NVIDIA GeForce GTX 1070

The NVIDIA GeForce GTX 1070, built on the Pascal architecture and 16 nm process, has a thermal design power (TDP) of 150 W. This power envelope is modest for a desktop graphics card of its era, positioning it as an efficient performer relative to its raw throughput capabilities. The data indicates that a suggested power supply of 450 W is required to support the card, which is a reasonable requirement for a mid-range system of its generation.

Power delivery is handled by a single 1x 8-pin power connector. This is a standard configuration for a card with a 150 W TDP, ensuring that most power supplies with an 8-pin PCIe cable can accommodate it without adapter complications. The card occupies a dual-slot width, which is typical for a cooling solution capable of managing the thermal output of the GP104 chip. With physical dimensions of 267 mm in length, 112 mm in height, and 40 mm in width, the GTX 1070 is designed to fit comfortably in most modern PC cases, though users with smaller form factors should verify clearance. The cooling requirements are standard for the performance class, relying on a capable air cooler rather than exotic liquid solutions, and the 450 W PSU recommendation aligns with the total system draw expectations for a single-GPU setup.

Ray Tracing and Feature Set

The GeForce GTX 1070 is a product of the Pascal generation, which predates NVIDIA's dedicated ray tracing and tensor core hardware. The fact pack confirms that the card has no RT cores and no tensor cores. Consequently, hardware-accelerated ray tracing is not part of its feature set, and any ray tracing workloads would rely on compute shaders, which is inefficient compared to dedicated hardware found in later generations. Similarly, AI-accelerated features that depend on tensor cores, such as deep learning super sampling, are not supported by this hardware.

The card's feature set is instead defined by its API support and display output capabilities. It supports DirectX 12 (12_1), indicating compatibility with the feature level required for modern gaming titles at the time of its release. OpenGL 4.6 and Vulkan 1.4 support are also present, ensuring broad compatibility with contemporary applications and games that utilize these low-level APIs. For display connectivity, the GTX 1070 provides a versatile array of outputs: 1x DVI, 1x HDMI 2.0, and 3x DisplayPort 1.4a. This configuration allows for multi-monitor setups and supports high refresh rate and high-resolution displays through the DisplayPort connections. The memory subsystem consists of 8 GB of GDDR5 memory on a 256-bit bus, delivering a bandwidth of 256.3 GB/s. This capacity and bandwidth were ample for the resolutions and texture loads common during its product cycle, ensuring that the card could handle large frame buffers without bottlenecking the GPU core.

Benchmark Performance

The GTX 1070's average benchmark score across the suite is 12331, placing it at the 51st percentile of all GPUs in the database. This percentile ranking indicates that it sits squarely in the middle of the performance distribution, outperforming half of the tracked graphics cards. An analysis of specific benchmark results shows a nuanced performance profile. In synthetic tests, the card achieves a score of 47490 in Geekbench OpenCL and 47395 in Vulkan, showing strong compute and graphics API performance. Its Passmark G3D score of 13498 is solid, while the DirectX 11 score of 100 is notably higher than the DirectX 12 score of 48, suggesting that the card's architecture is better optimized for legacy DirectX 11 workloads than for the more modern DirectX 12 path in this specific test.

Relative to its nearest rivals, the GTX 1070's performance is tightly clustered. It is only 0.7% faster than the NVIDIA Quadro K4200, a professional-grade card, with an average score of 12241. This margin is negligible, indicating near-identical performance in average workloads. Conversely, it trails the NVIDIA GeForce GTX 880M by a delta of -1.3%, as the GTX 880M achieves a score of 12490. The gap widens slightly against the NVIDIA GeForce GTX 1650 SUPER, which is 1.4% faster with a score of 12504. The largest deficit is against the NVIDIA Tesla K20Xm, a server compute card, which leads by 1.7% with a score of 12547. These deltas, all under two percentage points, demonstrate that the GTX 1070 is firmly entrenched in a highly competitive performance tier where architectural differences and driver optimizations can swing the lead. In terms of specific compute tasks, the GPU compute score of 6102 shows a capable compute unit, while the pixel rate of 107.7 GPixel/s and texture rate of 202.0 GTexel/s provide a baseline for its rasterization throughput. The FP32 performance is rated at 6.463 TFLOPS, with FP16 performance significantly lower at 101.0 GFLOPS due to a 1:64 ratio, indicating that the card is not optimized for half-precision compute tasks.

FAQ

Q: What is the launch MSRP of the NVIDIA GeForce GTX 1070?

A: The launch MSRP for the NVIDIA GeForce GTX 1070 was 379 USD.

Q: Does the GTX 1070 support hardware-accelerated ray tracing?

A: No. The fact pack indicates that the card has no RT cores, meaning it lacks dedicated hardware for ray tracing.

Q: What is the memory bandwidth of the GTX 1070?

A: The card features a 256-bit memory bus with GDDR5 memory, providing a total bandwidth of 256.3 GB/s.

Q: How does the GTX 1070 compare in performance to the GeForce GTX 1650 SUPER?

A: The GTX 1070 is 1.4% slower than the GeForce GTX 1650 SUPER, based on average benchmark scores.

Q: What are the power supply requirements for the GTX 1070?

A: The suggested power supply unit size is 450 W, and the card requires a single 8-pin power connector.

Q: Which API versions are supported by the GTX 1070?

A: The card supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4.

How It Compares

NVIDIA Quadro K4200: The GTX 1070 holds a razor-thin 0.7% performance advantage over the Quadro K4200. With average scores of 12331 and 12241 respectively, the two cards are effectively indistinguishable in raw compute performance. The GTX 1070's edge is within the margin of error for most benchmarks, making the choice between them dependent on factors other than raw speed, such as driver support and feature set.

NVIDIA GeForce GTX 880M: The GTX 1070 is 1.3% slower than the GTX 880M, which posts an average score of 12490. This is a surprising result given the generational gap, as the GTX 880M is a mobile part from an earlier series. The data suggests that the GTX 880M's high-end mobile configuration is able to match or slightly exceed the desktop GTX 1070 in average benchmark aggregation, though the GTX 1070 likely benefits from better sustained performance in longer workloads due to its desktop cooling.

NVIDIA GeForce GTX 1650 SUPER: The GTX 1650 SUPER outperforms the GTX 1070 by 1.4%, with a score of 12504 versus 12331. This newer, lower-tier card manages to edge out the older GTX 1070, highlighting the efficiency gains of newer architectures. The performance delta is small enough that the GTX 1070 remains competitive, but it no longer holds a clear lead over newer budget-oriented options.

NVIDIA Tesla K20Xm: The Tesla K20Xm, a compute-focused server card, leads the GTX 1070 by 1.7%, achieving a score of 12547. This is the largest delta among the nearest rivals. The Tesla's advantage is likely due to its optimized compute performance, which boosts its average score in compute-heavy benchmarks. For gaming, the GTX 1070 would be the more practical choice, but the raw benchmark data places the Tesla slightly ahead.