AMD Radeon Pro 5600M vs NVIDIA GeForce GTX 660 Ti Comparison

The AMD Radeon Pro 5600M and NVIDIA GeForce GTX 660 Ti represent two very different eras of GPU design, and the benchmark data reflects that divide clearly. The 5600M is a modern mobile professional part built on a 7nm process, while the GTX 660 Ti is a desktop GPU from 2012 built on 28nm. In the two shared benchmarks available, the Radeon Pro 5600M dominates by a wide margin, but the GTX 660 Ti still holds relevance in legacy DirectX workloads where its older architecture remains functional. This analysis breaks down where each card wins, what separates their architectures, and who should consider each one based strictly on the data provided.

Where Each One Wins

The AMD Radeon Pro 5600M wins decisively in modern compute and cross-platform graphics APIs. In Geekbench OpenCL, it scores 47,783 against the GTX 660 Ti’s 15,113, a 216.2% advantage. In Geekbench Vulkan, it scores 46,425 against 15,012, a 209.3% lead. These are not marginal wins—they are generational leaps. The 5600M also has a broader benchmark footprint, with scores in PassMark DirectX 10, 11, 12, and 9, plus G2D and G3D tests, while the GTX 660 Ti only has two recorded Geekbench results. This means the 5600M can be evaluated across a wider range of workloads, while the GTX 660 Ti’s data is limited to compute and Vulkan.

The NVIDIA GeForce GTX 660 Ti’s only "win" is in legacy API support where it has no direct comparison. It supports DirectX 12 (11_0), while the 5600M supports DirectX 12 (12_1), so the older card is not ahead in feature level. However, the GTX 660 Ti does have a higher base clock (915 MHz vs 822 MHz) and boost clock (980 MHz vs 1144 MHz for the AMD), but raw clock speed does not translate to performance here. In practice, the GTX 660 Ti wins in the sense that it is a desktop card with standard display outputs (2x DVI, 1x HDMI 1.4a, 1x DisplayPort 1.2), whereas the 5600M’s outputs are "Portable Device Dependent," meaning it is tied to a laptop chassis. For a fixed desktop setup, the GTX 660 Ti offers plug-and-play connectivity that the mobile AMD part cannot match.

Architecture Differences

The architectural gap between these two is vast. The AMD Radeon Pro 5600M uses the Navi 12 chip on RDNA 1.0 architecture, fabricated on TSMC’s 7nm process. The NVIDIA GeForce GTX 660 Ti uses the GK104 chip on Kepler architecture, fabricated on TSMC’s 28nm process. This process difference alone explains much of the efficiency gap: the 5600M has a 50W TDP while the GTX 660 Ti draws 150W, yet the AMD part delivers over 2x the compute performance in shared tests.

Memory configurations differ sharply. The 5600M has 8GB of HBM2 on a 2048-bit bus, yielding 394.2 GB/s bandwidth. The GTX 660 Ti has 2GB of GDDR5 on a 192-bit bus, yielding 144.2 GB/s. The AMD part has 2560 shading units, 160 TMUs, and 64 ROPs, versus 1344 shading units, 112 TMUs, and 24 ROPs on the NVIDIA card. Pixel rate is 73.22 GPixel/s for the AMD versus 27.44 GPixel/s for the NVIDIA, and texture rate is 183.0 GTexel/s versus 109.8 GTexel/s. FP32 compute is 5.857 TFLOPS for the AMD versus 2.634 TFLOPS for the NVIDIA. The 5600M also supports FP16 at 11.71 TFLOPS (2:1), while the GTX 660 Ti has no FP16 data.

Interface and power delivery also differ. The 5600M uses PCIe 4.0 x16 and requires no power connectors, fitting as an IGP (integrated graphics processor) in a laptop. The GTX 660 Ti uses PCIe 3.0 x16, requires 2x 6-pin power connectors, and a 450W suggested PSU. The GTX 660 Ti is a dual-slot card at 241mm (9.5 inches) long, while the 5600M has no dimensions listed because it is not a standalone card. API support shows the AMD part supports Vulkan 1.3 versus 1.2.175 on the NVIDIA, and both support OpenGL 4.6.

Head-to-Head Benchmarks

The two head-to-head benchmarks are both landslide wins for the AMD Radeon Pro 5600M. In Geekbench OpenCL, the AMD scores 47,783 versus the NVIDIA’s 15,113. That is a delta of 216.2%, meaning the AMD part is more than three times faster in raw compute throughput. This test likely stresses the 5600M’s massive memory bandwidth (394.2 GB/s vs 144.2 GB/s) and higher shading unit count (2560 vs 1344). The 2048-bit HBM2 bus versus the 192-bit GDDR5 bus is a fundamental advantage that shows up in compute-heavy workloads.

In Geekbench Vulkan, the AMD scores 46,425 versus the NVIDIA’s 15,012, a delta of 209.3%. The gap is slightly smaller than OpenCL but still overwhelming. Vulkan is a modern low-level API, and the 5600M’s RDNA 1.0 architecture is designed for it, while Kepler was never optimized for Vulkan—the GTX 660 Ti only supports Vulkan 1.2.175, and its DirectX 12 support is limited to level 11_0. The 5600M’s DirectX 12 (12_1) support gives it access to more advanced features, though the PassMark DirectX tests show it scores surprisingly low: 62 in DX10, 59 in DX11, 39 in DX12, and 118 in DX9. These low scores are likely due to driver overhead or the mobile nature of the card, but they are the only DirectX data points available.

