NVIDIA GeForce GTX 650 Ti vs NVIDIA GeForce GTX 660 Comparison

FAQ

Q: How does the NVIDIA GeForce GTX 660 compare to the GTX 650 Ti in overall benchmark performance?

A: The database shows the GTX 660 has an average benchmark score of 9022, while the GTX 650 Ti scores 8053. That puts the GTX 660 roughly 12% higher overall, and it places in the 45th percentile of all GPUs versus the 42nd percentile for the GTX 650 Ti.

Q: Which card wins in the OpenCL and Vulkan benchmarks, and by how much?

A: The GTX 660 wins both head-to-head tests. In Geekbench OpenCL, it scores 11347 against 7877 for the GTX 650 Ti, a 44.1% advantage. In Geekbench Vulkan, it scores 11415 versus 8229, a 38.7% lead.

Q: What are the memory configurations of the two cards?

A: The GTX 660 comes with 2 GB of GDDR5 memory on a 192-bit bus, delivering 144.2 GB/s of bandwidth. The GTX 650 Ti has 1024 MB (1 GB) of GDDR5 on a 128-bit bus, with 86.40 GB/s of bandwidth. The GTX 660 has both more capacity and nearly 67% more bandwidth.

Q: Are the shading resources different between the two?

A: Yes. The GTX 660 has 960 shading units, 80 texture mapping units (TMUs), and 24 ROPs. The GTX 650 Ti is cut down to 768 shading units, 64 TMUs, and 16 ROPs. The GTX 660 has 25% more shaders, 25% more TMUs, and 50% more ROPs.

Q: How do the compute rates compare?

A: The GTX 660 achieves 1.981 TFLOPS of FP32 performance, a pixel rate of 20.64 GPixel/s, and a texture rate of 82.56 GTexel/s. The GTX 650 Ti delivers 1,425.4 GFLOPS (about 1.425 TFLOPS), 14.85 GPixel/s, and 59.39 GTexel/s. The GTX 660 leads by roughly 39% in FP32, 39% in pixel rate, and 39% in texture rate.

Q: What are the physical and power differences?

A: The GTX 660 is a dual-slot card, 241 mm (9.5 inches) long, with a 140 W TDP. The GTX 650 Ti is a single-slot card, 145 mm (5.7 inches) long, with a 110 W TDP. Both use a single 6-pin power connector and require a 300 W power supply. The GTX 660 also has a DisplayPort 1.2 output, while the GTX 650 Ti has a mini-HDMI 1.4a instead.

The Verdict

The data is unambiguous: the NVIDIA GeForce GTX 660 is the stronger card in every measured benchmark. It wins both available head-to-head tests, and its average score sits roughly 969 points higher than the GTX 650 Ti. For any workload that touches compute, shader throughput, or memory bandwidth, the GTX 660 is the clear choice.

The GTX 650 Ti is not without merit, however. Its 110 W TDP and single-slot, 145 mm design make it a far more compact and power-efficient option for small-form-factor builds or secondary machines. It also uses the same 6-pin power connector and 300 W suggested PSU, so it drops into similar power envelopes.

Who should pick which? If the priority is raw performance in OpenCL or Vulkan tasks, the GTX 660 is the only rational pick, with a 38.7% to 44.1% lead in those tests. If the priority is physical size, thermal footprint, and lower power draw, the GTX 650 Ti is the one to grab. But for gaming or general GPU compute, the data shows no contest.

Head-to-Head Benchmarks

The database lists two direct comparative tests: Geekbench OpenCL and Geekbench Vulkan. The GTX 660 wins both, and the margins are substantial.

In Geekbench OpenCL, the GTX 660 posts 11347 points, while the GTX 650 Ti manages 7877. That is a delta of 44.1%, meaning the GTX 660 scores nearly one and a half times as high. The GTX 650 Ti's score is closer to the performance of an NVIDIA GTX 880M (averaging 8040) or a Quadro P5000 (8039), both of which are within 0.2% of the GTX 650 Ti's average.

In Geekbench Vulkan, the GTX 660 records 11415 points against the GTX 650 Ti's 8229. The delta is 38.7%. The GTX 650 Ti's Vulkan score is near the performance of the GTX 650 Ti Boost (which averages 8067), a card that sits just 0.2% higher in average score.

