760's 192.3 GB/s, did not hold it back in this specific test. It is a minor upset, but it does not overturn the overall trend.

Aggregating the three results, the GTX 760's average benchmark score stands at 9,458, while the GTX 660 averages 9,022. That is a 4.8% difference in the GTX 760's favor, aligning closely with its percentile ranking of 46 versus the GTX 660's 45. When placed against external rivals, the GTX 760 sits just 0.2% behind the NVIDIA GeForce GTX TITAN BLACK (average score 9,474) and 1.6% ahead of the AMD Radeon R7 M380 (9,313). The GTX 660, meanwhile, is 0.2% behind the NVIDIA TITAN V CEO Edition (9,037) and 0.4% behind the NVIDIA GeForce GTX 560 (9,058). These external comparisons show both cards are clustered tightly with similar-generation hardware, but the GTX 760 consistently holds a small edge in the head-to-head matchups.

FAQ

Q: Which card wins the majority of benchmark tests?

A: The NVIDIA GeForce GTX 760 wins two out of three head-to-head benchmarks, taking Geekbench Metal and Geekbench Vulkan, while the GTX 660 wins only Geekbench OpenCL.

Q: How large is the GTX 760's lead in Vulkan performance?

A: The GTX 760 scores 12,551 in Geekbench Vulkan versus 11,415 for the GTX 660, producing a 10% delta in favor of the GTX 760.

Q: Is there any test where the GTX 660 outperforms the GTX 760?

A: Yes, in Geekbench OpenCL the GTX 660 scores 11,347 against the GTX 760's 11,299, giving the GTX 660 a -0.4% advantage, which is within run-to-run noise.

Q: Do both cards have the same core clock speeds?

A: Yes, both the GTX 760 and GTX 660 feature a base clock of 980 MHz and a boost clock of 1032 MHz, with identical memory clocks of 1502 MHz (6 Gbps effective).

Q: How do the two cards compare in average benchmark score?

A: The GTX 760 has an average benchmark score of 9,458, while the GTX 660 averages 9,022, a difference of approximately 4.8% in favor of the GTX 760.

Q: What are the percentile rankings for each card?

A: The GTX 760 ranks in the 46th percentile among all GPUs, while the GTX 660 ranks in the 45th percentile, showing they are closely matched in overall standing.

Architecture Differences

Both cards are built on NVIDIA's Kepler architecture and manufactured by TSMC on a 28 nm process, but they use different chips. The GTX 760 is powered by the GK104 chip, while the GTX 660 uses the smaller GK106. This difference in silicon is substantial: the GK104 contains 3,540 million transistors on a 294 mm² die, whereas the GK106 has 2,540 million transistors on a 221 mm² die. Consequently, the GTX 760 achieves a higher transistor density of 12.0M / mm² compared to 11.5M / mm² for the GTX 660.

The shading unit count is another major divergence. The GTX 760 features 1,152 shading units, 96 texture mapping units (TMUs), and 32 render output units (ROPs). The GTX 660 scales down to 960 shading units, 80 TMUs, and 24 ROPs. These reductions directly impact fill rates, as the GTX 760 delivers a pixel rate of 24.77 GPixel/s and a texture rate of 99.07 GTexel/s, while the GTX 660 falls to 20.64 GPixel/s and 82.56 GTexel/s, respectively. The FP32 compute throughput also favors the GTX 760 at 2.378 TFLOPS versus 1.981 TFLOPS for the GTX 660.

Memory configuration further separates the two. The GTX 760 uses a 256-bit memory bus with a bandwidth of 192.3 GB/s, while the GTX 660 is constrained to a 192-bit bus with 144.2 GB/s. Both cards ship with 2 GB of GDDR5 memory, but the wider bus gives the GTX 760 a clear advantage in bandwidth-heavy scenarios. The GTX 760 also draws more power, rated at a TDP of 170 W compared to the GTX 660's 140 W, and requires two 6-pin power connectors versus a single 6-pin connector on the GTX 660. The suggested power supply rating reflects this, with the GTX 760 calling for 450 W and the GTX 660 for 300 W.

