NVIDIA GeForce GTX 670M vs NVIDIA GeForce GTX 680M Comparison

The benchmark data clearly establishes the NVIDIA GeForce GTX 680M as the superior mobile graphics solution, decisively outperforming the GTX 670M in every measurable category. This is not a close contest; the 680M leads by a substantial margin in raw compute power, memory bandwidth, and architectural efficiency, making it the unequivocal choice for users prioritizing maximum performance in a laptop form factor.

Head-to-Head Benchmarks

The only directly comparable benchmark between the two GPUs is the Geekbench OpenCL test, and the results are starkly one-sided. The GTX 680M scores 9230 points, while the GTX 670M trails significantly with 6513 points. This represents a 41.7% performance advantage for the 680M, a massive gap that underscores the generational and architectural leap between the two parts. In practical terms, this means the 680M can handle significantly more demanding compute workloads, from complex 3D rendering tasks to GPU-accelerated video encoding, with far greater fluidity.

This single head-to-head result is reinforced by the average benchmark scores for each card. The GTX 680M achieves an average score of 7023, while the GTX 670M settles at 6513. The fact that the 680M's average is pulled down by its lower Geekbench Metal score of 4815 (compared to its OpenCL score of 9230) highlights that its peak performance in a compute-heavy API is exceptionally high. The 670M, lacking a Metal benchmark result, shows a more consistent but lower performance profile across its available tests.

Contextualizing these scores against their respective rivals further clarifies the performance hierarchy. The GTX 680M's average score of 7023 places it in a competitive position, trading blows with the NVIDIA T600, which scores 7035 (a -0.2% delta for the 680M). It is also just 1.9% behind the desktop-class NVIDIA GeForce GTX 970, which scores 7157. This indicates that the 680M, despite being a mobile part, delivers performance that is surprisingly close to a well-regarded desktop GPU from a similar era. In contrast, the GTX 670M's average score of 6513 places it in a much lower tier, competing with entry-level parts like the NVIDIA GeForce GT 555M (6493, a 0.3% delta) and the AMD Radeon Vega 10 Mobile (6476, a 0.6% delta). The 670M is even 0.6% behind the Intel UHD Graphics P750 (6554), an integrated solution, which underscores its position as a lower-mid-range performer.

The percentile ranking tells a similar story, albeit with a smaller margin than the raw scores suggest. The GTX 680M sits in the 39th percentile of all GPUs, while the GTX 670M is in the 38th percentile. While this difference seems narrow, it is crucial to remember that the percentile is based on the average score, which includes the 680M's relatively weak Metal result. The OpenCL test, which is a more direct measure of the card's compute capabilities, shows the 680M's true, much larger advantage.

Where Each One Wins

Given the data, the GTX 680M wins in virtually every scenario where raw performance is paramount. Its 41.7% lead in OpenCL compute makes it the superior choice for any GPU-accelerated application, including video editing, 3D modeling, scientific simulations, and demanding game physics. The 680M’s higher texture rate of 84.90 GTexel/s versus the 670M's 33.49 GTexel/s suggests it can handle high-resolution textures and complex visual scenes with much less performance degradation. Similarly, its pixel rate of 21.22 GPixel/s dwarfs the 670M's 8.372 GPixel/s, indicating a strong advantage in fill-rate-bound scenarios, such as high-resolution gaming with anti-aliasing enabled.

The GTX 670M, on the other hand, does not win any benchmark in the provided data. Its only potential advantage lies outside of performance: its lower TDP of 75 W compared to the 680M's 100 W. This means that laptops housing the 670M could potentially have slightly better battery life or produce less heat under sustained load, though this is a qualitative inference from the TDP figures, not a measured outcome in the data. For users on a stricter power budget, the 670M is the more efficient choice, but this comes at the cost of a massive performance deficit. The 670M is not a "winner" in any performance metric; it is merely a lower-power alternative that sacrifices a significant amount of capability.

Architecture Differences

The performance gulf between the two cards is rooted in fundamentally different architectures and manufacturing processes. The GTX 680M is built on the Kepler architecture, utilizing the GK104 chip, while the GTX 670M is based on the older Fermi 2.0 architecture with the GF114 chip. This architectural shift is a primary driver of the performance difference, as Kepler was designed to be significantly more efficient and powerful per clock than Fermi.

This efficiency is evident in the manufacturing process. The 680M is fabricated on a 28 nm process at TSMC, compared to the 670M's 40 nm process. This smaller node allows the 680M to pack 3,540 million transistors onto a 294 mm² die, resulting in a transistor density of 12.0M / mm². In contrast, the 670M has 1,950 million transistors on a larger 332 mm² die, yielding a much lower density of 5.9M / mm². The combination of a newer architecture and a denser, more advanced process gives the 680M a staggering advantage in raw computational resources.