Looking at wider context, the GTX 660's average score of 9022 places it just 0.2% below the NVIDIA TITAN V CEO Edition (9037) and 0.4% below the GTX 560 (9058). It sits 1.2% above the AMD Radeon 550X (8918) and 1.8% above the AMD Radeon Pro WX 5100 (8863). The GTX 650 Ti, by comparison, averages 8053, which is 0.2% above the GTX 880M (8040) and 0.2% above the Quadro P5000 (8039), while sitting 0.2% below the GTX 650 Ti Boost (8067) and 0.3% below the GRID K2 (8080).

The takeaway: the GTX 660 does not just beat the GTX 650 Ti; it outperforms it by margins that are consistent across compute APIs. The GTX 650 Ti's best showing, 8229 in Vulkan, is still lower than the GTX 660's worst benchmark in the database, which is 11347 in OpenCL.

Specification Differences

The two cards share the same architecture, but the numerical differences in their specifications are substantial.

Memory: The GTX 660 has 2 GB of GDDR5, while the GTX 650 Ti has 1 GB. The bus width is 192-bit versus 128-bit, and bandwidth is 144.2 GB/s versus 86.40 GB/s. The GTX 660 also runs its memory at 6 Gbps effective, whereas the GTX 650 Ti runs at 5.4 Gbps.

Compute units: The GTX 660 has 960 shading units, 80 TMUs, and 24 ROPs. The GTX 650 Ti has 768 shading units, 64 TMUs, and 16 ROPs. The GTX 660 also leads in pixel rate (20.64 GPixel/s vs 14.85 GPixel/s), texture rate (82.56 GTexel/s vs 59.39 GTexel/s), and FP32 throughput (1.981 TFLOPS vs 1,425.4 GFLOPS).

Clocks: The GTX 660 has a base clock of 980 MHz and a boost clock of 1032 MHz. The GTX 650 Ti has no base or boost clock listed in the database, so the comparison is limited to memory clock, which is higher on the GTX 660.

Physical design: The GTX 660 is dual-slot with a 241 mm length (9.5 inches). The GTX 650 Ti is single-slot with a 145 mm length (5.7 inches). The GTX 660 has a 140 W TDP, the GTX 650 Ti a 110 W TDP. Both use a single 6-pin connector and a suggested 300 W PSU.

Display outputs: The GTX 660 offers 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2. The GTX 650 Ti offers 2x DVI and 1x mini-HDMI 1.4a, with no DisplayPort.

Release dates: The GTX 660 launched on 2012-09-05, the GTX 650 Ti on 2012-10-08. Both are end-of-life products. The launch MSRP for the GTX 660 is 229 USD; for the GTX 650 Ti it is 149 USD.

Architecture Differences

Both cards are built on the same Kepler architecture, use the same 28 nm process node at TSMC, and share the same 2,540 million transistor count on a 221 mm² die with a transistor density of 11.5M / mm². The fundamental silicon is identical in size and process.

The key architectural difference lies in the chip configuration. The GTX 660 uses the GK106 chip, while the GTX 650 Ti uses the GK106S. The "S" suffix indicates a cut-down die, and the specifications confirm this: the GTX 650 Ti has fewer shading units (768 vs 960), fewer TMUs (64 vs 80), and fewer ROPs (16 vs 24). This is a classic segmentation strategy within the same generation.

The GTX 660 has a higher memory clock (1502 MHz vs 1350 MHz) and a wider bus (192-bit vs 128-bit), which yields a larger bandwidth advantage than the raw clock difference alone. The GTX 660 also carries a base and boost clock, while the GTX 650 Ti has no listed base or boost clock.

Both cards support the same API levels: DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. Neither card has ray tracing or tensor cores. They share the same PCIe 3.0 x16 interface and the same suggested PSU of 300 W.

The GTX 660's larger TDP (140 W vs 110 W) is a direct consequence of the additional shader units, TMUs, and ROPs running at higher clocks. The GTX 650 Ti's reduced power draw and single-slot profile make it a lower-power, lower-performance sibling with the same basic architecture. The GTX 660 is the fully enabled GK106, while the GTX 650 Ti is the cut-down GK106S.