Both cards share identical display outputs (2x DVI, 1x HDMI 1.4a, 1x DisplayPort 1.2), a dual-slot form factor, and the same PCIe 3.0 x16 bus interface. Their API support is also identical, with DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. The GTX 760 belongs to the GeForce 700 generation with the GK104 architecture, while the GTX 660 is from the GeForce 600 generation with GK106, though both are end-of-life products. Release timing differs, with the GTX 660 launching earlier, but the GTX 760's launch MSRP of 249 USD was only slightly above the GTX 660's 229 USD.

The Verdict

The data presents a straightforward recommendation for most users: the NVIDIA GeForce GTX 760 is the stronger card. It wins the majority of benchmark tests, including a decisive 10% victory in Vulkan and a 5.1% win in Metal, and it holds a meaningful lead in average benchmark score at 9,458 versus 9,022. Its superior hardware—more shading units, higher texture and pixel rates, and a wider memory bus—translates into measurable performance gains that would be noticeable in real-world workloads, particularly those leveraging Vulkan or Metal APIs. The GTX 760 also sits in a slightly higher percentile (46 vs 45), reinforcing its position as the more capable of the two.

However, the GTX 660 is not without merit. Its OpenCL score of 11,347 edges past the GTX 760's 11,299, and it does so while consuming less power (140 W TDP vs 170 W) and requiring a less demanding power supply (300 W vs 450 W). For users with a power supply that lacks a second 6-pin connector, or those who prioritize compute tasks that favor OpenCL, the GTX 660 could be a sensible pick. Its lower power draw also means less heat output, which may appeal in compact builds.

Yet, the broader picture favors the GTX 760. The 10% Vulkan lead is too large to ignore, and the consistent wins in two of three tests suggest it is the more future-proof option for graphics-intensive applications. The GTX 660's single OpenCL victory is marginal, and its lower bandwidth and reduced shading resources place it at a structural disadvantage. For gaming and general GPU compute, the GTX 760 is the clear choice, while the GTX 660 remains a viable alternative only for those constrained by power delivery or specifically targeting OpenCL workloads.

Specification Differences

The following specifications differ between the two cards:

• Chip: GK104 (GTX 760) vs GK106 (GTX 660)
• Generation: GeForce 700 (GTX 760) vs GeForce 600 (GTX 660)
• Transistors: 3,540 million (GTX 760) vs 2,540 million (GTX 660)
• Die Size: 294 mm² (GTX 760) vs 221 mm² (GTX 660)
• Transistor Density: 12.0M / mm² (GTX 760) vs 11.5M / mm² (GTX 660)
• Memory Bus Width: 256 bit (GTX 760) vs 192 bit (GTX 660)
• Memory Bandwidth: 192.3 GB/s (GTX 760) vs 144.2 GB/s (GTX 660)
• Shading Units: 1,152 (GTX 760) vs 960 (GTX 660)
• TMUs: 96 (GTX 760) vs 80 (GTX 660)
• ROPs: 32 (GTX 760) vs 24 (GTX 660)
• Pixel Rate: 24.77 GPixel/s (GTX 760) vs 20.64 GPixel/s (GTX 660)
• Texture Rate: 99.07 GTexel/s (GTX 760) vs 82.56 GTexel/s (GTX 660)
• FP32: 2.378 TFLOPS (GTX 760) vs 1.981 TFLOPS (GTX 660)
• TDP: 170 W (GTX 760) vs 140 W (GTX 660)
• Power Connectors: 2x 6-pin (GTX 760) vs 1x 6-pin (GTX 660)
• Suggested PSU: 450 W (GTX 760) vs 300 W (GTX 660)
• Release Date: 2013-06-24 (GTX 760) vs 2012-09-05 (GTX 660)
• Predecessor: GeForce 600 (GTX 760) vs GeForce 500 (GTX 660)
• Successor: GeForce 900 (GTX 760) vs GeForce 700 (GTX 660)