NVIDIA GeForce GTX 680 vs NVIDIA GeForce GTX 965M Comparison

The GeForce GTX 965M and the GeForce GTX 680 occupy opposite ends of the NVIDIA laptop-desktop spectrum, yet their benchmark averages land nearly on top of each other. The GTX 965M, a Maxwell 2.0 mobile part from the GeForce 900M generation, posts an average benchmark score of 14,404, placing it in the 56th percentile of all GPUs. The GTX 680, a Kepler desktop flagship from the GeForce 600 series, averages 14,150, sitting in the 55th percentile. The 254-point gap (roughly 1.8%) is statistically trivial, but the two GPUs achieve parity through radically different designs: one is a power-sipping MXM module with a 128-bit memory bus, the other a 195 W dual-slot card with a 256-bit interface and a launch MSRP of 499 USD. Benchmark results show the desktop card wins both direct comparisons, but the mobile chip's efficiency and modern feature set tell a more nuanced story.

Where Each One Wins

The head-to-head benchmark data is unambiguous: the GTX 680 wins both recorded tests. In Geekbench OpenCL, the GTX 680 scores 16,906 against the GTX 965M's 14,509, a delta of -14.2% from the mobile chip's perspective. In Geekbench Vulkan, the gap widens to -19%, with the GTX 680 scoring 17,646 versus 14,299. The desktop card's wins are substantial, representing a 16.5% advantage in OpenCL and a 23.4% advantage in Vulkan when measured from the GTX 965M's scores.

However, the GTX 965M's wins are not in raw performance but in its surroundings. The mobile chip's nearest rivals include the AMD Radeon RX Vega 11 (average score 14,385, delta 0.1%), the NVIDIA GeForce GTX TITAN (14,373, delta 0.2%), and the AMD Radeon Vega 11 (14,352, delta 0.4%). The GTX 680's nearest rivals include the NVIDIA Tesla K10 (14,029, delta 0.9%), the GTX 1070 Ti (14,277, delta -0.9%), and the Intel Iris Xe MAX Graphics (14,315, delta -1.2%). Interestingly, the GTX 965M sits slightly above the GTX 680's average, meaning the mobile part edges the desktop card in aggregate scoring despite losing the direct tests.

For use cases, the GTX 680 is the pick for raw compute throughput. Its FP32 performance of 3.250 TFLOPS dwarfs the GTX 965M's 1.946 TFLOPS, and its texture rate of 135.4 GTexel/s more than doubles the mobile chip's 60.80 GTexel/s. The desktop card also offers 1536 shading units versus 1024, and 128 TMUs versus 64. Memory bandwidth is another decisive win: 192.3 GB/s against 80.19 GB/s. The GTX 965M, by contrast, wins on API support, offering DirectX 12 (12_1) versus the GTX 680's DirectX 12 (11_0), and Vulkan 1.4 versus 1.2.175. The mobile chip also has a higher transistor density at 13.1M per mm² versus 12.0M per mm².

Architecture Differences

The two GPUs come from different architectural generations. The GTX 965M uses the GM204 chip built on Maxwell 2.0, while the GTX 680 uses the GK104 chip on the older Kepler architecture. Both are fabricated by TSMC on a 28 nm process, but the similarities end there. The GM204 packs 5,200 million transistors on a 398 mm² die, whereas the GK104 contains 3,540 million transistors on a 294 mm² die. That means the Maxwell chip has nearly 47% more transistors in a 35% larger die, resulting in a higher density of 13.1M transistors per mm² versus 12.0M.

The memory subsystems diverge sharply. The GTX 965M uses a 128-bit bus with 2 GB of GDDR5 running at 5 Gbps effective, yielding 80.19 GB/s of bandwidth. The GTX 680 uses a 256-bit bus with 2 GB of GDDR5 at 6 Gbps effective, producing 192.3 GB/s — a 2.4x advantage. Clock speeds tell a different story: the GTX 680 runs higher at 1006 MHz base and 1058 MHz boost, while the GTX 965M sits at 924 MHz base and 950 MHz boost. The mobile chip's lower clocks and narrower bus are partially offset by its higher transistor count and architectural efficiency.

The feature sets differ in API support. The GTX 965M supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The GTX 680 supports DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. The Maxwell part also benefits from a newer mobile-oriented design, using an MXM-B (3.0) bus interface and requiring no power connectors, while the Kepler card uses PCIe 3.0 x16 and demands two 6-pin connectors. The GTX 680's TDP is listed at 195 W with a suggested PSU of 450 W, whereas the GTX 965M has no TDP listed, reflecting its laptop-centric positioning.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The GTX 965M averages 14,404, which is 254 points higher than the GTX 680's 14,150. This places the mobile chip in the 56th percentile versus the desktop card's 55th.

