NVIDIA GeForce GTX 1070 vs NVIDIA GeForce GTX 1080 Comparison

Where Each One Wins

The recorded data is unambiguous: the NVIDIA GeForce GTX 1080 wins every single head-to-head benchmark in the database. Across 11 direct comparisons, the GTX 1080 takes all 11, while the GTX 1070 records zero wins. This is not a close contest with split results; it is a complete sweep in favor of the higher-tier card.

Look at the compute-oriented tests first. In PassMark GPU Compute, the GTX 1080 scores 7614 against 6102, a 24.8% advantage. That matters for users running GPGPU workloads, video encoding, or any task that stresses raw shader throughput. The GTX 1070 is not slow in absolute terms, but it trails by a quarter in this metric, which is a substantial gap for productivity tasks.

The 3DMark Steel Nomad DX12 test shows the largest single delta between the two cards. The GTX 1080 posts 1560, while the GTX 1070 manages 1082. That is a 44.2% lead for the GTX 1080, the biggest margin in the entire comparison. For modern DirectX 12 titles, this result suggests the GTX 1080 has significantly more headroom in API-heavy scenes.

Vulkan performance follows a similar pattern. The Geekbench Vulkan score for the GTX 1080 is 30398, versus 22121 for the GTX 1070, a 37.4% difference. Users targeting Vulkan-based games or emulators should expect a much smoother experience on the GTX 1080.

The GTX 1080 also leads in legacy and general-purpose tests, though by smaller margins. PassMark DirectX 9 shows 211 versus 197, only 7.1% apart. PassMark G2D, a 2D graphics test, sees 888 versus 846, a 5% gap. These older workloads do not favor one card strongly, but the GTX 1080 still comes out ahead.

The overall picture: the GTX 1080 is the clear choice for maximum performance in every measured category. The GTX 1070 is the more modest option, and its wins are absent from the data entirely.

Architecture Differences

Both cards share the same fundamental design. They use the GP104 chip, the Pascal architecture, a 16 nm process from TSMC, 7,200 million transistors, and a 314 mm² die size. Transistor density is identical at 22.9M per mm². This means the architectural foundation is the same, and the performance gap comes from how each card is configured within that framework.

The core counts differ significantly. The GTX 1080 has 2560 shading units, 160 texture mapping units, and 64 ROPs. The GTX 1070 has 1920 shading units, 120 TMUs, and the same 64 ROPs. The GTX 1080 thus has 25% more shaders and 33% more TMUs, which directly explains its higher texture rate of 277.3 GTexel/s against 202.0 GTexel/s.

Clocks also favor the GTX 1080. Its base clock is 1607 MHz with a boost of 1733 MHz, while the GTX 1070 runs at 1506 MHz base and 1683 MHz boost. The combination of more cores and higher clocks gives the GTX 1080 a substantial raw throughput advantage. Its FP32 performance is 8.873 TFLOPS, compared to 6.463 TFLOPS for the GTX 1070, a difference of roughly 37%.

Memory is another major differentiator. Both cards have 8 GB and a 256-bit bus, but the GTX 1080 uses GDDR5X at 10 Gbps effective, yielding 320.3 GB/s of bandwidth. The GTX 1070 uses GDDR5 at 8 Gbps effective, yielding 256.3 GB/s. That is a 25% bandwidth advantage for the GTX 1080, which helps in high-resolution textures and memory-heavy workloads. The memory clock is listed as 1251 MHz for the GTX 1080 and 2002 MHz for the GTX 1070, but the effective data rate is what matters: 10 Gbps versus 8 Gbps.

Pixel rates are nearly identical: 110.9 GPixel/s for the GTX 1080 and 107.7 GPixel/s for the GTX 1070. Since both have 64 ROPs, the small difference comes from clock speeds. FP16 performance is also in the same ratio, with 138.6 GFLOPS for the GTX 1080 and 101.0 GFLOPS for the GTX 1070, both at a 1:64 ratio to FP32.

Power draw differs: the GTX 1080 is rated at 180 W TDP, the GTX 1070 at 150 W. Both use a single 8-pin power connector and a suggested 450 W PSU. Both are dual-slot cards with identical dimensions of 267 mm length, 112 mm height, and 40 mm width. Both have the same display outputs, bus interface, and API support. The GTX 1080 simply does more per watt, but also consumes more.

Head-to-Head Benchmarks

The biggest win for the GTX 1080 is in 3DMark Steel Nomad DX12, where it scores 1560 against 1082. That is a 44.2% lead, the largest delta in the entire set. This test stresses DirectX 12 features heavily, and the GTX 1080's extra shading units and faster memory translate directly into a dominant result.

Geekbench Vulkan shows a 37.4% gap, with the GTX 1080 at 30398 and the GTX 1070 at 22121. Vulkan is a low-overhead API, so the raw compute advantage of the GTX 1080 shows through clearly. Users running Vulkan titles, especially esports or emulation workloads, will see a major difference.

Geekbench Metal gives the GTX 1080 a 26.7% lead, scoring 23824 versus 18801. This is relevant for macOS or iOS-adjacent workloads, though both cards are primarily Windows-oriented. The margin is still large enough to matter for Metal-based rendering.

PassMark GPU Compute is another notable result: 7614 versus 6102, a 24.8% advantage. This test measures general compute performance, and the GTX 1080 delivers nearly a quarter more throughput. For non-gaming tasks like rendering or simulation, this is the number to watch.