This is most apparent in the sheer number of compute units. The GTX 680M features 1344 shading units, 112 texture mapping units (TMUs), and 32 render output units (ROPs). The GTX 670M, by comparison, is far more modest, with only 336 shaders, 56 TMUs, and 24 ROPs. The 680M has four times the shaders and double the TMUs, explaining its dominant lead in texture and pixel fill rates (84.90 vs 33.49 GTexel/s and 21.22 vs 8.372 GPixel/s, respectively). This translates to an FP32 compute throughput of 2.038 TFLOPS for the 680M, versus a mere 803.7 GFLOPS for the 670M.

Memory configurations also differ significantly, reinforcing the 680M's advantage. The 680M comes equipped with 4 GB of GDDR5 memory on a 256-bit bus, providing a bandwidth of 115.2 GB/s. The 670M, in contrast, has only 1536 MB of GDDR5 on a narrower 192-bit bus, yielding a substantially lower bandwidth of 72.00 GB/s. This wider memory interface and higher bandwidth allow the 680M to feed its more powerful GPU core far more effectively, especially at high resolutions and with large texture sets. Furthermore, the 680M has a base clock of 719 MHz and a boost clock of 758 MHz, while the 670M's core clocks are not listed in the data. The 680M's effective memory speed is also higher at 3.6 Gbps versus the 670M's 3 Gbps.

FAQ

Q: Which GPU is faster in compute workloads, the GTX 680M or the GTX 670M?

A: The GTX 680M is significantly faster. In the Geekbench OpenCL test, it scores 9230 compared to the GTX 670M's 6513, a 41.7% performance advantage.

Q: How does the memory bandwidth of the two cards compare?

A: The GTX 680M has a much higher memory bandwidth of 115.2 GB/s, thanks to its 256-bit bus and 4 GB of GDDR5 memory. The GTX 670M has a bandwidth of 72.00 GB/s from its 192-bit bus and 1536 MB of memory.

Q: What are the architectural differences between the two GPUs?

A: The GTX 680M uses the newer Kepler architecture on a 28 nm process, while the GTX 670M uses the older Fermi 2.0 architecture on a 40 nm process. The 680M also features 1344 shading units versus the 670M's 336.

Q: Which GPU is more power-efficient?

A: Based on TDP, the GTX 670M is more power-efficient, with a 75 W rating compared to the GTX 680M's 100 W. This could translate to longer battery life in a laptop.

Q: Does the GTX 670M support any newer APIs that the GTX 680M does not?

A: No. Both GPUs support DirectX 12 (11_0) and OpenGL 4.6. The GTX 680M also supports Vulkan 1.2.175, while the GTX 670M does not have a listed Vulkan version.

Q: How does the GTX 680M compare to the desktop NVIDIA GeForce GTX 970?

A: The GTX 680M's average benchmark score of 7023 is only 1.9% lower than the GTX 970's score of 7157, indicating that the mobile GPU is nearly as fast in this specific test.

The Verdict

The data unequivocally points to the NVIDIA GeForce GTX 680M as the superior GPU. It wins the only direct head-to-head benchmark by a commanding 41.7%, has a higher average benchmark score (7023 vs 6513), and offers dramatically more computational resources. Users who require the best possible gaming performance or need to accelerate demanding creative and professional workloads should choose the GTX 680M without hesitation. Its higher power draw is the only significant trade-off.

The GTX 670M, on the other hand, is the choice for users for whom power efficiency is the absolute priority. Its 75 W TDP is a clear advantage over the 680M's 100 W. However, this efficiency comes at the cost of a massive performance deficit. The data shows the 670M is not competitive with the 680M in any benchmark. It is a legacy part that offers basic acceleration for less demanding tasks, but it cannot deliver the high-end experience of the 680M. For any user who values performance, the GTX 680M is the definitive winner.

Specification Differences

| Specification | NVIDIA GeForce GTX 680M | NVIDIA GeForce GTX 670M |

| :--- | :--- | :--- |

| Chip | GK104 | GF114 |

| Architecture | Kepler | Fermi 2.0 |

| Process Node | 28 nm | 40 nm |

| Transistors | 3,540 million | 1,950 million |

| Die Size | 294 mm² | 332 mm² |

| Transistor Density | 12.0M / mm² | 5.9M / mm² |

| Base Clock | 719 MHz | Not Listed |

| Boost Clock | 758 MHz | Not Listed |

| Memory Size | 4 GB | 1536 MB |

| Memory Bus Width | 256 bit | 192 bit |

| Memory Bandwidth | 115.2 GB/s | 72.00 GB/s |

| Shading Units | 1344 | 336 |

| TMUs | 112 | 56 |

| ROPs | 32 | 24 |

| Pixel Rate | 21.22 GPixel/s | 8.372 GPixel/s |

| Texture Rate | 84.90 GTexel/s | 33.49 GTexel/s |

| FP32 Performance | 2.038 TFLOPS | 803.7 GFLOPS |

| TDP | 100 W | 75 W |

| Vulkan API | 1.2.175 | Not Listed |