Q: How does the GTX 680 perform in Vulkan compared to the GTX 965M?

A: In Geekbench Vulkan, the GTX 680 scores 17,646 against the GTX 965M's 14,299, a 19% advantage for the desktop card. This is the larger of the two direct benchmark wins.

Q: What are the memory bandwidth figures for each card?

A: The GTX 680 delivers 192.3 GB/s of bandwidth via a 256-bit bus and 6 Gbps GDDR5. The GTX 965M provides 80.19 GB/s through a 128-bit bus and 5 Gbps GDDR5, less than half the desktop card's throughput.

Q: Which GPU supports newer API versions?

A: The GTX 965M supports DirectX 12 (12_1) and Vulkan 1.4, while the GTX 680 supports DirectX 12 (11_0) and Vulkan 1.2.175. Both support OpenGL 4.6.

Q: Does the GTX 680 require additional power connectors?

A: Yes, the GTX 680 uses two 6-pin power connectors and has a 195 W TDP with a suggested 450 W PSU. The GTX 965M requires no power connectors and uses an MXM module slot.

Specification Differences

| Specification | NVIDIA GeForce GTX 965M | NVIDIA GeForce GTX 680 |

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

| Architecture | Maxwell 2.0 | Kepler |

| Chip | GM204 | GK104 |

| Transistors | 5,200 million | 3,540 million |

| Die Size | 398 mm² | 294 mm² |

| Transistor Density | 13.1M / mm² | 12.0M / mm² |

| Base Clock | 924 MHz | 1006 MHz |

| Boost Clock | 950 MHz | 1058 MHz |

| Memory Clock | 1253 MHz (5 Gbps effective) | 1502 MHz (6 Gbps effective) |

| Memory Bus Width | 128 bit | 256 bit |

| Memory Bandwidth | 80.19 GB/s | 192.3 GB/s |

| Shading Units | 1024 | 1536 |

| TMUs | 64 | 128 |

| Pixel Rate | 30.40 GPixel/s | 33.86 GPixel/s |

| Texture Rate | 60.80 GTexel/s | 135.4 GTexel/s |

| FP32 | 1.946 TFLOPS | 3.250 TFLOPS |

| TDP | Not listed | 195 W |

| Slot Width | MXM Module | Dual-slot |

| Power Connectors | None | 2x 6-pin |

| Suggested PSU | Not listed | 450 W |

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

| Display Outputs | Portable Device Dependent | 2x DVI, 1x HDMI 1.4a, 1x DisplayPort 1.2 |

| DirectX | 12 (12_1) | 12 (11_0) |

| Vulkan | 1.4 | 1.2.175 |

| Dimensions | Not listed | 256 mm x 111 mm x 38 mm |

| Release Date | 2015-01-08 | 2012-03-21 |

Head-to-Head Benchmarks

The two available benchmark comparisons both favor the GTX 680, but the margins differ meaningfully. In Geekbench OpenCL, the GTX 680 scores 16,906 against the GTX 965M's 14,509. The delta of -14.2% means the desktop card outperforms the mobile chip by roughly one-seventh. This result aligns with the raw compute specifications: the GTX 680's 3.250 TFLOPS of FP32 performance and 135.4 GTexel/s texture rate give it a substantial throughput advantage that OpenCL workloads can exploit.

In Geekbench Vulkan, the gap widens to -19%, with the GTX 680 scoring 17,646 versus 14,299. The larger delta suggests the Vulkan test leverages the GTX 680's higher memory bandwidth (192.3 GB/s versus 80.19 GB/s) and greater shading unit count (1536 versus 1024). Interestingly, the GTX 965M's newer Vulkan 1.4 API support does not translate into a performance win here — the older Kepler card's brute-force resources win out.

Despite losing both direct tests, the GTX 965M's average benchmark score of 14,404 is higher than the GTX 680's 14,150. This apparent contradiction stems from the benchmark sets: the GTX 965M has two recorded scores (14,509 and 14,299), while the GTX 680 has three, including a much lower Geekbench Metal score of 7,897. The GTX 680's Metal result drags down its average, even though its OpenCL and Vulkan scores are both higher than the GTX 965M's equivalents.

The GTX 680's nearest rival data shows it sits between the Tesla K10 (14,029, delta 0.9%) and the GTX 1070 Ti (14,277, delta -0.9%), with the Iris Xe MAX (14,315, delta -1.2%) and Radeon Vega 11 (14,352, delta -1.4%) close behind. The GTX 965M, meanwhile, edges out the RX Vega 11 (14,385, delta 0.1%), TITAN (14,373, delta 0.2%), and Radeon Vega 11 (14,352, delta 0.4%). In practical terms, both cards trade blows with the same cluster of integrated and entry-level discrete GPUs, making their overall standing nearly identical despite the desktop card's direct-test dominance.