NVIDIA GeForce GTX 1070 Ti vs NVIDIA GeForce GTX 680 Comparison

The NVIDIA GeForce GTX 1070 Ti and the NVIDIA GeForce GTX 680 represent two distinct eras of graphics hardware, separated by five generations of architectural evolution. The benchmark data confirms a decisive generational leap, but the magnitude of the performance gap and the specific reasons behind it are worth quantifying precisely.

Head-to-Head Benchmarks

The head-to-head comparison is a clean sweep. The GTX 1070 Ti wins all three shared benchmark tests, with the margin widening considerably in compute-heavy workloads. In the Geekbench Metal test, the GTX 1070 Ti scores 22,180 points against the GTX 680’s 7,897 points, a delta of 180.9%. This is not a marginal improvement; it is a near-tripling of raw performance in a graphics API context.

The Geekbench OpenCL result tells a similar story, with the 1070 Ti posting 49,219 points versus 16,906 points for the 680, a 191.1% advantage. This test exercises general-purpose GPU compute, and the data shows that the Pascal architecture’s execution resources are vastly more capable than Kepler’s. The gap becomes even more pronounced in Geekbench Vulkan, where the 1070 Ti reaches 61,006 points against the 680’s 17,646 points, a 245.7% lead. Vulkan’s low-level overhead favors the newer architecture’s scheduling and command-processing efficiency.

It is important to contextualize these wins against the broader market. The 1070 Ti’s average benchmark score is 14,277, placing it in the 56th percentile of all GPUs. Its nearest rival, the Intel Iris Xe MAX Graphics, scores 14,315, which is just 0.3% higher. The AMD Radeon Vega 11 and NVIDIA GeForce GTX TITAN sit at 14,352 and 14,373, respectively, with deltas of -0.5% and -0.7% relative to the 1070 Ti. This means the 1070 Ti is essentially at parity with its immediate competitors, trading blows within a 1% margin.

The GTX 680, meanwhile, has an average score of 14,150, placing it in the 55th percentile. Its nearest rival is the NVIDIA Tesla K10 at 14,029, which the 680 beats by 0.9%. Interestingly, the 680’s rival list includes the 1070 Ti itself, with a delta of -0.9%, meaning the 680 is only 0.9% behind the 1070 Ti in average score. This seems contradictory to the head-to-head results, but it is explained by the fact that the 680 only has three benchmark entries (Metal, OpenCL, Vulkan) while the 1070 Ti has eleven, including DirectX and Passmark tests that pull its average down. The head-to-head data is the more reliable indicator of their relative performance in identical workloads.

The per-test deltas are stark. A 180.9% lead in Metal, 191.1% in OpenCL, and 245.7% in Vulkan are not incremental gains. They indicate that the GTX 1070 Ti is in a completely different performance class for these workloads. The GTX 680’s best showing is in the Metal test, where it still loses by a factor of 2.8. There is no scenario in the shared test suite where the GTX 680 comes close to matching the 1070 Ti.

Architecture Differences

The performance chasm stems from fundamental architectural differences. The GTX 1070 Ti is built on the Pascal architecture using the GP104 chip, fabricated on a 16 nm process at TSMC. The GTX 680 uses the Kepler architecture with the GK104 chip, on a 28 nm process. The process node shrink alone is significant, allowing the 1070 Ti to pack 7,200 million transistors into a 314 mm² die, yielding a transistor density of 22.9 million per mm². The 680 has 3,540 million transistors on a 294 mm² die, with a density of 12.0 million per mm². The 1070 Ti has more than double the transistor count in only a slightly larger die.

This transistor budget translates directly into execution resources. The 1070 Ti has 2,432 shading units, 152 texture mapping units (TMUs), and 64 render output units (ROPs). The 680 has 1,536 shading units, 128 TMUs, and 32 ROPs. The 1070 Ti has 58% more shading units, 19% more TMUs, and double the ROPs. The ROP count is particularly important for fill-rate-bound scenarios, and the pixel rate figures confirm this: the 1070 Ti achieves 107.7 GPixel/s versus the 680’s 33.86 GPixel/s.

Clock speeds also favor the newer card. The 1070 Ti has a base clock of 1607 MHz and a boost of 1683 MHz, while the 680 runs at 1006 MHz base and 1058 MHz boost. Combined with the higher core count, this yields a texture rate of 255.8 GTexel/s for the 1070 Ti versus 135.4 GTexel/s for the 680. The FP32 compute throughput is 8.186 TFLOPS for the 1070 Ti, more than 2.5 times the 680’s 3.250 TFLOPS. The 1070 Ti also has FP16 capability at 127.9 GFLOPS (a 1:64 ratio to FP32), which the 680 lacks entirely.

Memory is another major differentiator. The 1070 Ti has 8 GB of GDDR5 on a 256-bit bus, with a memory clock of 2002 MHz (8 Gbps effective), yielding a bandwidth of 256.3 GB/s. The 680 has 2 GB of GDDR5 on the same 256-bit bus, but at 1502 MHz (6 Gbps effective), for a bandwidth of 192.3 GB/s. The 1070 Ti has four times the capacity and 33% more bandwidth. The API support also differs: the 1070 Ti supports DirectX 12 (12_1) and Vulkan 1.4, while the 680 is limited to DirectX 12 (11_0) and Vulkan 1.2.175. Both support OpenGL 4.6.