PassMark DirectX 11 shows 124 versus 100, a 24% lead for the GTX 1080. DirectX 11 remains common in many titles, and this result confirms the GTX 1080 handles them with more headroom. DirectX 12 and DirectX 10 tests both show 14.6% gaps, with scores of 55 versus 48 and 93 versus 82 respectively.

PassMark G3D, a general 3D performance metric, has the GTX 1080 at 15586 and the GTX 1070 at 13498, a 15.5% difference. This aligns with the average benchmark score gap: the GTX 1080 averages 11960 across all tests, while the GTX 1070 averages 9780, roughly 22% higher for the GTX 1080.

Geekbench OpenCL shows a 14.6% lead, with 51204 versus 44700. This is a compute-heavy test, and the GTX 1080 again comes out ahead. PassMark DirectX 9 is the closest result: 211 versus 197, only 7.1% apart. PassMark G2D is even closer in percentage terms, 888 versus 846, a 5% gap. These older tests do not stress the hardware as much, so the differences shrink.

FAQ

Q: Which card has higher raw compute performance?

A: The GTX 1080. Its FP32 throughput is 8.873 TFLOPS versus 6.463 TFLOPS for the GTX 1070, a difference of about 37%. This shows up in compute benchmarks like PassMark GPU Compute, where the GTX 1080 leads by 24.8%.

Q: Do the cards use the same memory bus width?

A: Yes, both use a 256-bit bus. However, the GTX 1080 uses GDDR5X at 10 Gbps effective, giving 320.3 GB/s bandwidth, while the GTX 1070 uses GDDR5 at 8 Gbps effective, giving 256.3 GB/s.

Q: What is the average benchmark score difference?

A: The GTX 1080 has an average benchmark score of 11960, while the GTX 1070 averages 9780. The GTX 1080 sits at the 51st percentile of all GPUs, the GTX 1070 at the 47th.

Q: Which card draws more power?

A: The GTX 1080 is rated at 180 W TDP, the GTX 1070 at 150 W. Both use a single 8-pin power connector and suggest a 450 W PSU.

Q: Are the cards identical in physical size?

A: Yes. Both are dual-slot, 267 mm long, 112 mm tall, and 40 mm wide. They also have identical display outputs and the same PCIe 3.0 x16 bus interface.

Q: Which card wins in DirectX 12?

A: The GTX 1080. In 3DMark Steel Nomad DX12, it scores 1560 against 1082, a 44.2% lead. In PassMark DirectX 12, it scores 55 versus 48, a 14.6% lead.

The Verdict

The data points to a clear hierarchy. The GTX 1080 is the stronger card in every measured benchmark, with an average score of 11960 against 9780 for the GTX 1070. Its nearest rivals are cards like the GTX 960A at 11998, only 0.3% ahead, and the Radeon RX 6500 XT at 11842, 1% behind. The GTX 1070, by contrast, sits near the AMD FirePro W5000 and NVIDIA Quadro M2000M, all within a percentage point of each other. This places the GTX 1080 in a higher performance tier entirely.

Pick the GTX 1080 if you want maximum frame rates in modern APIs, especially DirectX 12 and Vulkan, where its leads are 44.2% and 37.4% respectively. Its higher FP32 throughput and memory bandwidth make it the better choice for compute workloads and high-resolution texture-heavy games. The 180 W TDP is a modest cost for that extra performance, and the suggested 450 W PSU is unchanged from the GTX 1070.

Pick the GTX 1070 if you are willing to trade measurable performance for lower power draw. Its 150 W TDP is 30 W less, and its average score of 9780 is still respectable. In older or lighter workloads, such as DirectX 9 and 2D tests, the gap shrinks to 7.1% and 5% respectively, so the GTX 1070 remains serviceable for less demanding software. But the database shows no scenario where the GTX 1070 outperforms the GTX 1080.

For most builders, the GTX 1080 is the straightforward recommendation. It wins every benchmark, has a higher percentile ranking, and its architectural advantages are consistent across APIs and compute workloads. The GTX 1070 is a fine card, but it is strictly a step below.

Specification Differences

| Specification | NVIDIA GeForce GTX 1080 | NVIDIA GeForce GTX 1070 |

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

| Base clock | 1607 MHz | 1506 MHz |

| Boost clock | 1733 MHz | 1683 MHz |

| Memory clock | 1251 MHz, 10 Gbps effective | 2002 MHz, 8 Gbps effective |

| Memory type | GDDR5X | GDDR5 |

| Memory bandwidth | 320.3 GB/s | 256.3 GB/s |

| Shading units | 2560 | 1920 |

| TMUs | 160 | 120 |

| Pixel rate | 110.9 GPixel/s | 107.7 GPixel/s |

| Texture rate | 277.3 GTexel/s | 202.0 GTexel/s |

| FP32 performance | 8.873 TFLOPS | 6.463 TFLOPS |

| FP16 performance | 138.6 GFLOPS (1:64) | 101.0 GFLOPS (1:64) |

| TDP | 180 W | 150 W |

| Launch MSRP | 599 USD | 379 USD |

Both cards share the same GP104 chip, Pascal architecture, 16 nm TSMC process, 7,200 million transistors, 314 mm² die size, 8 GB memory capacity, 256-bit bus width, 64 ROPs, dual-slot design, single 8-pin power connector, 450 W suggested PSU, and identical dimensions. The GTX 1080 was released on 2016-05-26, the GTX 1070 on 2016-06-09. Both are end-of-life products with the same API support: DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4.