NVIDIA GeForce RTX 2070 vs NVIDIA GeForce RTX 3070 Comparison

The Verdict

The recorded data separates these two NVIDIA cards clearly. The RTX 3070 wins 9 of the 10 head-to-head benchmark comparisons, with the RTX 2070 securing a single, albeit massive, victory in the Geekbench Vulkan test. For users prioritizing modern DirectX 12 workloads, compute tasks, and overall rasterization performance, the RTX 3070 is the definitive choice. The RTX 2070 remains relevant only in the specific scenario where Vulkan API performance is the sole criterion, as its score of 82521 dwarfs the RTX 3070's 21022, a 292.5% advantage. However, this outlier does not compensate for the RTX 3070's commanding leads elsewhere, such as a 50.4% improvement in 3DMark Steel Nomad DX12 and a 42.7% lead in Passmark GPU Compute. The database positions the RTX 3070 in the 61st percentile versus the RTX 2070's 63rd percentile, but this metric reflects the entire GPU landscape, not the direct comparison. The RTX 2070's average benchmark score of 18789 slightly exceeds the RTX 3070's 17208, a quirk driven entirely by that anomalous Vulkan result. For any buyer choosing between these two specific cards, the data overwhelmingly favors the RTX 3070 for its generational leap in throughput and API efficiency.

Architecture Differences

The two cards represent distinct architectural generations. The RTX 2070 utilizes the TU106 chip built on Turing architecture, manufactured on a 12 nm process at TSMC. The RTX 3070 employs the GA104 chip with Ampere architecture, fabricated on an 8 nm node at Samsung. This process shrink enables a significant transistor density increase: the RTX 3070 packs 17,400 million transistors into a 392 mm² die, yielding 44.4M transistors per mm², while the RTX 2070 houses 10,800 million transistors on a larger 445 mm² die, achieving only 24.3M per mm². The RTX 3070's Ampere design also doubles the FP32 throughput per clock compared to Turing, which is reflected in its raw specifications. The RTX 3070 features 5888 shading units, 184 TMUs, and 96 ROPs, versus the RTX 2070's 2304 shading units, 144 TMUs, and 64 ROPs. Ray tracing hardware also scales up: the RTX 3070 has 46 RT cores and 184 tensor cores, while the RTX 2070 offers 36 RT cores and 288 tensor cores. Interestingly, the tensor core count is higher on the older card, but the Ampere tensor cores operate with different efficiency characteristics. The RTX 3070's FP16 performance matches its FP32 at 20.31 TFLOPS (1:1 ratio), whereas the RTX 2070 achieves 14.93 TFLOPS FP16 via a 2:1 ratio from its 7.465 TFLOPS FP32. Both cards use 8 GB of GDDR6 memory on a 256-bit bus with identical 448.0 GB/s bandwidth and 1750 MHz memory clock, but the RTX 3070's memory subsystem benefits from the newer architecture's improved access patterns.

FAQ

Q: Which card wins the most benchmarks in the database?

A: The NVIDIA GeForce RTX 3070 wins 9 out of 10 head-to-head benchmark comparisons, with the RTX 2070 winning only the Geekbench Vulkan test.

Q: Is the RTX 2070 ever faster than the RTX 3070?

A: Yes, in the Geekbench Vulkan test, the RTX 2070 scores 82521 versus the RTX 3070's 21022, a 292.5% advantage for the older card.

Q: How do the process nodes differ between the two cards?

A: The RTX 2070 uses a 12 nm process from TSMC, while the RTX 3070 uses an 8 nm process from Samsung.

Q: What is the difference in FP32 compute performance?

A: The RTX 3070 delivers 20.31 TFLOPS FP32, which is 2.72 times the RTX 2070's 7.465 TFLOPS.

Q: Do the cards have the same memory configuration?

A: Yes, both have 8 GB GDDR6, a 256-bit bus width, 448.0 GB/s bandwidth, and 14 Gbps effective memory speed.

Q: What are the power requirements?

