AMD Radeon 660M vs NVIDIA GeForce GTX 680 Comparison

# NVIDIA GeForce GTX 680 vs AMD Radeon 660M

The data presents a clear performance hierarchy across the two available benchmark comparisons, with the NVIDIA GeForce GTX 680 holding a decisive lead in both OpenCL and Vulkan workloads. The GTX 680 achieves an average benchmark score of 14,150 against the Radeon 660M's 13,812, a difference of roughly 2.4% that understates its head-to-head margins. Both GPUs sit at the 55th percentile among all GPUs, placing them in the same overall performance tier despite their vastly different architectures, release dates, and design philosophies. The GTX 680, a 2012 desktop flagship, outperforms the 2022 integrated mobile part in every measured test, but the Radeon 660M delivers that near-comparable performance at a fraction of the power envelope — 40 W versus 195 W — which fundamentally changes how each should be evaluated.

The Verdict

For desktop users who already have a PCIe 3.0 slot and a 450 W power supply, the GTX 680 remains the superior choice purely on benchmark scores. It wins both head-to-head tests, with a 31.3% advantage in OpenCL (16,906 vs 12,876) and a 19.7% lead in Vulkan (17,646 vs 14,748). The GTX 680 also holds a higher average benchmark score of 14,150 versus 13,812, and it edges out the Radeon 660M's nearest rival comparison — the GTX 680 sits within 0.9% of the NVIDIA Tesla K10 (14,029), while the Radeon 660M sits within 0.1% of the NVIDIA RTX A2000 Mobile (13,821). None of these deltas exceed 1.5%, however, which means the two cards occupy essentially the same performance band in aggregate.

The Radeon 660M is the rational pick for compact or mobile systems, where its integrated design requires no power connectors, no slot width, and no dedicated cooling solution beyond what the laptop or small-form-factor chassis already provides. Its 40 W TDP is less than a quarter of the GTX 680's 195 W, and its PCIe 4.0 x8 interface offers a modern bus standard even though the GTX 680 uses PCIe 3.0 x16. The Radeon 660M also supports DirectX 12 Ultimate (12_2) and Vulkan 1.4, whereas the GTX 680 is limited to DirectX 12 (11_0) and Vulkan 1.2.175 — a meaningful feature gap for newer titles that leverage ray tracing, since the Radeon 660M includes 6 dedicated ray accelerators while the GTX 680 has none. For users prioritizing raw frame rates in legacy or non-ray-traced workloads, the GTX 680 wins. For users prioritizing portability, modern API support, and power efficiency, the Radeon 660M is the only sensible option.

Architecture Differences

The GTX 680 uses the GK104 chip built on NVIDIA's Kepler architecture, fabricated on a 28 nm TSMC process with 3,540 million transistors across a 294 mm² die. The Radeon 660M uses the Rembrandt chip based on RDNA 2.0, also fabricated by TSMC but on a far more advanced 6 nm node, packing 13,100 million transistors into a 208 mm² die. This yields a transistor density of 63.0M per mm² for the Radeon 660M versus 12.0M per mm² for the GTX 680 — a 5.25x density advantage that reflects the decade-long process technology gap between their 2012 and 2022 release dates.

Core counts differ substantially. The GTX 680 fields 1,536 shading units, 128 texture mapping units, and 32 ROPs. The Radeon 660M has 384 shading units, 24 TMUs, and 16 ROPs — exactly one-quarter the shading units, roughly one-fifth the TMUs, and half the ROPs. Despite these deficits, the Radeon 660M compensates with much higher clocks: a 1,500 MHz base and 1,900 MHz boost versus the GTX 680's 1,006 MHz base and 1,058 MHz boost. The Radeon 660M also includes 6 ray tracing cores, a feature entirely absent from the GTX 680's Kepler design.

Memory architecture is fundamentally different. The GTX 680 has a dedicated 2 GB GDDR5 frame buffer on a 256-bit bus, delivering 192.3 GB/s of bandwidth. The Radeon 660M uses system-shared memory, with its size, type, bus width, and bandwidth all listed as "System Shared" or "System Dependent." This makes the Radeon 660M's effective memory performance entirely contingent on the host system's RAM configuration, whereas the GTX 680's bandwidth is fixed and predictable.

The GTX 680's pixel rate is 33.86 GPixel/s and its texture rate is 135.4 GTexel/s, versus 30.40 GPixel/s and 45.60 GTexel/s for the Radeon 660M. The GTX 680's FP32 throughput is 3.250 TFLOPS, more than double the Radeon 660M's 1,459.2 GFLOPS. The Radeon 660M does offer FP16 compute at 2.918 TFLOPS (2:1 ratio), a feature the GTX 680 does not list. The GTX 680's memory clock is 1,502 MHz with 6 Gbps effective, while the Radeon 660M's memory clock is tied to the system.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA GeForce GTX 680 has an average benchmark score of 14,150, while the AMD Radeon 660M scores 13,812. Both GPUs rank at the 55th percentile among all GPUs.

Q: How large is the GTX 680's lead in OpenCL performance?

