NVIDIA GeForce GTX 1080 vs NVIDIA GeForce GTX 680 Comparison

Head-to-Head Benchmarks

The benchmark data is unambiguous: the NVIDIA GeForce GTX 1080 outperforms the NVIDIA GeForce GTX 680 in every recorded test. Across three cross-platform compute workloads, the GTX 1080 delivers margins ranging from roughly 42% to 67%, making this a decisive generational victory rather than a close contest.

The largest gap appears in the Geekbench OpenCL test. The GTX 680 records a score of 16,906, while the GTX 1080 reaches 51,204. That is a 67% advantage for the newer card, a delta that reflects fundamental changes in raw compute throughput. The GTX 1080's FP32 performance is listed at 8.873 TFLOPS versus 3.250 TFLOPS for the GTX 680, more than 2.7 times the theoretical floating-point output, and the OpenCL result tracks that hardware advantage closely.

Geekbench Metal shows a similar pattern, though slightly less extreme. The GTX 680 scores 7,897, while the GTX 1080 posts 23,824. The delta here is 66.9%, nearly identical to the OpenCL margin. Even in the closest head-to-head test, Geekbench Vulkan, the GTX 1080 wins by a substantial 42%: 30,398 versus 17,646. While Vulkan narrows the gap compared to the other two workloads, the GTX 680 still trails by a wide margin in absolute terms.

The GTX 1080 also appears in broader database results with additional workloads not shared by the GTX 680. For instance, the GTX 1080 records a Passmark G3D score of 15,586 and a Passmark GPU Compute score of 7,614. The GTX 680 has no corresponding entries in those tests, so a direct comparison is not possible, but the presence of these results shows the GTX 1080 has been characterized across a wider range of legacy and compute scenarios.

When placed against the wider field, the two cards occupy different positions in the database's percentile rankings. The GTX 680 sits at the 55th percentile among all GPUs, with an average benchmark score of 14,150. Its nearest rivals include the NVIDIA Tesla K10 (average score 14,029, a 0.9% gap) and the NVIDIA GeForce GTX 1070 Ti (average score 14,277, a 0.9% gap in the other direction). The GTX 1080, by contrast, sits at the 51st percentile with an average score of 11,960, with nearest rivals including the NVIDIA GeForce GTX 960A (average score 11,998, a 0.3% gap) and the AMD Radeon RX 6500 XT (average score 11,842, a 1% gap). This is an important nuance: even though the GTX 1080 crushes the GTX 680 in every shared test, the GTX 680's aggregate score across its recorded benchmarks happens to be higher, because the two cards are measured against different sets of workloads and the GTX 680's average is inflated by its Vulkan and OpenCL results relative to its Metal score.

FAQ

Q: Which card wins in Geekbench Metal?

A: The GTX 1080 wins decisively. It scores 23,824 versus 7,897 for the GTX 680, a 66.9% advantage.

Q: How large is the performance gap in OpenCL compute?

A: The GTX 1080 scores 51,204, while the GTX 680 scores 16,906. That is a 67% lead for the GTX 1080, the largest margin of any shared test.

Q: Does the GTX 680 win any benchmark against the GTX 1080?

A: No. In the three head-to-head tests recorded (Metal, OpenCL, and Vulkan), the GTX 680 wins zero tests and the GTX 1080 wins all three.

Q: Why does the GTX 680 have a higher percentile rank than the GTX 1080?

A: The GTX 680 sits at the 55th percentile with an average benchmark score of 14,150, while the GTX 1080 sits at the 51st percentile with an average score of 11,960. The percentile ranking reflects each card's average across its own set of recorded benchmarks, not the head-to-head results; the GTX 680's average is driven up by its relatively strong Vulkan and OpenCL scores compared to its Metal score.

Q: What is the closest head-to-head result between the two cards?

A: The closest is Geekbench Vulkan, where the GTX 1080 wins by 42%. The GTX 680 scores 17,646 and the GTX 1080 scores 30,398. Even the smallest margin is still a large defeat.

Q: Are there any benchmarks where only the GTX 1080 has recorded scores?

A: Yes. The GTX 1080 has results for Passmark DirectX 9 (211), DirectX 10 (93), DirectX 11 (124), DirectX 12 (55), G2D (888), G3D (15,586), and GPU Compute (7,614). The GTX 680 has no entries for those tests in the database.

Where Each One Wins

The GTX 1080 wins everywhere in direct comparison. It takes all three shared benchmarks and does so with margins between 42% and 67%. If the workload is Metal-based, OpenCL compute, or Vulkan rendering, the GTX 1080 is the clear choice. The data also shows the GTX 1080 has been validated across a broader suite of legacy DirectX tests, including DirectX 9 through DirectX 12, plus 2D and compute workloads. That breadth of coverage suggests the GTX 1080 remains a more versatile option for users running older APIs or mixed workloads.

