NVIDIA GeForce RTX 2060 SUPER vs NVIDIA GeForce RTX 3070 Comparison

The benchmark data presents a clear generational shift. The NVIDIA GeForce RTX 3070 wins 9 of the 10 head-to-head tests, often by significant margins, while the RTX 2060 SUPER manages a single, but enormous, victory in one specific API test. The data suggests that while the newer card is broadly dominant, the older Turing architecture retains a peculiar strength in a particular Vulkan workload.

The Verdict

The data points decisively toward the RTX 3070 for nearly all use cases. Its average benchmark score of 17208, while slightly lower than the RTX 2060 SUPER's 18093, is misleading due to the latter's outlier Vulkan result. In the more common DirectX and compute workloads, the RTX 3070 is consistently faster. For gamers targeting high-refresh-rate 1440p or smooth 4K experiences in modern DirectX 12 titles, the RTX 3070 is the clear choice. Its 25.9% lead in the Passmark G3D test and its 28.2% advantage in DirectX 12 benchmarks show a substantial performance cushion.

However, the RTX 2060 SUPER is not without its niche. Its staggering 268.2% lead in the Geekbench Vulkan test is a data anomaly that cannot be ignored. This suggests that for specific applications or games that are heavily optimized for the Vulkan API in a way that favors Turing's design, the older card could offer a unique advantage. This is a rare scenario, but the data is unambiguous about its existence. The RTX 2060 SUPER, with its launch MSRP of 399 USD, was positioned as a more accessible high-performance card, but the RTX 3070, at a launch MSRP of 499 USD, commands a significant performance premium that the benchmarks justify.

Architecture Differences

The two cards represent distinct architectural generations. The RTX 2060 SUPER is built on the Turing architecture and uses the TU106 chip, fabricated on a 12 nm process at TSMC. In contrast, the RTX 3070 is an Ampere-based card with the GA104 chip, built on Samsung's 8 nm process. This process shift is a major factor in the performance difference. The RTX 3070 packs 17,400 million transistors into a 392 mm² die, yielding a transistor density of 44.4M / mm². The RTX 2060 SUPER, meanwhile, has 10,800 million transistors on a larger 445 mm² die, resulting in a much lower density of 24.3M / mm².

This architectural leap translates into a massive increase in raw compute resources. The RTX 3070 features 5888 shading units, 184 texture mapping units (TMUs), and 96 raster output units (ROPs). This is a significant jump from the RTX 2060 SUPER's 2176 shading units, 136 TMUs, and 64 ROPs. The ray tracing and tensor core configurations also differ: the RTX 3070 has 46 RT cores and 184 tensor cores, while the RTX 2060 SUPER has 34 and 272, respectively. Notably, the RTX 2060 SUPER has more tensor cores, but the RTX 3070's are from a newer, more capable generation.

The memory subsystems are identical in capacity and bandwidth. Both cards have 8 GB of GDDR6 memory on a 256-bit bus, providing 448.0 GB/s of bandwidth. The clock speeds are also similar, with the RTX 3070 having a slightly higher base and boost clock (1500 MHz / 1725 MHz) compared to the RTX 2060 SUPER (1470 MHz / 1650 MHz). The key difference lies in the FP32 and FP16 compute throughput. The RTX 3070 achieves 20.31 TFLOPS for both FP32 and FP16 (1:1), while the RTX 2060 SUPER offers 7.181 TFLOPS FP32 and 14.36 TFLOPS FP16 (2:1). This indicates a fundamental difference in how the two architectures handle compute workloads.

FAQ

Q: Which card is faster in the 3DMark Steel Nomad DirectX 12 test?

A: The NVIDIA GeForce RTX 3070 is significantly faster, scoring 3162 compared to the RTX 2060 SUPER's 2011, a difference of 36.4%.

Q: Is there any benchmark where the RTX 2060 SUPER beats the RTX 3070?

A: Yes, in the Geekbench Vulkan test, the RTX 2060 SUPER scores 77402, which is 268.2% higher than the RTX 3070's score of 21022.

Q: How do the cards compare in general compute performance?

A: The RTX 3070 is substantially stronger. In the Passmark GPU Compute test, it scores 11195, which is 40% higher than the RTX 2060 SUPER's 6721.

Q: Do the cards have the same amount of memory and bandwidth?

A: Yes, both the RTX 2060 SUPER and RTX 3070 feature 8 GB of GDDR6 memory on a 256-bit bus, providing identical 448.0 GB/s of bandwidth.

Q: What is the difference in their power requirements?

A: The RTX 3070 has a TDP of 220 W and requires a 550 W power supply, while the RTX 2060 SUPER has a TDP of 175 W and suggests a 450 W power supply.