For comparison, the 5600M’s average benchmark score across all tests is 16,351, placing it at the 59th percentile of all GPUs. The GTX 660 Ti’s average is 15,063, at the 57th percentile. The 5600M’s nearest rivals include the AMD Radeon RX 5700 XT (avg 16,361, -0.1%) and the NVIDIA RTX PRO 6000 Blackwell (avg 16,408, -0.3%), while the GTX 660 Ti’s rivals include the AMD Radeon Pro 560X (avg 15,082, -0.1%) and the NVIDIA GeForce RTX 3050 OEM (avg 15,199, -0.9%). The 5600M is effectively tied with those modern cards in average score, while the GTX 660 Ti sits near the bottom of modern low-end GPUs.

The Verdict

From the data, the choice is clear for anyone needing modern compute performance: the AMD Radeon Pro 5600M is the superior GPU by every measurable metric. It is 216% faster in OpenCL and 209% faster in Vulkan than the GTX 660 Ti. It has 4x the memory, 2.7x the bandwidth, and 2.2x the FP32 compute. It does all this at one-third the power draw (50W vs 150W). If you are running compute workloads, modern games, or anything that uses Vulkan or OpenCL, the 5600M is the only rational pick from these two.

However, the GTX 660 Ti is not without a niche. Its 2GB GDDR5 memory and 192-bit bus are small by modern standards, but it is a desktop card with standard display outputs and a 241mm length that fits in most cases. If you are building a legacy system that only needs DirectX 9 or 10 support, the PassMark scores for the GTX 660 Ti are not available, but the 5600M’s DirectX 9 score of 118 is the highest of its PassMark results, suggesting both cards can handle older APIs. The GTX 660 Ti also has a launch MSRP of 299 USD, which is the only pricing data available—the 5600M has no launch MSRP listed.

For the vast majority of users, the 5600M is the answer. It is end-of-life, but so is the GTX 660 Ti. The 5600M’s percentile ranking (59th) is only two points higher than the GTX 660 Ti’s (57th), but that average is dragged down by its weak PassMark DirectX scores. In the tests that matter for modern workloads, the 5600M is in a different league. Pick the GTX 660 Ti only if you need a physical desktop GPU with traditional display outputs and cannot use a mobile IGP solution.

FAQ

Q: Which GPU is faster in Geekbench OpenCL?

A: The AMD Radeon Pro 5600M scores 47,783 versus the NVIDIA GeForce GTX 660 Ti’s 15,113, a 216.2% advantage for the AMD part.

Q: What is the memory bandwidth difference?

A: The 5600M has 394.2 GB/s bandwidth from 8GB HBM2 on a 2048-bit bus, while the GTX 660 Ti has 144.2 GB/s from 2GB GDDR5 on a 192-bit bus.

Q: Which card has a higher TDP?

A: The GTX 660 Ti has a 150W TDP and requires 2x 6-pin power connectors with a 450W suggested PSU, while the 5600M has a 50W TDP and no power connectors.

Q: Do both cards support Vulkan?

A: Yes, but the 5600M supports Vulkan 1.3 while the GTX 660 Ti supports Vulkan 1.2.175. In the Geekbench Vulkan test, the 5600M scores 46,425 versus 15,012.

Q: Which card has more shading units?

A: The 5600M has 2560 shading units, 160 TMUs, and 64 ROPs. The GTX 660 Ti has 1344 shading units, 112 TMUs, and 24 ROPs.

Q: What is the average benchmark score for each?

A: The 5600M averages 16,351 across all tests, placing at the 59th percentile. The GTX 660 Ti averages 15,063, placing at the 57th percentile.

Specification Differences

| Specification | AMD Radeon Pro 5600M | NVIDIA GeForce GTX 660 Ti |

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

| Architecture | RDNA 1.0 | Kepler |

| Process Node | 7 nm | 28 nm |

| Transistors | Not listed | 3,540 million |

| Die Size | Not listed | 294 mm² |

| Transistor Density | Not listed | 12.0M / mm² |

| Base Clock | 822 MHz | 915 MHz |

| Boost Clock | 1144 MHz | 980 MHz |

| Memory Size | 8 GB | 2 GB |

| Memory Type | HBM2 | GDDR5 |

| Memory Bus Width | 2048 bit | 192 bit |

| Memory Bandwidth | 394.2 GB/s | 144.2 GB/s |

| Shading Units | 2560 | 1344 |

| TMUs | 160 | 112 |

| ROPs | 64 | 24 |

| Pixel Rate | 73.22 GPixel/s | 27.44 GPixel/s |

| Texture Rate | 183.0 GTexel/s | 109.8 GTexel/s |

| FP32 Performance | 5.857 TFLOPS | 2.634 TFLOPS |

| FP16 Performance | 11.71 TFLOPS (2:1) | Not listed |

| TDP | 50 W | 150 W |

| Slot Width | IGP | Dual-slot |

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

| Suggested PSU | Not listed | 450 W |

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

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

| DirectX Support | 12 (12_1) | 12 (11_0) |

| Vulkan Support | 1.3 | 1.2.175 |

| Length | Not listed | 241 mm (9.5 inches) |

| Release Date | 2020-06-14 | 2012-08-15 |

| Launch MSRP | Not listed | 299 USD |