NVIDIA GeForce GTX 680M vs NVIDIA RTX A400 Comparison

# NVIDIA GeForce GTX 680M vs NVIDIA RTX A400: A Generational Leap in Mobile Graphics

The database comparison between the NVIDIA GeForce GTX 680M and the NVIDIA RTX A400 showcases a fascinating intersection of mobile and workstation graphics. The GTX 680M, released in 2012, was a high-end mobile part built on the Kepler architecture, while the RTX A400, arriving in 2024, is a modern low-profile workstation card based on Ampere. The gap between them is not merely a matter of clock speeds, but a fundamental shift in architecture, process node, and feature set. Benchmark results place the RTX A400 an astonishing 59.6% ahead of the GTX 680M in the OpenCL test, a lead that reflects the massive technological progress of the intervening years.

FAQ

Q: Which GPU is faster in the geekbench_opencl benchmark?

A: The NVIDIA RTX A400 wins decisively. It scores 22,844 points, while the GeForce GTX 680M scores 9,230. This represents a 59.6% performance advantage for the newer card.

Q: What is the average benchmark score for each GPU?

A: The GeForce GTX 680M has an average benchmark score of 7,023, placing it in the 39th percentile of all GPUs. The RTX A400, with an average score of 6,078, sits in the 35th percentile. Interestingly, the GTX 680M's average is actually slightly higher than the RTX A400's average in the database.

Q: Does the RTX A400 support the latest APIs?

A: Yes. The RTX A400 supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The GTX 680M is limited to DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175.

Q: What are the key differences in memory technology?

A: The GTX 680M uses 4 GB of GDDR5 on a 256-bit bus, delivering 115.2 GB/s of bandwidth. The RTX A400 moves to 4 GB of GDDR6 on a 64-bit bus, achieving 96.00 GB/s. While the newer card has less total bandwidth, the GDDR6 memory operates at a higher frequency.

Q: Which card has a higher boost clock?

A: The RTX A400 boosts to 1762 MHz, significantly higher than the GTX 680M's 758 MHz boost. The base clock is also higher on the RTX card at 1417 MHz versus 719 MHz.

Architecture Differences

The two cards represent a major architectural leap. The GeForce GTX 680M is built on the Kepler architecture with the GK104 chip, manufactured on a 28nm process at TSMC. This is a large die: 294 mm², housing 3,540 million transistors, resulting in a density of 12.0 million per mm². The card uses an MXM-B (3.0) interface and gets power from the MXM slot, requiring no additional power connectors.

In contrast, the RTX A400 is a radically different design. It is based on the Ampere architecture, using the GA107 chip, and is fabricated on Samsung's 8nm process. This allows for a much denser package: 8,700 million transistors on a 200 mm² die, a density of 43.5M per mm². The A400 is a low-profile, single-slot card, powered entirely by its PCIe 4.0 x8 bus, with no external power connectors. The recommended PSU is 250W.

The compute core of the A400 is also fundamentally different. It includes 768 shading units, 24 TMUs, and 16 ROPs. It does include 6 RT cores and 24 tensor cores, features the older card completely lacks. The A400's pixel rate is listed at 28.19 GPixel/s and its texture rate is 42.29 GTexel/s.

Head-to-Head Benchmarks

The best recorded performance for the GTX 680M comes from the Geekbench OpenCL test, where it scores 9,230. In this same test, the RTX A400 posts a score of 22,844, a 59.6% lead. That is the single biggest win for the newer card in the database.

However, the average figures tell a different and more nuanced story. The GTX 680M has a higher average benchmark score of 7,023 across a mix of legacy tests, including Passmark's DX9, DX10, DX11, DX12, G2D, G3D, GPU Compute, and the older Geekbench suite. The A400's average score of 6,078 is actually 1.2% lower than that of the GTX 680M. This is primarily due to the inclusion of legacy DirectX benchmarks where older architectures hold an advantage.

The RTX A400 wins a single head-to-head benchmark in the database, taking the Geekbench OpenCL test. The GTX 680M does not win any of the recorded head-to-head comparisons.

Specification Differences

| Field | GeForce GTX 680M | RTX A400 |

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

| Chip | GK104 | GA107 |

| Architecture | Kepler | Ampere |

| Process | 28 nm | 8 nm |

| Transistors | 3,540 million | 8,700 million |

| Die size | 294 mm² | 200 mm² |

| Transistor density | 12.0M/mm² | 43.5M/mm² |

| Base clock | 719 MHz | 1417 MHz |

| Boost clock | 758 MHz | 1762 MHz |

| Memory clock | 900 MHz (3.6 Gbps effective) | 1500 MHz (12 Gbps effective) |

| Memory | 4GB GDDR4 | 4GB GDDR6 |

| Bus width | 256-bit | 64-bit |

| Bandwidth | 115.2 GB/s | 96.00 GB/s |

| Shading units | 1344 | 768 |

| TMUs | 112 | 24 |

| ROPs | 32 | 16 |

| Pixel rate | 21.22 GPixel/s | 28.19 GPixel/s |

| Texture rate | 84.90 GTexel/s | 42.29 GTexel/s |

| FP32 | 2.038 TFLOPS | 2.706 TFLOPS |

| TDP | 100 W | 50 W |

| Bus interface | MXM-B (3.0) | PCIe 4.0 x8 |

| Display outputs | Portable Device Dependent | 4x mini-DisplayPort 1.4a |

| Release date | 2012-06-03 | 2024-04-15 |

Where Each One Wins

Choose the NVIDIA RTX A400 if the workload targets modern APIs and compute features that the older GPU cannot handle. Its support for DirectX 12 Ultimate, Vulkan 1.4, and OpenGL 4.6, combined with its RT and tensor cores, makes it the choice for contemporary professional workloads. It also takes the crown for OpenCL performance, delivering 59.6% more performance than the GTX 680M. The card also has a significantly higher pixel rate and a much better power efficiency profile: the GTX 680M's 100W is cut in half to 50W.

Choose the GeForce GTX 680M for general 3D rendering and legacy API compatibility. Its average benchmark score of 7,023 is 15.7% higher than the A400's average. It provides a wider 256-bit memory bus, which translates to a higher memory bandwidth of 115.2 GB/s, and it is the only one of the two that supports the older MXM-B slot. For users with flexibility in display outputs, the GTX 680M is the most flexible, with a device-dependent display output.