The GTX 680 has no benchmark wins in this comparison. Its only notable position is in the aggregate percentile ranking, where its average score of 14,150 exceeds the GTX 1080's 11,960, but that is an artifact of different benchmark sets rather than evidence of superiority. In any shared test, the GTX 680 loses. Therefore, there is no use case in the recorded data where the GTX 680 is the faster card.

Specification Differences

The two cards differ across nearly every hardware specification. Memory capacity jumps from 2 GB on the GTX 680 to 8 GB on the GTX 1080, a fourfold increase. Memory type also changes from GDDR5 to GDDR5X. The bus width stays at 256 bit for both, but memory bandwidth rises from 192.3 GB/s to 320.3 GB/s, a 66.5% improvement.

Clock speeds are substantially higher on the GTX 1080. The base clock is 1607 MHz versus 1006 MHz, and the boost clock is 1733 MHz versus 1058 MHz. Memory clock also differs: the GTX 680 runs at 1502 MHz with 6 Gbps effective, while the GTX 1080 runs at 1251 MHz with 10 Gbps effective, the higher effective rate enabled by GDDR5X.

The compute configuration is larger on the GTX 1080. Shading units increase from 1,536 to 2,560, texture mapping units from 128 to 160, and ROPs from 32 to 64. Pixel rate jumps from 33.86 GPixel/s to 110.9 GPixel/s, and texture rate from 135.4 GTexel/s to 277.3 GTexel/s. FP32 throughput rises from 3.250 TFLOPS to 8.873 TFLOPS, and the GTX 1080 adds FP16 capability at 138.6 GFLOPS (1:64), a feature the GTX 680 does not list.

Power and physical specifications also shift. The GTX 680 has a TDP of 195 W with dual 6-pin power connectors, while the GTX 1080 has a lower TDP of 180 W with a single 8-pin connector. Both cards are dual-slot and share a suggested PSU of 450 W. The GTX 1080 is slightly longer (267 mm versus 256 mm) and slightly thicker (40 mm versus 38 mm), with the same height of 112 mm versus 111 mm. The bus interface is PCIe 3.0 x16 for both.

Display outputs differ. The GTX 680 offers 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2. The GTX 1080 offers 1x DVI, 1x HDMI 2.0, and 3x DisplayPort 1.4a. API support also diverges: the GTX 680 is listed as DirectX 12 (11_0) with Vulkan 1.2.175, while the GTX 1080 is listed as DirectX 12 (12_1) with Vulkan 1.4. Both support OpenGL 4.6.

Architecture Differences

The GTX 680 is built on the Kepler architecture with the GK104 chip, fabricated on a 28 nm process at TSMC. The GTX 1080 uses the Pascal architecture with the GP104 chip, also at TSMC but on a 16 nm process. The transistor count nearly doubles: 3,540 million on the GTX 680 versus 7,200 million on the GTX 1080, despite a modest die size increase from 294 mm² to 314 mm². Transistor density consequently rises from 12.0M per mm² to 22.9M per mm², reflecting the process node shrink.

The generation gap is also clear in the product lineage. The GTX 680 belongs to the GeForce 600 generation with a predecessor in GeForce 500 and a successor in GeForce 700. The GTX 1080 belongs to the GeForce 10-series with a predecessor in GeForce 900 and a successor in GeForce 20. Neither card includes ray tracing cores or tensor cores, so the architectural differences are limited to the core layout, process, and feature set described above.

The GTX 1080 also supports a higher DirectX feature level (12_1 versus 11_0) and a newer Vulkan version (1.4 versus 1.2.175), which matters for compatibility with modern titles that rely on those APIs. The GTX 1080's FP16 support, even at a 1:64 ratio, is an additional architectural capability absent from the GTX 680.

The Verdict

The data is one-sided. The GTX 1080 wins every shared benchmark by a margin no smaller than 42%, and in two of three tests the margin exceeds 66%. It offers four times the memory capacity, nearly double the bandwidth, more than double the FP32 throughput, and a newer architecture on a smaller process node. The GTX 680's only statistical advantage is its higher aggregate percentile rank, which stems from its unique benchmark set and does not reflect any head-to-head victory.

For users prioritizing compute performance, modern API support, or high-resolution texture workloads, the GTX 1080 is the only defensible choice from the recorded data. The GTX 680 remains a functional card for legacy DirectX 11-era workloads, but it cannot match the GTX 1080 in any measured scenario. The verdict is straightforward: the GTX 1080 is the superior product, and the GTX 680 should only be considered if the workload is confined to the GTX 680's specific benchmark profile, where no competing GTX 1080 result exists to contradict its standing.