A: The GTX 680 scores 16,906 in Geekbench OpenCL versus 12,876 for the Radeon 660M, a 31.3% advantage. This is the largest delta between the two GPUs in any benchmark.

Q: Does the Radeon 660M support ray tracing?

A: Yes, the Radeon 660M includes 6 ray tracing cores as part of its RDNA 2.0 architecture. The GTX 680 has no ray tracing cores, as its Kepler architecture predates that feature.

Q: What are the power consumption figures for each GPU?

A: The GTX 680 has a TDP of 195 W and requires two 6-pin power connectors with a suggested 450 W power supply. The Radeon 660M has a TDP of 40 W, requires no power connectors, and has no suggested PSU rating.

Q: Which GPU supports newer graphics APIs?

A: The Radeon 660M supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The GTX 680 supports DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. Both share OpenGL 4.6, but the Radeon 660M has newer DirectX and Vulkan versions.

Q: How do the two GPUs compare in Vulkan performance?

A: The GTX 680 scores 17,646 in Geekbench Vulkan versus 14,748 for the Radeon 660M, a 19.7% advantage for the NVIDIA card.

Specification Differences

| Specification | NVIDIA GeForce GTX 680 | AMD Radeon 660M |

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

| Chip | GK104 | Rembrandt |

| Architecture | Kepler | RDNA 2.0 |

| Generation | GeForce 600 | Navi II IGP (Rembrandt Mobile) |

| 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 | 1006 MHz | 1500 MHz |

| Boost Clock | 1058 MHz | 1900 MHz |

| Memory Size | 2 GB | System Shared |

| Memory Type | GDDR5 | System Shared |

| Memory Bus Width | 256 bit | System Shared |

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

| Memory Clock | 1502 MHz (6 Gbps effective) | System Shared |

| Shading Units | 1536 | 384 |

| TMUs | 128 | 24 |

| ROPs | 32 | 16 |

| Ray Tracing Cores | None | 6 |

| Pixel Rate | 33.86 GPixel/s | 30.40 GPixel/s |

| Texture Rate | 135.4 GTexel/s | 45.60 GTexel/s |

| FP32 | 3.250 TFLOPS | 1,459.2 GFLOPS |

| FP16 | Not listed | 2.918 TFLOPS (2:1) |

| TDP | 195 W | 40 W |

| Slot Width | Dual-slot | IGP |

| Power Connectors | 2x 6-pin | None |

| Suggested PSU | 450 W | Not listed |

| Bus Interface | PCIe 3.0 x16 | PCIe 4.0 x8 |

| Display Outputs | 2x DVI, 1x HDMI 1.4a, 1x DisplayPort 1.2 | Portable Device Dependent |

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

| Vulkan | 1.2.175 | 1.4 |

| Release Date | 2012-03-21 | 2022-01-03 |

| Launch MSRP | 499 USD | Not listed |

Head-to-Head Benchmarks

The GTX 680 wins both available benchmark comparisons, but the margins tell a more nuanced story than a simple win-loss record. In Geekbench OpenCL, the GTX 680 posts 16,906 against the Radeon 660M's 12,876, a 31.3% delta that represents the largest performance gap between the two in any test. This result aligns with the GTX 680's raw compute advantages: its FP32 throughput of 3.250 TFLOPS is more than double the Radeon 660M's 1,459.2 GFLOPS, and its dedicated 192.3 GB/s memory bandwidth avoids the system-shared bottleneck that the Radeon 660M must contend with.

The Vulkan benchmark narrows the gap considerably. The GTX 680 scores 17,646 versus 14,748 for the Radeon 660M, a 19.7% advantage that still clearly favors NVIDIA but demonstrates the Radeon 660M's relative strength in modern API workloads. The Radeon 660M's newer architecture — RDNA 2.0 with Vulkan 1.4 support — likely narrows the efficiency gap in API-overhead-sensitive scenarios, even though its shading unit count (384) is a quarter of the GTX 680's (1,536).

Looking at the nearest rival data for context, the GTX 680's 14,150 average score places it between the NVIDIA Tesla K10 (14,029, +0.9% for the GTX 680) and the NVIDIA GeForce GTX 1070 Ti (14,277, -0.9%) and Intel Iris Xe MAX Graphics (14,315, -1.2%). The Radeon 660M's 13,812 average score puts it between the NVIDIA RTX A2000 Mobile (13,821, -0.1%) and the AMD Radeon RX 570X (13,871, -0.4%), with the AMD Radeon RX 7900 XT (13,745, +0.5%) and NVIDIA Tesla K10 (14,029, -1.5%) flanking the extremes. These tight deltas — all within 1.5% — indicate that both GPUs are firmly mid-pack performers, with the GTX 680 holding a slight edge in aggregate.

The GTX 680's dual wins in head-to-head testing, combined with its higher average score, make it the benchmark leader. The Radeon 660M's 40 W power draw and integrated form factor, however, represent a fundamentally different value proposition that no benchmark score can capture: it delivers 97.6% of the GTX 680's average score (13,812 vs 14,150) while consuming 79.5% less power (40 W vs 195 W) and requiring no additional cooling or power delivery hardware.