AMD Radeon 660M vs AMD Radeon 680M Comparison

Head-to-Head Benchmarks

The recorded data shows a decisive advantage for the AMD Radeon 680M across the two shared benchmark tests. In Geekbench OpenCL, the 680M scores 23,468 points against 12,876 for the 660M, a delta of 82.3%. This is the largest performance gap between the two integrated GPUs in the database. The Vulkan test shows a narrower but still substantial lead: 21,965 versus 14,748, a 48.9% advantage for the 680M.

These are not marginal differences. The 680M's OpenCL result is nearly double that of the 660M, which indicates a fundamental shift in compute capability rather than a simple clock speed bump. The Vulkan gap, while smaller, still places the 680M firmly ahead in graphics workloads that rely on modern API efficiency.

Looking at the broader database context, the 680M's average benchmark score of 15,270 places it at the 57th percentile of all GPUs. Its nearest rivals include the NVIDIA GeForce GTX 580 (15,283, a -0.1% delta), the NVIDIA GeForce RTX 2060 (15,290, -0.1%), the NVIDIA GeForce RTX 3050 OEM (15,199, +0.5%), and the AMD Radeon RX 7600 (15,171, +0.7%). This means the 680M sits in a performance band occupied by dedicated desktop graphics cards from several generations, despite being an integrated processor.

The 660M, by contrast, averages 13,812, which places it at the 55th percentile. Its nearest rivals are the NVIDIA RTX A2000 Mobile (13,821, -0.1%), the AMD Radeon RX 570X (13,871, -0.4%), the AMD Radeon RX 7900 XT (13,745, +0.5%), and the NVIDIA Tesla K10 (14,029, -1.5%). The 660M is therefore within 1.5% of several very different products, from a professional mobile GPU to a flagship desktop card, which reflects the fact that average scores compress a wide range of workloads into a single number.

The head-to-head data shows 2 wins for the 680M and 0 for the 660M. No benchmark in the shared set favors the smaller chip. The 680M's advantage in OpenCL (82.3%) is particularly striking because OpenCL tends to stress raw compute throughput, where the 680M's doubled shader count and higher clocks become decisive. The Vulkan result (48.9%) is more moderate, suggesting that API overhead and memory latency play a larger role in that test, but the 680M still wins clearly.

FAQ

Q: How much faster is the AMD Radeon 680M than the 660M in OpenCL?

A: The 680M scores 23,468 versus 12,876, a delta of 82.3%. This is the largest recorded performance gap between the two in any shared benchmark.

Q: Does the 660M win any benchmark against the 680M?

A: No. The database shows 2 wins for the 680M (Geekbench OpenCL and Geekbench Vulkan) and 0 wins for the 660M.

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

A: The 680M averages 15,270 across all recorded benchmarks, while the 660M averages 13,812. The 680M leads by approximately 1,458 points, or roughly 10.6% based on the 660M's average.

Q: How do these integrated GPUs compare to dedicated graphics cards?

A: The 680M's average score places it within 0.7% of the AMD Radeon RX 7600 and within 0.5% of the NVIDIA GeForce RTX 3050 OEM. The 660M sits within 0.5% of the AMD Radeon RX 7900 XT, though this is likely due to averaging effects across very different test conditions.

Q: Which GPU has the higher percentile ranking among all GPUs?

A: The 680M ranks at the 57th percentile, while the 660M ranks at the 55th percentile. The difference is small, but the 680M sits above the median in a more comfortable position.

Q: What is the Vulkan performance delta between the two?

A: The 680M scores 21,965 against 14,748 for the 660M, a 48.9% advantage. This is the smaller of the two head-to-head gaps, but it is still a substantial margin.

The Verdict

The data points to a clear choice for most users: the AMD Radeon 680M offers significantly higher performance in every shared benchmark. The 82.3% lead in OpenCL and 48.9% lead in Vulkan are not subtle differences; they represent a generational step within the same RDNA 2.0 architecture family. The 680M's average score of 15,270 versus 13,812 for the 660M reinforces this conclusion.

However, the 660M is not without merit. It draws 40 W versus 50 W for the 680M, a 10 W reduction that could matter in thin-and-light laptops where thermal headroom is scarce. The 660M also carries a lower base clock (1500 MHz versus 2000 MHz) and a lower boost clock (1900 MHz versus 2200 MHz), which means it will run cooler and potentially quieter under sustained load. For users who prioritize battery life and surface temperatures over raw frame rates, the 660M remains a viable option.

