AMD Radeon RX 570X vs NVIDIA GeForce GTX 680 Comparison

NVIDIA GeForce GTX 680 and AMD Radeon RX 570X land within 2% of each other in average benchmark score, yet their individual test results reveal starkly different strengths. The GTX 680, a Kepler-generation part from 2012, posts an average of 14150 across three Geekbench compute tests, while the RX 570X, a Polaris refresh from 2018, averages 13871 across a mix of Geekbench and Passmark tests. Both cards sit at the 55th percentile among all GPUs, placing them in the same performance tier overall, but the data shows they achieve that parity through entirely different architectural approaches and workload specialties.

Where Each One Wins

The RX 570X is the clear winner in raw compute throughput, as demonstrated by the single head-to-head benchmark available. In Geekbench OpenCL, the AMD card scores 38939 against the GTX 680's 16906, a massive 56.6% advantage. This suggests the RX 570X excels in general-purpose compute workloads that leverage OpenCL, such as rendering, physics simulation, or data-parallel tasks. The GTX 680's compute capabilities are far more modest in this specific test, which may reflect its older Kepler architecture's weaker OpenCL implementation relative to GCN 4.0.

The GTX 680, however, holds its own in the broader benchmark suite by posting competitive scores across Geekbench Metal, OpenCL, and Vulkan tests. Its Geekbench Vulkan score of 17646 is its highest, indicating reasonable modern API performance despite its age. The RX 570X lacks a Vulkan or Metal score in the data, so its API-specific performance in those areas remains unmeasured. This creates an asymmetric comparison: the AMD card wins decisively where tested head-to-head, but the NVIDIA card demonstrates broader API coverage across the available data points.

The wins break down as 0 for the GTX 680 and 1 for the RX 570X in the head-to-head table, but the average scores tell a more nuanced story. The GTX 680's average of 14150 is actually 2% higher than the RX 570X's 13871, meaning the NVIDIA card performs better across the aggregate of all its tested workloads. This suggests that while the RX 570X dominates in pure OpenCL compute, the GTX 680 may be more balanced or better optimized for the specific mix of tests it underwent.

Architecture Differences

The two cards come from fundamentally different design philosophies. The GTX 680 uses the GK104 chip built on TSMC's 28 nm process, packing 3,540 million transistors into a 294 mm² die with a transistor density of 12.0M per mm². The RX 570X uses the Polaris 20 chip fabricated by GlobalFoundries on a more advanced 14 nm node, containing 5,700 million transistors in a smaller 232 mm² die, yielding a much higher density of 24.6M per mm². The RX 570X crams 61% more transistors into a 21% smaller area, reflecting the generational leap in manufacturing between 2012 and 2018.

The compute resources differ substantially. The RX 570X has 2048 shading units versus the GTX 680's 1536, a 33% advantage in raw shader count. Both cards have 128 texture mapping units and 32 ROPs, but the AMD card's higher clocks push its fill rates ahead: 39.81 GPixel/s pixel rate and 159.2 GTexel/s texture rate versus 33.86 GPixel/s and 135.4 GTexel/s on the GTX 680. The RX 570X also delivers 5.095 TFLOPS of FP32 performance with 1:1 FP16 support, while the GTX 680 manages only 3.250 TFLOPS of FP32 with no FP16 capability listed.

Memory configuration is another major divider. The GTX 680 has 2 GB of GDDR5 on a 256-bit bus with 192.3 GB/s bandwidth, while the RX 570X has 8 GB of GDDR5 on the same 256-bit bus but with 224.0 GB/s bandwidth due to its 7 Gbps effective memory clock versus 6 Gbps. The RX 570X's fourfold capacity increase and 16.5% bandwidth advantage give it a clear edge in memory-intensive workloads and higher-resolution textures. The RX 570X also draws less power at 150 W TDP versus 195 W, despite its higher performance, thanks to the more efficient 14 nm process.

Head-to-Head Benchmarks

The single direct comparison available is Geekbench OpenCL, and it is a landslide. The RX 570X scores 38939, which is 56.6% higher than the GTX 680's 16906. This delta is enormous — the AMD card delivers more than double the OpenCL compute performance of the NVIDIA card. To put this in context, the GTX 680's nearest rivals include the NVIDIA Tesla K10 at 14029 (0.9% behind) and the GeForce GTX 1070 Ti at 14277 (0.9% ahead), meaning the GTX 680's OpenCL score is roughly in line with its average performance. The RX 570X, by contrast, produces an OpenCL score that is nearly triple its own average of 13871, suggesting this specific workload plays directly to the GCN architecture's strengths.

This imbalance raises questions about the other benchmarks. The GTX 680's Geekbench Vulkan score of 17646 and Metal score of 7897 indicate that its compute performance varies significantly by API. The RX 570X's Passmark scores of 750 in G2D and 1923 in G3D show a different performance profile, with the G3D score being particularly low relative to its OpenCL output. These divergent results imply that benchmark choice dramatically influences which card appears superior, and the data does not provide a unified picture of overall gaming or compute capability.

