AMD Radeon 680M vs AMD Radeon RX 580 Comparison

Head-to-Head Benchmarks

The recorded data shows a clear and consistent advantage for the AMD Radeon RX 580 across every shared benchmark. In the 3DMark Steel Nomad DX12 test, the RX 580 scores 1005, while the Radeon 680M manages only 378. That is a 62.4% deficit for the integrated part, the largest gap in the comparison. The RX 580 leads by more than two and a half times in this modern DirectX 12 workload, which stresses raw rasterization throughput and memory bandwidth.

The Geekbench compute results tell a similar story, though the margins narrow slightly. In the OpenCL test, the RX 580 scores 37453 against the 680M's 23468, a 37.3% lead. The Vulkan test shows an even larger spread: the RX 580 posts 45173, while the 680M reaches 21965, a 51.4% difference. Across all three head-to-head benchmarks, the RX 580 wins every round. The 680M records zero wins in the direct comparison.

Context from the database's nearest rival data reinforces the positioning. The 680M's average benchmark score of 15270 places it within 0.1% of the NVIDIA GeForce GTX 580 and RTX 2060, and 0.7% above the AMD Radeon RX 7600. The RX 580's average of 12928 sits 0.4% below the GTX 1660 SUPER and 0.5% above the AMD Radeon 740M. Both parts occupy similar percentile territory, with the 680M at the 57th percentile of all GPUs and the RX 580 at the 53rd. The direct head-to-head deltas, however, are unambiguous: the discrete card is substantially faster in every measured scenario.

Where Each One Wins

The AMD Radeon RX 580 wins in outright performance. Its 6.175 TFLOPS of FP32 compute is nearly double the 680M's 3.379 TFLOPS. Texture throughput also favors the discrete card heavily: 193.0 GTexel/s versus 105.6 GTexel/s. The RX 580's 256.0 GB/s of dedicated GDDR5 bandwidth is a decisive advantage over the 680M's system-shared memory, whose bandwidth is marked as system dependent. In any workload that saturates memory, such as high-resolution textures or compute-heavy shaders, the RX 580 should pull further ahead.

The Radeon 680M wins in efficiency and integration. Its 50 W TDP is a fraction of the RX 580's 185 W, and it requires no power connectors or suggested PSU rating. The RX 580 needs a 450 W suggested PSU and a single 8-pin connector. The 680M is an IGP, meaning it occupies no slot width and produces no add-in-card footprint. It also supports newer API features: DirectX 12 Ultimate (12_2) versus the RX 580's DirectX 12 (12_0), and Vulkan 1.4 versus 1.3. The 680M has 12 ray tracing cores, while the RX 580 has none. For a compact or low-power build where maximum frame rates are not the priority, the 680M is the only viable choice between these two.

Architecture Differences

The two GPUs come from different design eras and foundries. The Radeon 680M uses the Rembrandt+ chip on TSMC's 6 nm process, packing 13,100 million transistors into a 208 mm² die. That yields a transistor density of 63.0 million per square millimeter. The RX 580 uses the Polaris 20 chip on GlobalFoundries' 14 nm process, with 5,700 million transistors on a 232 mm² die, for a density of 24.6 million per square millimeter. The 680M's architecture is RDNA 2.0, while the RX 580 is GCN 4.0. The 680M belongs to the Navi II IGP generation (Rembrandt Mobile); the RX 580 is part of the Polaris (RX 500) generation.

Their compute configurations differ substantially. The RX 580 has 2304 shading units, 144 texture mapping units, and 32 ROPs. The 680M has 768 shading units, 48 TMUs, and 32 ROPs. Despite having three times the shader count, the RX 580's pixel rate is lower at 42.88 GPixel/s versus 70.40 GPixel/s for the 680M, a result of the integrated part's much higher boost clock of 2200 MHz versus 1340 MHz. The RX 580 compensates with a texture rate of 193.0 GTexel/s against 105.6 GTexel/s.

Memory architecture is a fundamental split. The 680M uses system-shared memory with a system-shared bus width and system-dependent bandwidth. The RX 580 has 8 GB of dedicated GDDR5 on a 256-bit bus, delivering 256.0 GB/s. The FP16 throughput also differs: the 680M achieves 6.758 TFLOPS at a 2:1 ratio, while the RX 580 is capped at 6.175 TFLOPS at a 1:1 ratio. The RX 580 is an end-of-life product from April 2017, whereas the 680M is active and launched in January 2023. The 680M uses a PCIe 4.0 x8 interface; the RX 580 uses PCIe 3.0 x16.

