NVIDIA GeForce GTX 650 Ti vs NVIDIA GeForce GTX 675M Comparison

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA GeForce GTX 650 Ti records an average benchmark score of 8053, while the NVIDIA GeForce GTX 675M scores 6946. The 650 Ti sits at the 42nd percentile among all GPUs, compared to the 675M's 39th percentile.

Q: How do the two compare in the only shared benchmark test?

A: In Geekbench OpenCL, the GTX 650 Ti scores 7877 versus the GTX 675M's 6946, giving the 650 Ti a 13.4% advantage. This is the sole head-to-head test recorded in the database.

Q: What are the nearest rivals for each card based on average score?

A: For the GTX 650 Ti, the nearest rivals are the GTX 880M (8040, 0.2% higher), GTX 650 Ti Boost (8067, 0.2% lower), Quadro P5000 (8039, 0.2% higher), and GRID K2 (8080, 0.3% lower). For the GTX 675M, they are the Radeon R5 M240 (6975, 0.4% lower), GTX 680M (7023, 1.1% lower), T600 (7035, 1.3% lower), and FirePro M5100 (6830, 1.7% higher).

Q: Which GPU has a wider memory bus?

A: The GTX 675M uses a 256-bit bus with 2 GB of GDDR5, while the GTX 650 Ti uses a 128-bit bus with 1 GB. The 675M's wider bus contributes to its higher memory bandwidth of 96.00 GB/s versus 86.40 GB/s for the 650 Ti.

Q: What architecture does each card use?

A: The GTX 650 Ti is built on Kepler (chip GK106S), while the GTX 675M uses Fermi 2.0 (chip GF114). This architectural split affects transistor counts, die sizes, and feature support.

Q: Which card supports Vulkan?

A: Only the GTX 650 Ti lists Vulkan support at version 1.2.175. The GTX 675M has no Vulkan entry in the database, though both support DirectX 12 (11_0) and OpenGL 4.6.

Where Each One Wins

The recorded data shows a clear split based on workload type. The GTX 650 Ti wins the only direct benchmark comparison, but the GTX 675M holds advantages in specific memory and throughput characteristics that matter for certain tasks.

GTX 650 Ti strengths:

• Higher raw compute: FP32 performance reaches 1,425.4 GFLOPS versus 952.3 GFLOPS for the 675M. This represents a 49.7% advantage in shader throughput, which directly impacts compute-heavy workloads such as physics simulations or rendering tasks that rely on parallel floating-point operations.
• Superior pixel and texture rates: The 650 Ti delivers 14.85 GPixel/s and 59.39 GTexel/s, compared to 9.920 GPixel/s and 39.68 GTexel/s for the 675M. These figures favor the 650 Ti in fill-rate-bound scenarios like high-resolution texture streaming or anti-aliasing.
• Higher memory clock: The 650 Ti's memory runs at 1350 MHz (5.4 Gbps effective) versus 750 MHz (3 Gbps effective) for the 675M. Faster memory clock helps latency-sensitive workloads despite the narrower bus.

GTX 675M strengths:

• Larger memory capacity: 2 GB versus 1 GB, which allows larger datasets or textures to reside locally without spilling to system memory.
• Wider memory bus: 256-bit versus 128-bit, resulting in higher total bandwidth (96.00 GB/s versus 86.40 GB/s) despite the lower clock. This favors bandwidth-bound tasks such as large matrix operations or video processing.
• More ROPs: 32 versus 16, which can improve performance in scenarios with heavy blending or depth buffer operations, even though the lower pixel rate suggests the 650 Ti's faster clock compensates in many cases.
• Lower TDP: 100 W versus 110 W, making it a more power-efficient option for sustained loads in constrained thermal environments.

Architecture Differences

The GTX 650 Ti uses the Kepler architecture on a 28 nm TSMC process. This node allows a transistor count of 2,540 million on a die size of 221 mm², yielding a transistor density of 11.5M per mm². Kepler introduces a more efficient shader scheduling design compared to older generations, which helps explain the 650 Ti's higher FP32 output despite having 768 shading units.

The GTX 675M uses the older Fermi 2.0 architecture on a 40 nm TSMC process. It packs 1,950 million transistors into a larger die of 332 mm², resulting in a lower density of 5.9M per mm². Fermi's design uses 384 shading units but achieves lower peak compute. The architectural gap is also visible in feature support: Kepler adds Vulkan 1.2.175 support, whereas the 675M has no Vulkan entry in the database.