A: The RTX 2070 has a 175 W TDP with a suggested 450 W PSU, while the RTX 3070 has a 220 W TDP with a suggested 550 W PSU.

Specification Differences

The database records the following key differences between the two cards. The RTX 2070 operates with a base clock of 1410 MHz and boost clock of 1620 MHz, while the RTX 3070 runs at 1500 MHz base and 1725 MHz boost. The RTX 3070 delivers a pixel rate of 165.6 GPixel/s and texture rate of 317.4 GTexel/s, far exceeding the RTX 2070's 103.7 GPixel/s and 233.3 GTexel/s. The RTX 3070's FP32 throughput is 20.31 TFLOPS versus 7.465 TFLOPS, and its FP16 output is 20.31 TFLOPS (1:1) versus 14.93 TFLOPS (2:1). The transistor counts differ substantially (17,400 million versus 10,800 million), as do die sizes (392 mm² versus 445 mm²) and transistor densities (44.4M per mm² versus 24.3M per mm²). The RTX 3070 uses a PCIe 4.0 x16 interface, while the RTX 2070 uses PCIe 3.0 x16. Display outputs also differ: the RTX 2070 offers 1x DVI, 1x HDMI 2.0, 2x DisplayPort 1.4a, and 1x USB Type-C, whereas the RTX 3070 provides 1x HDMI 2.1 and 3x DisplayPort 1.4a. The power connector changed from 1x 8-pin on the RTX 2070 to 1x 12-pin on the RTX 3070. Physical dimensions vary slightly: the RTX 2070 measures 229 mm long, 113 mm high, and 35 mm wide, while the RTX 3070 is 242 mm long and 112 mm high with no recorded width. The RTX 2070 launched on 2018-10-16, and the RTX 3070 on 2020-08-31. Both cards are end-of-life and share the same launch MSRP of 499 USD. Both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Head-to-Head Benchmarks

The benchmark data reveals a lopsided contest. In 3DMark Steel Nomad DX12, the RTX 3070 scores 3162 against the RTX 2070's 1569, a 50.4% advantage. This is the largest performance gap in the modern API tests and highlights the Ampere architecture's efficiency in DirectX 12 workloads. The Geekbench OpenCL test shows the RTX 3070 scoring 112821 versus 79966, a 29.1% lead, indicating substantially better general-purpose compute performance. The Passmark GPU Compute test reinforces this trend: the RTX 3070 achieves 11195, which is 42.7% higher than the RTX 2070's 6411. This compute advantage stems from the RTX 3070's 5888 shading units and doubled FP32 throughput. In DirectX 11, the RTX 3070 scores 182 versus 129, a 29.1% gain, while DirectX 10 shows a 26% improvement (150 versus 111). The DirectX 12 Passmark test records 85 for the RTX 3070 and 60 for the RTX 2070, a 29.4% delta. Legacy DirectX 9 performance favors the RTX 3070 as well, with a score of 247 compared to 211, a 14.6% difference. The 2D graphics test (Passmark G2D) shows the RTX 3070 ahead at 1001 versus 819, an 18.2% margin. The Passmark G3D test, which aggregates overall 3D performance, gives the RTX 3070 a 22214 score versus 16094, a 27.6% advantage. The only exception to this pattern is the Geekbench Vulkan test, where the RTX 2070 scores 82521 and the RTX 3070 scores only 21022. This 292.5% difference is anomalous and likely reflects driver or implementation quirks in the Vulkan path for the Ampere card at the time of testing. It is worth noting the RTX 2070's average benchmark score of 18789 is higher than the RTX 3070's 17208, entirely due to this Vulkan outlier pulling the average up. The nearest rival data places the RTX 2070 within 0.9% of the AMD Radeon RX 560X and 0.6% of the AMD Radeon Pro 5700 XT, while the RTX 3070 sits 0.7% above the AMD Radeon RX 7600 XT and 1.1% above the AMD Radeon HD 7970M. These rival comparisons contextualize the generational leap: the RTX 3070's performance class is closer to newer mainstream cards, while the RTX 2070 aligns with older professional and mid-range parts.