NVIDIA GeForce GT 645M vs NVIDIA GeForce GT 755M Comparison

The NVIDIA GeForce GT 645M and NVIDIA GeForce GT 755M are two end-of-life mobile graphics chips from the same Kepler family, yet the benchmark data reveals they have distinct performance personalities. The GT 645M takes the Geekbench Metal test decisively, while the GT 755M dominates Geekbench OpenCL. This split is unusual for sibling GPUs sharing the same core configuration, and the underlying reasons lie in clock speeds and memory architecture.

Head-to-Head Benchmarks

The head-to-head results show a clean split between the two APIs. In Geekbench Metal, the GT 645M scores 5,679 points against the GT 755M’s 3,132 points. That is an 81.3% advantage for the older 600M-series part — a massive margin that suggests the Metal workload heavily favors the GT 645M’s specific clock and memory characteristics. The GT 755M does not just lose this test; it is outperformed by a substantial factor, making the Metal result the single largest differentiator between these two cards.

In Geekbench OpenCL, the tables turn completely. The GT 755M scores 4,934 points, while the GT 645M manages only 2,680 points. The delta percentage here is -45.7%, meaning the GT 645M trails by roughly 46% of the GT 755M’s score. This is a commanding win for the newer chip, and the margin is nearly as large as the GT 645M’s Metal victory, just in the opposite direction. OpenCL appears to reward the GT 755M’s higher clocks and faster memory bandwidth, while Metal seems to favor whatever the GT 645M does differently.

With one win each, the head-to-head record is tied at 1–1. The average benchmark score, however, favors the GT 645M overall. It posts an average of 4,411 points across its three available tests (adding Geekbench Vulkan at 4,875), whereas the GT 755M’s average is 4,033 from just two tests. This puts the GT 645M at the 26th percentile of all GPUs, while the GT 755M sits at the 24th percentile. The percentile gap is narrow, but the average score difference of 378 points is not trivial, given that the GT 755M lacks a Vulkan score to bolster its average.

Architecture Differences

Both chips come from the same GK107 silicon, fabricated on TSMC’s 28 nm process. The transistor count is identical at 1,270 million, and the die size is the same 118 mm², yielding a transistor density of 10.8M per mm². The core config is also unchanged: 384 shading units, 32 texture mapping units, and 16 raster output units. Neither chip has ray tracing or tensor cores, and both support DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175.

The differences begin with the clock speeds. The GT 755M runs at a base clock of 980 MHz with a boost of 1020 MHz, while the GT 645M is significantly slower at 709 MHz base and 780 MHz boost. That is a 271 MHz gap at base and a 240 MHz gap at boost — roughly 38% higher base clock for the GT 755M. This clock advantage directly feeds into the compute rates: the GT 755M delivers 783.4 GFLOPS of FP32 performance versus 599.0 GFLOPS for the GT 645M, a 31% lead. The pixel rate follows suit, with the GT 755M at 8.160 GPixel/s against 6.240 GPixel/s, and the texture rate at 32.64 GTexel/s versus 24.96 GTexel/s.

Memory is the other major divergence. The GT 645M pairs its 2 GB frame buffer with DDR3 memory running at 900 MHz (1800 Mbps effective) across a 128-bit bus, yielding 28.80 GB/s of bandwidth. The GT 755M uses GDDR5 memory at 1350 MHz (5.4 Gbps effective) on the same 128-bit bus, but the bandwidth jumps to 86.40 GB/s — exactly three times the GT 645M’s throughput. This memory bandwidth gap is enormous and likely explains the OpenCL result, where the GT 755M’s faster memory can feed the shaders more effectively.

Power and physical specs also differ. The GT 645M is rated at 32 W TDP and uses an IGP slot width, while the GT 755M draws 50 W and comes as an MXM Module. Both lack power connectors and use a PCIe 3.0 x16 interface, and display outputs are portable-device dependent for both. The GT 645M launched in September 2012 as part of the GeForce 600M generation, succeeding the GeForce 500M line, while the GT 755M arrived in June 2013 in the GeForce 700M generation, succeeding the 600M series.

Where Each One Wins

The GT 645M wins in Metal-based workloads. Its Geekbench Metal score of 5,679 is far ahead of the GT 755M’s 3,132, and this is not a marginal edge — it is a dominant performance in that specific API. For any application that relies on Metal, the GT 645M is clearly the stronger choice, despite its lower clocks and slower memory. The GT 645M also holds a higher average benchmark score (4,411 vs 4,033) and a better percentile ranking (26th vs 24th), suggesting that across the full suite of tests it runs, it performs more consistently.

