NVIDIA GeForce 945M vs NVIDIA GeForce GTX 675MX Comparison

# FAQ

Q: How do the two GPUs compare in raw benchmark performance?

A: The GeForce GTX 675MX holds a decisive edge. In the Geekbench OpenCL test, it scores 10,723 against 8,099 for the GeForce 945M, a 32.4% advantage. The GTX 675MX also has a Geekbench Metal score of 6,131, while no Metal result is available for the 945M.

Q: Which GPU has the higher average benchmark score and percentile ranking?

A: The GTX 675MX averages 8,427 across all benchmarks, placing it in the 43rd percentile of all GPUs. The 945M averages 8,099, sitting in the 42nd percentile. The gap is narrow in percentile terms, but the OpenCL delta is substantial.

Q: How does the GTX 675MX compare to its nearest rivals?

A: The GTX 675MX trails the AMD Radeon 880M by 0.1%, the NVIDIA GeForce MX330 by 0.4%, and the AMD Radeon HD 8870M by 0.4%, while leading the AMD Radeon R9 M375X by 1.2%. These are tight margins, indicating the 675MX sits in a competitive mid-range band.

Q: What about the 945M's competitive positioning?

A: The 945M is essentially neck-and-neck with its closest competitors: it trails the AMD Radeon R9 M360 by 0.4% and the NVIDIA GeForce GTX 950M by 0.4%, while leading the NVIDIA GRID K2 by 0.2% and the NVIDIA GeForce GTX 650 Ti Boost by 0.4%. The 945M is a solidly mid-pack mobile part.

Q: Which GPU is more power-efficient per the specifications?

A: The 945M has a lower thermal design power (TDP) at 75 W, compared to 100 W for the GTX 675MX. Despite the lower power draw and a smaller die, the 945M's FP32 throughput is actually slightly higher at 1,305.6 GFLOPS versus 1,255.7 GFLOPS for the 675MX.

Q: Are there differences in memory technology that affect performance?

A: Yes. The GTX 675MX uses 2 GB of GDDR5 over a 256-bit bus, yielding 115.2 GB/s of bandwidth. The 945M also has 2 GB, but it is DDR3 on a 128-bit bus, providing only 28.80 GB/s — a 75% reduction in memory bandwidth that explains much of the benchmark gap.

# Architecture Differences

The two GPUs represent different architectural generations from NVIDIA. The GeForce GTX 675MX is built on the Kepler architecture using the GK104 chip, while the GeForce 945M leverages the Maxwell architecture with the GM107 chip. Both are fabricated on TSMC's 28 nm process, and both use the same MXM-B (3.0) bus interface.

The chip-level differences are stark. The GK104 packs 3,540 million transistors on a 294 mm² die, resulting in a transistor density of 12.0 million per mm². By contrast, the GM107 contains 1,870 million transistors on a much smaller 148 mm² die, achieving a slightly higher density of 12.6 million per mm². This means the 675MX has nearly twice the transistor count, but spread across roughly double the silicon area.

The execution resources differ significantly. The GTX 675MX fields 960 shading units, 80 texture mapping units, and 32 raster output units. The 945M, meanwhile, offers 640 shading units, 40 TMUs, and 16 ROPs. The 675MX therefore has 50% more shading units, double the TMUs, and double the ROPs. This hardware advantage translates directly into higher pixel and texture rates: the 675MX produces 13.08 GPixel/s and 52.32 GTexel/s, while the 945M manages 16.32 GPixel/s and 40.80 GTexel/s.

Clock behavior also differs. The 945M has explicit base and boost clocks of 928 MHz and 1020 MHz, respectively. The 675MX lists no base or boost clock in the specifications. Interestingly, despite the 945M's higher pixel rate, the 675MX wins on texture throughput, suggesting that the Kepler part's extra TMUs overcome its lower fill-rate ceiling.

Both GPUs support DirectX 12 (11_0), OpenGL 4.6, and Vulkan, though the 675MX lists Vulkan 1.2.175 while the 945M supports Vulkan 1.4. Neither has ray tracing or tensor cores. The memory clock is the same for both at 900 MHz, though the effective data rate differs: 3.6 Gbps for the 675MX's GDDR5 versus 1.8 Gbps for the 945M's DDR3.

# Head-to-Head Benchmarks

The only direct comparison available is the Geekbench OpenCL test, and it is a one-sided affair. The GTX 675MX scores 10,723, while the 945M manages 8,099. That is a 32.4% advantage for the older Kepler chip. This is a substantial margin, especially considering the 945M has a higher FP32 compute rating (1,305.6 GFLOPS) than the 675MX (1,255.7 GFLOPS). The data suggests that raw compute throughput is not the limiting factor — rather, the memory subsystem likely is.

The 675MX's GDDR5 memory delivers 115.2 GB/s of bandwidth, exactly four times the 28.80 GB/s of the 945M's DDR3. For OpenCL workloads, which often involve large data transfers and memory-bound kernels, this 4x bandwidth advantage is the most plausible explanation for the 32.4% score gap. The 675MX also has a wider 256-bit memory bus versus 128-bit, and double the ROPs, which further aids memory-intensive operations.