Both chips share the same TMU count at 64, but the ROP count differs (16 for 650 Ti, 32 for 675M). The 650 Ti compensates with higher clock-driven rates. The 675M is designed as a mobile MXM module with no power connectors, while the 650 Ti is a single-slot desktop card requiring a 6-pin connector. Display outputs also differ: the 650 Ti offers 2x DVI and 1x mini-HDMI 1.4a, while the 675M is dependent on the portable device's own outputs.

Specification Differences

| Field | GTX 650 Ti | GTX 675M |

|---|---|---|

| Architecture | Kepler | Fermi 2.0 |

| Process node | 28 nm | 40 nm |

| Transistors | 2,540 million | 1,950 million |

| Die size | 221 mm² | 332 mm² |

| Transistor density | 11.5M / mm² | 5.9M / mm² |

| Memory size | 1024 MB | 2 GB |

| Memory clock | 1350 MHz (5.4 Gbps effective) | 750 MHz (3 Gbps effective) |

| Bus width | 128 bit | 256 bit |

| Bandwidth | 86.40 GB/s | 96.00 GB/s |

| Shading units | 768 | 384 |

| ROPs | 16 | 32 |

| Pixel rate | 14.85 GPixel/s | 9.920 GPixel/s |

| Texture rate | 59.39 GTexel/s | 39.68 GTexel/s |

| FP32 | 1,425.4 GFLOPS | 952.3 GFLOPS |

| TDP | 110 W | 100 W |

| Slot width | Single-slot | MXM Module |

| Power connectors | 1x 6-pin | None |

| Bus interface | PCIe 3.0 x16 | MXM-B (3.0) |

| Display outputs | 2x DVI, 1x mini-HDMI 1.4a | Portable Device Dependent |

| Vulkan | 1.2.175 | None |

| Launch MSRP | 149 USD | Not listed |

The table shows that the 650 Ti leads in compute and fill rates, while the 675M leads in memory capacity, bus width, bandwidth, and ROP count. The 650 Ti also has a higher TDP by 10 W, which correlates with its higher peak performance.

Head-to-Head Benchmarks

The database records only one shared benchmark test: Geekbench OpenCL. In that test, the GTX 650 Ti scores 7877 against the GTX 675M's 6946. The delta is 13.4% in favor of the 650 Ti. This is a decisive win, but the margin is modest enough to suggest the 675M is not far behind in compute workloads that are memory-sensitive.

Looking at the nearest rivals for each card provides additional context. The 650 Ti's average score of 8053 places it within 0.3% of the GRID K2 and GTX 650 Ti Boost, which are slightly ahead or behind by fractions of a percent. The 675M's average of 6946 sits 1.7% above the FirePro M5100 but 1.3% below the T600. These clusters show that the 650 Ti competes with higher-end workstation and mobile parts, while the 675M sits closer to entry-level professional GPUs.

In the Geekbench OpenCL test, the 650 Ti's higher FP32 throughput (1,425.4 GFLOPS) likely drives its win, as OpenCL workloads often scale with shader compute. The 675M's wider memory bus (256-bit) and higher bandwidth (96.00 GB/s) do not overcome the 650 Ti's compute advantage in this specific test. The data does not include a Vulkan comparison, but the 650 Ti's Vulkan capability suggests it may also win in Vulkan-based workloads if tested.

The Verdict

The GTX 650 Ti is the stronger performer in the recorded benchmark data. It wins the only head-to-head test by 13.4%, holds a higher average score (8053 versus 6946), and reaches a higher percentile rank (42nd versus 39th). Its Kepler architecture delivers superior compute density and fill rates, which are critical for most graphics and compute tasks.

The GTX 675M is not without merit. Its 2 GB memory capacity and 256-bit bus provide 11.1% higher bandwidth than the 650 Ti. Users who work with very large textures or datasets may find the 675M more practical, especially if memory capacity is a bottleneck. Its lower TDP of 100 W also makes it an easier fit for thin laptops or systems with limited cooling.

For desktop users, the 650 Ti is the obvious choice based on raw performance. For laptop users with a fixed MXM slot, the 675M is the only option of the two, and its memory advantages are notable. The data suggests that the 650 Ti is the more capable GPU overall, but the 675M's niche lies in bandwidth-heavy and capacity-sensitive scenarios. The 13.4% gap in OpenCL is significant but not overwhelming, meaning the 675M remains competitive in specific workloads. Ultimately, the 650 Ti wins on compute, while the 675M wins on memory configuration. Choose based on whether shader throughput or memory capacity matters more for the intended application.