AMD Radeon 680M vs NVIDIA GeForce GTX 690 Comparison

Where Each One Wins

The benchmark data splits cleanly between these two very different designs. The AMD Radeon 680M wins both recorded head-to-head tests, and it does so by a substantial margin. In Geekbench OpenCL, the 680M scores 23468 against the GTX 690's 17399, a 25.9% advantage. In Geekbench Vulkan, the 680M scores 21965 against 16675, a 24.1% lead. That is a clean sweep in the database's direct comparison.

However, the GTX 690 holds a higher overall percentile ranking. It sits at the 60th percentile of all GPUs, while the 680M sits at the 57th. The GTX 690's average benchmark score of 17037 is also higher than the 680M's 15270. This is an unusual split: the newer integrated part wins the individual compute tests, but the older discrete card places higher in the aggregate database ranking. The explanation lies in the test set. The GTX 690 only has Geekbench OpenCL and Vulkan entries, while the 680M also has a 3DMark Steel Nomad DX12 result of 378, which drags its average down. For pure compute workloads in OpenCL and Vulkan, the 680M is clearly ahead. For the GTX 690, those two Geekbench tests are the only recorded data points, so its average reflects only its stronger tests.

The practical takeaway: if your workload is OpenCL or Vulkan compute, the 680M wins decisively. If you are looking at the broader database ranking across mixed workloads, the GTX 690 appears stronger, but that is partly because its weak results are not recorded. The 680M is the more consistently modern performer in the tests that both cards actually share.

Architecture Differences

These two GPUs come from different eras and different design philosophies. The NVIDIA GeForce GTX 690 uses the GK104 chip, built on Kepler architecture, on a 28 nm process at TSMC. It packs 3,540 million transistors into a 294 mm² die, giving a transistor density of 12.0M per mm². The AMD Radeon 680M uses the Rembrandt+ chip, built on RDNA 2.0 architecture, on a 6 nm process, also at TSMC. It contains 13,100 million transistors in a 208 mm² die, for a density of 63.0M per mm². The density gap is enormous: the 680M crams more than five times as many transistors per square millimeter.

The GTX 690 is a dual-GPU card, which explains its physical footprint. It is 279 mm long, 111 mm tall, and 38 mm wide, a dual-slot design requiring 2x 8-pin power connectors and a 700 W suggested PSU. The 680M is an integrated graphics processor, an IGP with no slot width, no power connectors, and no published dimensions. Its memory is system shared, with bandwidth described as system dependent, whereas the GTX 690 has 2 GB of dedicated GDDR5 on a 256-bit bus with 192.3 GB/s of bandwidth.

Clock speeds also tell the story of the generational leap. The GTX 690 runs at a 915 MHz base and 1019 MHz boost, with memory at 1502 MHz (6 Gbps effective). The 680M runs at 2000 MHz base and 2200 MHz boost, more than double the base clock. The 680M also has 12 ray tracing cores, a feature the GTX 690 lacks entirely. The GTX 690 supports DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. The 680M supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The 680M also has FP16 capability at 6.758 TFLOPS (2:1), while the GTX 690 has no recorded FP16 throughput.

The power envelope is the most striking difference. The GTX 690 is rated at 300 W TDP. The 680M is rated at 50 W. That is a sixfold difference in power draw, and the 680M still wins the shared compute benchmarks. The GTX 690 was released on 2012-05-02 and is end-of-life, with the GeForce 700 series as its successor. The 680M was released on 2023-01-02, is still active, and follows the Vega II IGP. The GTX 690 had a launch MSRP of 999 USD; the 680M has no recorded launch MSRP because it ships as part of a mobile processor.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA GeForce GTX 690 has an average benchmark score of 17037, compared to the AMD Radeon 680M's 15270. The GTX 690 sits at the 60th percentile of all GPUs, while the 680M sits at the 57th.

Q: Why does the 680M win the head-to-head tests if its average score is lower?

A: The head-to-head tests only cover Geekbench OpenCL and Vulkan, where the 680M wins by 25.9% and 24.1% respectively. The 680M also has a 3DMark Steel Nomad DX12 result of 378, which is included in its average but not in the shared head-to-head set. That additional weak result lowers the 680M's overall average.

Q: How do the two compare in raw compute throughput?

A: The 680M has 3.379 TFLOPS FP32, slightly higher than the GTX 690's 3.130 TFLOPS. The 680M also provides 6.758 TFLOPS FP16 (2:1), while the GTX 690 has no recorded FP16 figure. The 680M has a higher pixel rate at 70.40 GPixel/s versus 32.61 GPixel/s, but the GTX 690 has a higher texture rate at 130.4 GTexel/s versus 105.6 GTexel/s.

Q: Which card has more shading units?

A: The GTX 690 has 1536 shading units, 128 TMUs, and 32 ROPs. The 680M has 768 shading units, 48 TMUs, and 32 ROPs. The GTX 690 has double the shading units and nearly triple the TMUs, yet the 680M still wins the recorded compute tests.

Q: What are the power requirements for each?

A: The GTX 690 has a 300 W TDP, requires 2x 8-pin power connectors, and needs a 700 W suggested PSU. The 680M has a 50 W TDP, no power connectors, and no suggested PSU, as it is an integrated processor.

Q: Which GPU supports newer graphics APIs?

A: The AMD Radeon 680M supports DirectX 12 Ultimate (12_2) and Vulkan 1.4. The NVIDIA GeForce GTX 690 supports DirectX 12 (11_0) and Vulkan 1.2.175. Both support OpenGL 4.6. The 680M also includes 12 ray tracing cores, which the GTX 690 does not have.