There is no Metal benchmark for the 945M, so the 675MX's Metal score of 6,131 stands alone. This absence limits cross-GPU comparison on Apple's Metal API, but the OpenCL result gives a clear picture of relative compute capability.

The wins tally is 1 for the GTX 675MX and 0 for the 945M. There are no benchmark categories where the 945M outperforms its rival in the data. The 945M's only consolation is its slightly higher FP32 peak, but that theoretical advantage does not materialize in the actual OpenCL test.

# Specification Differences

| Specification | GTX 675MX | 945M |

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

| Architecture | Kepler | Maxwell |

| Chip | GK104 | GM107 |

| Generation | GeForce 600M | GeForce 900M |

| Transistors | 3,540 million | 1,870 million |

| Die Size | 294 mm² | 148 mm² |

| Transistor Density | 12.0M / mm² | 12.6M / mm² |

| Base Clock | — | 928 MHz |

| Boost Clock | — | 1020 MHz |

| Memory Type | GDDR5 | DDR3 |

| Memory Bus Width | 256 bit | 128 bit |

| Memory Bandwidth | 115.2 GB/s | 28.80 GB/s |

| Effective Memory Rate | 3.6 Gbps | 1.8 Gbps |

| Shading Units | 960 | 640 |

| TMUs | 80 | 40 |

| ROPs | 32 | 16 |

| Pixel Rate | 13.08 GPixel/s | 16.32 GPixel/s |

| Texture Rate | 52.32 GTexel/s | 40.80 GTexel/s |

| FP32 | 1,255.7 GFLOPS | 1,305.6 GFLOPS |

| TDP | 100 W | 75 W |

| Vulkan Version | 1.2.175 | 1.4 |

| Release Date | 2012-09-30 | 2015-10-26 |

| Predecessor | GeForce 500M | GeForce 800M |

| Successor | GeForce 700M | GeForce 10 Mobile |

Both GPUs share several traits: 2 GB memory capacity, 28 nm process, TSMC foundry, MXM-B (3.0) interface, MXM Module slot width, no power connectors, portable-device-dependent display outputs, DirectX 12 (11_0), OpenGL 4.6, and end-of-life production status. Neither has a launch MSRP listed.

# Where Each One Wins

The GTX 675MX wins in every benchmark category where data exists. Its 32.4% OpenCL lead is the headline result, and its Metal score of 6,131 adds another data point in its favor. The 675MX also wins on memory bandwidth by a factor of four, has double the TMUs and ROPs, and packs 50% more shading units. For any workload that is bandwidth-limited or fill-rate-heavy, the 675MX is the clear choice.

The 945M's advantages are more subtle. It has a lower TDP at 75 W versus 100 W, making it a better fit for thermally constrained laptops. Its FP32 peak is slightly higher (1,305.6 GFLOPS), and its pixel rate is 16.32 GPixel/s versus 13.08 GPixel/s for the 675MX. The 945M also supports a newer Vulkan version (1.4 vs 1.2.175), which could matter for future compatibility with Vulkan-based applications. Its smaller die size and lower transistor count suggest a more efficient design, and its higher transistor density indicates a more compact layout.

In practical terms, the 675MX is the performance pick for compute-heavy tasks like OpenCL acceleration, where its massive bandwidth advantage dominates. The 945M is the efficiency pick, offering comparable compute with less power draw and newer API support. For gaming, the 675MX's higher texture rate and memory bandwidth should deliver smoother frame pacing, though the 945M's higher pixel rate could help at lower resolutions.

# The Verdict

The data points to a clear performance hierarchy: the NVIDIA GeForce GTX 675MX is the faster GPU, period. Its 32.4% OpenCL lead over the 945M is decisive, and its memory subsystem is in a different class entirely. Anyone choosing between these two purely on benchmark scores should select the 675MX without hesitation.

However, the 945M is not without merit. It draws 25 W less power, runs cooler in theory, and has a higher FP32 peak. For users who prioritize battery life and thermal headroom over raw performance, the 945M makes sense. Its newer Vulkan support also gives it a slight future-proofing edge.

The 675MX's position among its rivals is interesting: it sits within 0.4% of three competitors (AMD Radeon 880M, NVIDIA GeForce MX330, AMD Radeon HD 8870M) and leads the AMD Radeon R9 M375X by 1.2%. This indicates it is a well-balanced mid-range part. The 945M is similarly competitive, sitting within 0.4% of four rivals.

The verdict depends on the use case. For compute tasks, content creation, or any workload that stresses memory bandwidth, the GTX 675MX is the only rational choice. For light gaming, general productivity, or ultra-portable laptops where power draw matters, the 945M offers a more efficient alternative. The 675MX is the performance winner; the 945M is the efficiency winner. Neither is obsolete, but they serve different masters. The data does not support calling the 945M a better GPU — it simply fills a different niche.