NVIDIA GeForce RTX 2070

The NVIDIA GeForce RTX 2070 is a Turing-architecture graphics card built on the TU106 chip at TSMC's 12 nm process. It packs 10,800 million transistors on a 445 mm² die, with 2304 shading units, 144 TMUs, and 64 ROPs. The card carries 8 GB of GDDR6 on a 256-bit bus, yielding 448.0 GB/s of bandwidth. With a base clock of 1410 MHz and boost of 1620 MHz, it delivers 7.465 TFLOPS of FP32 compute and 14.93 TFLOPS of FP16. It sits in the 62nd percentile of all GPUs in the database, with an average benchmark score of 19,680. Its launch MSRP was 499 USD.

Benchmark Performance

The average benchmark score of 19,680 places the RTX 2070 in the 62nd percentile of all GPUs, meaning it outperforms a majority of the field. In the 3DMark Steel Nomad DX12 test, it scores 1,569 points, a result that reflects solid modern API performance. Geekbench returns 88,108 in OpenCL and 83,291 in Vulkan, indicating strong compute and cross-platform performance. Passmark's G3D score of 16,094 and GPU compute score of 6,411 further confirm its position. However, the legacy DirectX scores are low—DX9 at 211, DX10 at 111, and DX12 at 60—suggesting the card's architecture favors newer workloads. The FP32 throughput of 7.465 TFLOPS and texture fill rate of 233.3 GTexel/s support these results, while the pixel rate of 103.7 GPixel/s helps with high-resolution rendering.

Compared to its nearest rivals, the RTX 2070 is 0.1% behind the GTX TITAN, 0.3% ahead of the Quadro K5200, 0.5% behind the RTX 2060 SUPER, and 0.8% ahead of the Tesla K40m. These sub-1% deltas place the card in a tightly contested performance tier, where small differences in driver optimization or workload can flip the order. The 3DMark Steel Nomad score, being a modern DX12 benchmark, suggests the card is well-prepared for current and upcoming titles that leverage DirectX 12 Ultimate features. The low DX9 and DX10 scores are not a concern for contemporary gaming, but they indicate that the card's strength lies in its modern compute and shading capabilities rather than legacy pipeline optimization.

How It Compares

The RTX 2070 trails the GTX TITAN by a negligible 0.1% in average score. This effectively ties the two cards, but the RTX 2070 brings hardware ray tracing and tensor cores, which the older TITAN lacks. In legacy benchmarks, the TITAN might have an edge, but the RTX 2070's modern feature set makes it the more versatile choice for gamers and content creators who want to leverage current APIs and AI-based technologies.

The RTX 2070 leads the Quadro K5200 by 0.3%. The K5200 is a professional workstation card, so this margin shows the RTX 2070 can deliver comparable compute performance while also handling gaming workloads. The RTX 2070's GDDR6 memory and higher bandwidth likely contribute to this edge, and its support for DirectX 12 Ultimate and Vulkan 1.4 gives it a broader software compatibility profile.

The RTX 2070 is 0.5% slower than the RTX 2060 SUPER in average score. This is a close call; the 2060 SUPER edges ahead, but the 2070 offers a larger memory bus and higher bandwidth (448.0 GB/s). In practice, the difference is within noise, and the choice may come down to other factors like display outputs or RT core count. Both cards are well-matched for high-resolution gaming, and the 2070's slightly higher memory bandwidth could be an advantage in texture-heavy scenes.

The RTX 2070 outperforms the Tesla K40m by 0.8%. The K40m is a compute-focused card, so this margin indicates the RTX 2070 holds its own in general workloads. The RTX 2070's higher FP32 throughput (7.465 TFLOPS) and modern architecture likely account for the advantage. The Tesla K40m is a much older design, and the RTX 2070's efficiency and feature set make it a more practical choice for any mixed-use scenario.

Memory Subsystem

The RTX 2070 is equipped with 8 GB of GDDR6 memory on a 256-bit bus, delivering 448.0 GB/s of bandwidth. The effective memory clock is 14 Gbps (1750 MHz). This configuration is well-suited for high-resolution gaming, where large textures and geometry data demand rapid transfer. The 256-bit bus width provides a balance between capacity and bandwidth, avoiding the bottlenecks seen in narrower 128-bit designs. For ultra-high resolution workloads, the 8 GB capacity is sufficient for most titles, though some extreme texture packs may exceed it. The pixel rate of 103.7 GPixel/s and texture rate of 233.3 GTexel/s further support high-resolution rendering, ensuring that fill-rate limits are not the primary constraint. The memory subsystem is a key strength of this card, offering ample bandwidth for the shading units and RT cores to operate without starvation.

Who Should Consider It

The RTX 2070's average score of 19,680 places it in the 62nd percentile, making it a strong choice for gamers targeting high-refresh mainstream gaming or high-settings high-resolution play. Its 3DMark Steel Nomad score of 1,569 indicates it can handle modern DX12 titles without issue. The memory bandwidth of 448.0 GB/s ensures that texture streaming at high resolutions is not a bottleneck. For users who prioritize ray tracing, the 36 RT cores provide hardware acceleration, though the performance cost is not quantified here. Those who already own a GTX 10-series card will see a significant uplift in compute and feature support, as the RTX 2070 supports DirectX 12 Ultimate and Vulkan 1.4. However, the sub-1% deltas against the RTX 2060 SUPER mean the latter is a very close competitor, and the choice may hinge on specific feature needs or availability. The card is marked as end-of-life, so prospective buyers should check for remaining stock. For those building a new system or upgrading from an older generation, the RTX 2070 offers a balanced mix of compute, memory, and ray tracing capability that remains relevant for current gaming workloads.

Ray Tracing and Feature Set

With 36 RT cores and 288 tensor cores, the RTX 2070 brings dedicated ray tracing and AI acceleration to the table. It supports DirectX 12 Ultimate (feature level 12_2), OpenGL 4.6, and Vulkan 1.4. The tensor cores enable DLSS and other AI-based features, which can improve performance in supported titles. Display outputs include one DVI, one HDMI 2.0, two DisplayPort 1.4a, and one USB Type-C, offering flexible connectivity for multi-monitor setups and VR headsets. The card's Turing architecture introduced these features to the consumer market, making it a foundational piece for modern gaming features. The inclusion of a USB Type-C port is notable for VR headsets and future-proofing, while the dual DisplayPort 1.4a outputs support high refresh rate monitors. The API support is comprehensive, ensuring compatibility with the latest game engines and compute frameworks.

Power and Cooling

The RTX 2070 has a TDP of 175 W, requiring a power supply rated at 450 W or higher. It uses a single 8-pin PCIe power connector. The card is dual-slot and measures 229 mm (9 inches) in length, 113 mm (4.4 inches) in height, and 35 mm (1.4 inches) in width, fitting most mid-tower cases. The cooling solution is not detailed in the data, but the dual-slot design suggests a capable air cooler. The 12 nm process and 10,800 million transistors contribute to the power envelope, which is moderate for the performance class. The single 8-pin connector simplifies cable management, and the 450 W PSU recommendation is within reach of most standard power supplies. The card's dimensions are compact enough for many small-form-factor builds, though the dual-slot width should be checked against case clearance.