The deltaPct of -56.6% in the head-to-head table is the largest gap recorded, and it drives the win count. However, the average benchmark scores suggest that if more tests were available, the GTX 680 might close the gap or even overtake the RX 570X in certain workloads. The GTX 680's Vulkan score of 17646, for instance, would likely be competitive with or exceed the RX 570X's performance in Vulkan-based applications, but no such comparison exists in the data.

FAQ

Q: Which card has the higher average benchmark score?

A: The NVIDIA GeForce GTX 680 averages 14150, which is 2% higher than the AMD Radeon RX 570X's 13871. Both cards sit at the 55th percentile among all GPUs.

Q: How large is the RX 570X's lead in Geekbench OpenCL?

A: The RX 570X scores 38939 compared to the GTX 680's 16906, a 56.6% advantage. This is the only direct head-to-head benchmark available in the data.

Q: What are the memory capacity differences?

A: The RX 570X has 8 GB of GDDR5 memory, while the GTX 680 has only 2 GB. Both use a 256-bit bus, but the RX 570X offers 224.0 GB/s bandwidth versus 192.3 GB/s on the GTX 680.

Q: How do their power requirements compare?

A: The RX 570X has a 150 W TDP with a single 6-pin power connector, while the GTX 680 has a 195 W TDP with dual 6-pin connectors. Both recommend a 450 W power supply.

Q: Which card has more shading units and higher FP32 performance?

A: The RX 570X has 2048 shading units and delivers 5.095 TFLOPS of FP32 compute, while the GTX 680 has 1536 shading units and 3.250 TFLOPS. The RX 570X also supports FP16 at a 1:1 ratio, which the GTX 680 does not.

Q: What are the API support differences?

A: The RX 570X supports DirectX 12 (12_0) and Vulkan 1.3, while the GTX 680 supports DirectX 12 (11_0) and Vulkan 1.2.175. Both support OpenGL 4.6.

The Verdict

The data points to a clear split: the AMD Radeon RX 570X is the superior choice for OpenCL compute workloads, offering more than double the performance of the GTX 680 in that specific test. Its 8 GB memory capacity, higher bandwidth, and 1:1 FP16 support make it better equipped for modern data-parallel tasks and larger datasets. The RX 570X also consumes less power at 150 W versus 195 W, despite delivering 56.6% higher compute throughput in the head-to-head test.

The NVIDIA GeForce GTX 680, however, shows a slight edge in average benchmark score across all its tested workloads, suggesting it may be more consistent or better optimized for the specific mix of Geekbench tests it underwent. Its Vulkan score of 17646 indicates competent modern API performance, and its 2 GB memory might suffice for older or less demanding applications. The GTX 680's launch MSRP was 499 USD, which provides context for its original positioning, though no pricing data exists for the RX 570X.

For users prioritizing raw compute performance in OpenCL-heavy applications, the RX 570X is the obvious pick based on the data. For those who value broader API coverage and a higher average score across mixed workloads, the GTX 680 edges ahead. The 55th percentile ranking for both cards means neither is a top-tier performer, but the RX 570X's architectural advantages in memory and compute make it the more future-proof option despite its lower average score.

Specification Differences

| Specification | NVIDIA GeForce GTX 680 | AMD Radeon RX 570X |

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

| Chip | GK104 | Polaris 20 |

| Architecture | Kepler | GCN 4.0 |

| Process Node | 28 nm (TSMC) | 14 nm (GlobalFoundries) |

| Transistors | 3,540 million | 5,700 million |

| Die Size | 294 mm² | 232 mm² |

| Transistor Density | 12.0M / mm² | 24.6M / mm² |

| Base Clock | 1006 MHz | 1168 MHz |

| Boost Clock | 1058 MHz | 1244 MHz |

| Memory Clock | 6 Gbps effective | 7 Gbps effective |

| Memory Size | 2 GB | 8 GB |

| Memory Type | GDDR5 | GDDR5 |

| Bandwidth | 192.3 GB/s | 224.0 GB/s |

| Shading Units | 1536 | 2048 |

| Pixel Rate | 33.86 GPixel/s | 39.81 GPixel/s |

| Texture Rate | 135.4 GTexel/s | 159.2 GTexel/s |

| FP32 Performance | 3.250 TFLOPS | 5.095 TFLOPS |

| FP16 Performance | N/A | 5.095 TFLOPS (1:1) |

| TDP | 195 W | 150 W |

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

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

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

| Vulkan Support | 1.2.175 | 1.3 |

| Release Date | 2012-03-21 | 2018-04-10 |

| Length | 256 mm (10.1 inches) | 241 mm (9.5 inches) |

| Width | 38 mm (1.5 inches) | Not specified |