FAQ

Q: Which GPU is faster in the 3DMark Steel Nomad DX12 test?

A: The AMD Radeon RX 580 scores 1005 versus 378 for the Radeon 680M, a 62.4% advantage for the RX 580.

Q: Does the Radeon 680M support ray tracing?

A: Yes, the 680M includes 12 ray tracing cores and supports DirectX 12 Ultimate (12_2). The RX 580 has no ray tracing cores and only supports DirectX 12 (12_0).

Q: How much memory does each GPU have?

A: The RX 580 has 8 GB of dedicated GDDR5 on a 256-bit bus with 256.0 GB/s bandwidth. The 680M uses system-shared memory with a system-shared bus width and system-dependent bandwidth.

Q: Which GPU requires less power?

A: The 680M has a 50 W TDP and needs no power connectors. The RX 580 has a 185 W TDP and requires a 450 W suggested PSU plus a single 8-pin connector.

Q: What is the average benchmark score difference between the two?

A: The 680M has an average benchmark score of 15270, while the RX 580 averages 12928. The RX 580 still wins every direct head-to-head test, but the averages reflect different benchmark suites.

Q: Which GPU is more recent?

A: The 680M launched in January 2023 and is still active. The RX 580 launched in April 2017 and is end-of-life.

Specification Differences

| Specification | AMD Radeon 680M | AMD Radeon RX 580 |

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

| Architecture | RDNA 2.0 | GCN 4.0 |

| Process node | 6 nm (TSMC) | 14 nm (GlobalFoundries) |

| Transistors | 13,100 million | 5,700 million |

| Die size | 208 mm² | 232 mm² |

| Transistor density | 63.0M / mm² | 24.6M / mm² |

| Base clock | 2000 MHz | 1257 MHz |

| Boost clock | 2200 MHz | 1340 MHz |

| Memory size | System Shared | 8 GB GDDR5 |

| Memory bus width | System Shared | 256 bit |

| Memory bandwidth | System Dependent | 256.0 GB/s |

| Shading units | 768 | 2304 |

| TMUs | 48 | 144 |

| ROPs | 32 | 32 |

| RT cores | 12 | None |

| Pixel rate | 70.40 GPixel/s | 42.88 GPixel/s |

| Texture rate | 105.6 GTexel/s | 193.0 GTexel/s |

| FP32 | 3.379 TFLOPS | 6.175 TFLOPS |

| FP16 | 6.758 TFLOPS (2:1) | 6.175 TFLOPS (1:1) |

| TDP | 50 W | 185 W |

| Slot width | IGP | Dual-slot |

| Power connectors | None | 1x 8-pin |

| Suggested PSU | None | 450 W |

| Bus interface | PCIe 4.0 x8 | PCIe 3.0 x16 |

| DirectX | 12 Ultimate (12_2) | 12 (12_0) |

| Vulkan | 1.4 | 1.3 |

| Production status | Active | End-of-life |

| Release date | 2023-01-02 | 2017-04-17 |

| Launch MSRP | None | 229 USD |

The Verdict

The AMD Radeon RX 580 is the performance pick. Every direct benchmark in the database shows it ahead, often by wide margins: 62.4% in 3DMark Steel Nomad DX12, 51.4% in Geekbench Vulkan, and 37.3% in Geekbench OpenCL. Its dedicated 8 GB GDDR5 frame buffer with 256.0 GB/s bandwidth and 6.175 TFLOPS of FP32 compute make it the stronger choice for gaming and compute workloads where memory bandwidth is critical. The caveat is power: 185 W TDP, a 450 W suggested PSU, and a dual-slot card footprint.

The AMD Radeon 680M is the efficiency and feature pick. At 50 W with no power connectors and an IGP form factor, it fits into systems where a discrete card cannot. It brings modern architecture benefits: RDNA 2.0, DirectX 12 Ultimate with ray tracing support, Vulkan 1.4, and a 6 nm process. Its 2200 MHz boost clock and 70.40 GPixel/s pixel rate are surprisingly competitive, but the system-shared memory and lower shader count limit its ceiling.

Choose the RX 580 if you need maximum frame rates, can accommodate a 241 mm dual-slot card, and have a PSU rated for at least 450 W. Choose the 680M if you prioritize low power, small footprint, and access to ray tracing and DirectX 12 Ultimate features, and if your workloads fit within system-shared memory limits. The data does not support a scenario where the 680M outperforms the RX 580 in raw speed, but the two serve fundamentally different build contexts.