Q: Which card has a higher percentile ranking among all GPUs?

A: The RTX 2060 SUPER has a slightly higher percentile ranking at 62, compared to the RTX 3070's 61, despite the RTX 3070 having a higher average score in most individual tests.

Specification Differences

The following table outlines the key specifications where the two cards differ.

| Specification | NVIDIA GeForce RTX 2060 SUPER | NVIDIA GeForce RTX 3070 |

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

| Architecture | Turing | Ampere |

| Chip | TU106 | GA104 |

| Process Node | 12 nm | 8 nm |

| Foundry | TSMC | Samsung |

| Transistors | 10,800 million | 17,400 million |

| Die Size | 445 mm² | 392 mm² |

| Transistor Density | 24.3M / mm² | 44.4M / mm² |

| Base Clock | 1470 MHz | 1500 MHz |

| Boost Clock | 1650 MHz | 1725 MHz |

| Shading Units | 2176 | 5888 |

| TMUs | 136 | 184 |

| ROPs | 64 | 96 |

| RT Cores | 34 | 46 |

| Tensor Cores | 272 | 184 |

| Pixel Rate | 105.6 GPixel/s | 165.6 GPixel/s |

| Texture Rate | 224.4 GTexel/s | 317.4 GTexel/s |

| FP32 Performance | 7.181 TFLOPS | 20.31 TFLOPS |

| FP16 Performance | 14.36 TFLOPS (2:1) | 20.31 TFLOPS (1:1) |

| TDP | 175 W | 220 W |

| Power Connectors | 1x 8-pin | 1x 12-pin |

| Suggested PSU | 450 W | 550 W |

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

| Display Outputs | 1x DVI, 1x HDMI 2.0, 2x DP 1.4a, 1x USB Type-C | 1x HDMI 2.1, 3x DisplayPort 1.4a |

| Length | 229 mm (9 inches) | 242 mm (9.5 inches) |

Head-to-Head Benchmarks

The RTX 3070 dominates the head-to-head benchmarks, but the margins vary significantly. The largest victory for the RTX 3070 is in the 3DMark Steel Nomad DX12 test, where it scores 3162 against the RTX 2060 SUPER's 2011, a 36.4% lead. This is closely followed by a 40% lead in the Passmark GPU Compute test (11195 vs 6721), showing its strength in non-graphics compute workloads. In the Geekbench OpenCL test, the RTX 3070 is 31.8% ahead, scoring 112821 versus 76957.

The DirectX benchmark suite shows consistent, but slightly smaller, wins for the RTX 3070. It leads by 26% in DirectX 10 (150 vs 111), 28.6% in DirectX 11 (182 vs 130), and 28.2% in DirectX 12 (85 vs 61). The Passmark G3D test, which is a general 3D graphics performance measure, shows a 25.9% advantage for the RTX 3070 (22214 vs 16462). Even in the older DirectX 9 test, the RTX 3070 wins by 11.7% (247 vs 218). The 2D performance test also favors the RTX 3070, with a 14.7% lead (1001 vs 854).

The one outlier is the Geekbench Vulkan test. Here, the RTX 2060 SUPER wins decisively, scoring 77402 compared to the RTX 3070's 21022. This is a 268.2% difference in favor of the older card. This result is so anomalous that it suggests a driver or software-specific optimization for the Turing architecture in this particular workload, rather than a general Vulkan superiority. It is the sole reason the RTX 2060 SUPER has a higher average benchmark score (18093) than the RTX 3070 (17208).

Where Each One Wins

The RTX 3070 is the winner in almost every scenario a user would encounter. Its performance in DirectX 10, 11, and 12, as well as OpenCL, makes it the superior choice for modern gaming and general-purpose GPU compute. The data shows that for any workload using these APIs, the RTX 3070 provides a substantial performance uplift, ranging from roughly 26% to 40%. Its higher pixel rate (165.6 GPixel/s vs 105.6 GPixel/s) and texture rate (317.4 GTexel/s vs 224.4 GTexel/s) further solidify its position as the more capable gaming and rendering card. It also has a higher boost clock, which helps in scenarios that are not purely shader-limited.

The RTX 2060 SUPER's single win is in the Geekbench Vulkan test. This is a very specific and narrow victory. It implies that in an application that is heavily optimized for Vulkan in a way that aligns with the Turing architecture's design, the older card could potentially outperform the newer one. This is a highly specialized case, and the data does not suggest it would translate to a broad advantage in Vulkan gaming. For any other purpose, including all other tested APIs and compute tasks, the RTX 3070 is the definitive winner. The RTX 2060 SUPER's higher percentile ranking (62 vs 61) is a quirk of the averaging method that includes this Vulkan score, and does not reflect its performance in the majority of benchmarks.