The GT 755M wins in OpenCL compute. Its score of 4,934 crushes the GT 645M’s 2,680, a 45.7% deficit for the older chip. This is the GT 755M’s primary strength, and it aligns with the hardware: higher clocks and triple the memory bandwidth are exactly what OpenCL workloads tend to reward. The GT 755M also has a higher raw compute ceiling, with 783.4 GFLOPS and faster pixel and texture rates, so any workload that is purely throughput-bound should favor it.

In terms of architecture, the GT 755M is the more capable silicon on paper. It has the same core count but runs faster and moves data three times quicker. The GT 645M’s wins appear to be API-specific rather than general, which raises the question of whether its Metal score is an anomaly or a genuine strength of the older driver stack.

FAQ

Q: Which GPU has a higher average benchmark score?

A: The GT 645M has an average benchmark score of 4,411, while the GT 755M scores 4,033. This puts the GT 645M at the 26th percentile of all GPUs, versus the 24th percentile for the GT 755M.

Q: How much faster is the GT 755M in OpenCL?

A: The GT 755M scores 4,934 in Geekbench OpenCL, against 2,680 for the GT 645M. That is a 45.7% lead for the GT 755M.

Q: Why does the GT 645M win in Metal despite lower clocks?

A: The GT 645M scores 5,679 in Geekbench Metal versus 3,132 for the GT 755M, an 81.3% difference. The data does not explain the cause, but the result is consistent across the available Metal test.

Q: What memory type does each card use?

A: The GT 645M uses 2 GB of DDR3 memory with 28.80 GB/s bandwidth, while the GT 755M uses 2 GB of GDDR5 memory with 86.40 GB/s bandwidth on the same 128-bit bus.

Q: Are the core configurations identical?

A: Yes, both have 384 shading units, 32 TMUs, and 16 ROPs, with the same 1,270 million transistors and 118 mm² die size on TSMC’s 28 nm process.

Q: Which card has a higher TDP?

A: The GT 755M is rated at 50 W, while the GT 645M is rated at 32 W. The GT 755M also uses an MXM Module slot, whereas the GT 645M uses an IGP slot.

The Verdict

The data does not crown a single winner. The GT 645M takes the Metal benchmark with an 81.3% margin, and its average score is 378 points higher, placing it two percentile points above the GT 755M. Anyone prioritizing Metal-based applications should choose the GT 645M, as its 5,679 score dwarfs the GT 755M’s 3,132.

The GT 755M is the pick for OpenCL compute. Its 4,934 score is 45.7% ahead of the GT 645M, and its memory bandwidth of 86.40 GB/s is three times higher. The GT 755M also offers superior raw throughput in FP32 (783.4 GFLOPS), pixel fill (8.160 GPixel/s), and texture fill (32.64 GTexel/s), making it the better all-around compute engine for workloads that are not Metal-bound.

For general use, the GT 645M’s higher average score and better percentile are notable, but they come from a test suite that includes Vulkan, where the GT 755M has no score at all. The GT 755M’s TDP of 50 W versus 32 W is a meaningful trade-off for portability, but the extra power buys substantial clock and memory advantages. Ultimately, the choice hinges on the API in use: Metal favors the GT 645M, OpenCL favors the GT 755M, and the tied head-to-head record reflects that split.

Specification Differences

| Specification | NVIDIA GeForce GT 645M | NVIDIA GeForce GT 755M |

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

| Generation | GeForce 600M | GeForce 700M |

| Base clock | 709 MHz | 980 MHz |

| Boost clock | 780 MHz | 1020 MHz |

| Memory clock | 900 MHz (1800 Mbps effective) | 1350 MHz (5.4 Gbps effective) |

| Memory type | DDR3 | GDDR5 |

| Memory bandwidth | 28.80 GB/s | 86.40 GB/s |

| Pixel rate | 6.240 GPixel/s | 8.160 GPixel/s |

| Texture rate | 24.96 GTexel/s | 32.64 GTexel/s |

| FP32 performance | 599.0 GFLOPS | 783.4 GFLOPS |

| TDP | 32 W | 50 W |

| Slot width | IGP | MXM Module |

| Release date | 2012-09-30 | 2013-06-24 |

| Predecessor | GeForce 500M | GeForce 600M |

| Successor | GeForce 700M | GeForce 800M |

| Geekbench Metal | 5679 | 3132 |

| Geekbench OpenCL | 2680 | 4934 |

| Geekbench Vulkan | 4875 | N/A |

| Average benchmark score | 4411 | 4033 |

| Percentile vs all GPUs | 26 | 24 |