Physical and power characteristics differ as well. The 1070 Ti has a TDP of 180 W and requires a single 8-pin power connector, while the 680 has a higher TDP of 195 W and requires two 6-pin connectors. Both are dual-slot cards, but the 1070 Ti is longer (267 mm vs 256 mm) and wider (40 mm vs 38 mm). The display outputs also differ, with the 1070 Ti offering 1x DVI, 1x HDMI 2.0, and 3x DisplayPort 1.4a, while the 680 has 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2.

FAQ

Q: Which card has a higher average benchmark score?

A: The GTX 1070 Ti has an average score of 14,277, which is 0.9% higher than the GTX 680’s average of 14,150. However, this includes different test sets; the head-to-head tests show a much larger gap.

Q: How much faster is the GTX 1070 Ti in Vulkan?

A: The GTX 1070 Ti scores 61,006 in Geekbench Vulkan, which is 245.7% higher than the GTX 680’s 17,646.

Q: What is the memory capacity difference?

A: The GTX 1070 Ti has 8 GB of GDDR5 memory, while the GTX 680 has 2 GB. Both use a 256-bit bus, but the 1070 Ti’s bandwidth is 256.3 GB/s versus 192.3 GB/s for the 680.

Q: Does the GTX 680 support the same DirectX version?

A: No. The GTX 1070 Ti supports DirectX 12 (12_1), while the GTX 680 is limited to DirectX 12 (11_0). Both support OpenGL 4.6, but Vulkan support differs (1.4 vs 1.2.175).

Q: Which card has a higher transistor density?

A: The GTX 1070 Ti has 22.9 million transistors per mm², compared to 12.0 million per mm² for the GTX 680. This is due to the 16 nm process versus the 28 nm process.

Q: What are the TDPs of these two cards?

A: The GTX 1070 Ti has a TDP of 180 W, while the GTX 680 has a TDP of 195 W. Both have a suggested PSU of 450 W, but the 680 uses two 6-pin connectors while the 1070 Ti uses one 8-pin.

The Verdict

The data supports an unambiguous conclusion: the GTX 1070 Ti is the superior card in every measurable head-to-head metric. It wins all three shared benchmarks with leads ranging from 180.9% to 245.7%. Its architectural advantages—more shading units, double the ROPs, higher clocks, more memory, and a smaller process node—are reflected in every performance figure. The GTX 680, while a capable card in its own era, is simply outclassed by a wide margin.

For a user choosing between these two, the GTX 1070 Ti is the clear pick for any workload involving Metal, OpenCL, or Vulkan. Its 8 GB memory capacity also makes it more suitable for modern texture-heavy workloads. The GTX 680’s only advantage lies in its lower TDP (195 W) and smaller physical footprint (256 mm length vs 267 mm), but these are minor physical characteristics that do not offset a 245.7% performance deficit in Vulkan.

The percentile rankings are close (56th vs 55th), but this is misleading due to the different benchmark suites. The GTX 680’s average score is buoyed by having only three tests, all of which are compute-oriented. The head-to-head data, which uses identical tests, is the definitive measure. The GTX 1070 Ti is the only rational choice based on the available data.

Specification Differences

| Specification | NVIDIA GeForce GTX 1070 Ti | NVIDIA GeForce GTX 680 |

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

| Architecture | Pascal | Kepler |

| Process Node | 16 nm | 28 nm |

| Transistors | 7,200 million | 3,540 million |

| Die Size | 314 mm² | 294 mm² |

| Transistor Density | 22.9M / mm² | 12.0M / mm² |

| Base Clock | 1607 MHz | 1006 MHz |

| Boost Clock | 1683 MHz | 1058 MHz |

| Memory Clock | 2002 MHz (8 Gbps effective) | 1502 MHz (6 Gbps effective) |

| Memory Size | 8 GB | 2 GB |

| Memory Type | GDDR5 | GDDR5 |

| Memory Bandwidth | 256.3 GB/s | 192.3 GB/s |

| Shading Units | 2432 | 1536 |

| TMUs | 152 | 128 |

| ROPs | 64 | 32 |

| Pixel Rate | 107.7 GPixel/s | 33.86 GPixel/s |

| Texture Rate | 255.8 GTexel/s | 135.4 GTexel/s |

| FP32 Performance | 8.186 TFLOPS | 3.250 TFLOPS |

| FP16 Performance | 127.9 GFLOPS (1:64) | N/A |

| TDP | 180 W | 195 W |

| Power Connectors | 1x 8-pin | 2x 6-pin |

| Display Outputs | 1x DVI, 1x HDMI 2.0, 3x DisplayPort 1.4a | 2x DVI, 1x HDMI 1.4a, 1x DisplayPort 1.2 |

| DirectX Support | 12 (12_1) | 12 (11_0) |

| Vulkan Support | 1.4 | 1.2.175 |

| Length | 267 mm (10.5 inches) | 256 mm (10.1 inches) |

| Width | 40 mm (1.6 inches) | 38 mm (1.5 inches) |

| Launch MSRP | 399 USD | 499 USD |