Specification Differences

| Specification | NVIDIA GeForce GTX 690 | AMD Radeon 680M |

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

| Architecture | Kepler | RDNA 2.0 |

| Process node | 28 nm | 6 nm |

| Transistors | 3,540 million | 13,100 million |

| Die size | 294 mm² | 208 mm² |

| Transistor density | 12.0M / mm² | 63.0M / mm² |

| Base clock | 915 MHz | 2000 MHz |

| Boost clock | 1019 MHz | 2200 MHz |

| Memory | 2 GB GDDR5 | System Shared |

| Memory bus | 256 bit | System Shared |

| Memory bandwidth | 192.3 GB/s | System Dependent |

| Shading units | 1536 | 768 |

| TMUs | 128 | 48 |

| ROPs | 32 | 32 |

| RT cores | None | 12 |

| Pixel rate | 32.61 GPixel/s | 70.40 GPixel/s |

| Texture rate | 130.4 GTexel/s | 105.6 GTexel/s |

| FP32 | 3.130 TFLOPS | 3.379 TFLOPS |

| FP16 | None recorded | 6.758 TFLOPS (2:1 |

| TDP | 300 W | 50 W |

| Slot width | Dual-slot | IGP |

| Power connectors | 2x 8-pin | None |

| Suggested PSU | 700 W | None |

| Bus interface | PCIe 3.0 x16 | PCIe 4.0 x8 |

| DirectX | 12 (11_0) | 12 Ultimate (12_2) |

| Vulkan | 1.2.175 | 1.4 |

| Release date | 2012-05-02 | 2023-01-02 |

| Production status | End-of-life | Active |

| Launch MSRP | 999 USD | None |

Head-to-Head Benchmarks

The only two tests shared between these GPUs in the database are Geekbench OpenCL and Geekbench Vulkan. The AMD Radeon 680M wins both, and the margins are consistent.

In Geekbench OpenCL, the 680M scores 23468 against the GTX 690's 17399. That is a 25.9% lead for the 680M. This is a significant gap in a compute workload that stresses raw throughput and memory handling. The 680M achieves this despite having half the shading units (768 versus 1536) and fewer TMUs (48 versus 128). Its advantage comes from much higher clocks (2200 MHz boost versus 1019 MHz) and a more modern architecture. The FP32 throughput numbers back this up: 3.379 TFLOPS for the 680M versus 3.130 TFLOPS for the GTX 690. The 680M is doing more work per clock per shader, and the clock speed advantage is enormous.

In Geekbench Vulkan, the 680M scores 21965 against 16675. That is a 24.1% lead. Vulkan is a lower-level API, and the 680M's support for Vulkan 1.4 (versus the GTX 690's 1.2.175) likely contributes to its efficiency here. The 680M's 12 ray tracing cores also indicate a hardware feature set the GTX 690 simply does not have, even if the Geekbench Vulkan test does not directly measure ray tracing performance.

For context on the GTX 690's standing, its nearest rivals in the database include the AMD Radeon HD 7970M (average score 17019, a 0.1% difference), the AMD Radeon RX 7600 XT (17083, 0.3% behind), and the NVIDIA GeForce RTX 3070 (17208, 1% behind). The GTX 690's average of 17037 places it right in that cluster. The 680M's nearest rivals include the NVIDIA GeForce GTX 580 (15283, 0.1% ahead of the 680M), the NVIDIA GeForce RTX 2060 (15290, 0.1% ahead), and the AMD Radeon RX 7600 (15171, 0.7% behind). So the 680M's average score puts it in the same performance tier as a GTX 580 or RTX 2060, while the GTX 690's average puts it near an RTX 3070. The inconsistency between the head-to-head wins and the overall ranking is entirely explained by the 680M's additional 3DMark Steel Nomad DX12 score of 378, which pulls down its average but does not appear in the shared tests.

The GTX 690's dual-GPU nature means its 300 W TDP covers two GK104 dies, each with 2 GB of GDDR5 memory. The 680M's system shared memory bandwidth is listed as system dependent, meaning it varies with the host platform. In a laptop context, that memory bandwidth could be a limiting factor for the 680M, yet it still beats the GTX 690's dedicated 192.3 GB/s bandwidth in the recorded compute tests.

The Verdict

The AMD Radeon 680M is the right choice if you need modern compute performance in a low-power package. The data shows it wins both shared benchmarks by roughly 25%, supports DirectX 12 Ultimate and Vulkan 1.4, includes ray tracing cores, and does it all at a 50 W TDP. The 680M is an integrated part with no power connectors and no slot requirement, making it suitable for compact or portable systems. Its 6 nm process and 63.0M / mm² transistor density represent a far more modern design than the GTX 690's 28 nm node.

The NVIDIA GeForce GTX 690 is the right choice if you are looking at the aggregate database ranking and want the higher percentile placement. It sits at the 60th percentile versus the 680M's 57th, and its average score of 17037 beats the 680M's 15270. It also has more shading units, more TMUs, and dedicated GDDR5 memory with a 256-bit bus. That said, it is a 300 W dual-slot card from 2012 that is end-of-life, and it loses every shared benchmark to the 680M. The GTX 690's launch MSRP was 999 USD, which reflects its original flagship positioning, but the recorded data does not show it outperforming the 680M in any direct test.

The verdict from the numbers is straightforward: the 680M is the better modern compute part. The GTX 690 holds a higher database percentile, but that ranking is based on a narrower set of tests and does not include the 680M's weaker 3DMark result. If your workload is OpenCL or Vulkan compute, the 680M wins by a wide margin. If you care about the overall database average across all recorded tests, the GTX 690 places higher, but that advantage is a statistical artifact of the test set rather than a demonstration of superior performance. For a new build, the 680M's active production status, lower power draw, and newer API support make it the practical pick. For a vintage system or a collector, the GTX 690 remains a historical flagship, but the benchmark data does not support choosing it for compute performance.