That said, the 680M's performance advantage is so pronounced that it should be the default pick for anyone who plans to game, render, or run compute workloads on an integrated GPU. The 57th percentile ranking means it outperforms the majority of all GPUs in the database, including several dedicated desktop cards. The 660M's 55th percentile is close, but the head-to-head results show that the 680M is simply in a different performance class.

The production status also matters: the 680M is listed as Active, while the 660M is End-of-life. This suggests that the 680M will continue to receive driver support and appear in current systems, whereas the 660M is being phased out. For a new purchase, the 680M is the safer long-term investment.

Specification Differences

The two GPUs share the same RDNA 2.0 architecture, 6 nm process node, TSMC foundry, and die size of 208 mm². Both have 13,100 million transistors and a transistor density of 63.0M per mm². The differences begin with the chip name: the 680M uses "Rembrandt+" while the 660M uses "Rembrandt."

Clock speeds differ substantially. The 680M has a base clock of 2000 MHz and a boost clock of 2200 MHz. The 660M runs at 1500 MHz base and 1900 MHz boost. This 500 MHz base clock gap and 300 MHz boost gap contribute directly to the performance delta.

The compute configuration doubles across the board. The 680M has 768 shading units, 48 TMUs, 32 ROPs, and 12 ray tracing cores. The 660M has 384 shading units, 24 TMUs, 16 ROPs, and 6 ray tracing cores. Every one of these counts is exactly double.

Pixel and texture rates reflect the doubled resources. The 680M achieves 70.40 GPixel/s and 105.6 GTexel/s. The 660M manages 30.40 GPixel/s and 45.60 GTexel/s. Floating point performance follows suit: 3.379 TFLOPS for the 680M versus 1,459.2 GFLOPS for the 660M in FP32, and 6.758 TFLOPS versus 2.918 TFLOPS in FP16.

Power draw differs by 10 W: 50 W for the 680M versus 40 W for the 660M. Both use the PCIe 4.0 x8 bus interface and have no power connectors. Display outputs are portable device dependent for both. Memory is system shared for both, with bandwidth described as system dependent.

Release dates place the 660M earlier: January 2022 for the 660M versus January 2023 for the 680M. Production status differs as noted: Active for the 680M, End-of-life for the 660M. Both share the same predecessor (Vega II IGP) and successor (Navi III IGP). Neither has a launch MSRP in the database.

Architecture Differences

Both GPUs are built on RDNA 2.0, but the 680M uses the "Rembrandt+" chip while the 660M uses the base "Rembrandt" chip. The process node is identical at 6 nm from TSMC, and the transistor count and die size are the same. The architectural difference is one of configuration, not fundamental design.

The 680M doubles the execution resources of the 660M. It has 768 shading units against 384, 48 TMUs against 24, 32 ROPs against 16, and 12 ray tracing cores against 6. This doubling is consistent across all compute units, which suggests that the 680M enables the full Rembrandt+ die while the 660M uses a partially disabled configuration.

Clock speeds add another layer of separation. The 680M's 2000 MHz base and 2200 MHz boost clocks are significantly higher than the 660M's 1500 MHz base and 1900 MHz boost. The combination of doubled resources and higher clocks explains the 82.3% OpenCL delta, which exceeds the 2x resource ratio due to the clock advantage.

Both GPUs support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. Neither has tensor cores. The memory subsystem is identical in type (system shared) but the bandwidth is system dependent, meaning the actual performance will vary based on the host laptop's memory configuration.

The 680M's higher TDP of 50 W versus 40 W for the 660M reflects the additional active hardware. This is a modest power increase for a large performance gain, which is typical of a fully enabled die versus a cut-down variant.

Where Each One Wins

The AMD Radeon 680M wins in raw performance across every shared benchmark. It is the choice for users who want the best possible integrated graphics experience for gaming, 3D rendering, or GPU-accelerated compute. Its 57th percentile ranking and proximity to dedicated GPUs like the RTX 2060 and RX 7600 mean it can handle modern titles at playable settings, provided the system has adequate memory bandwidth.

The 680M's 82.3% lead in OpenCL makes it particularly suited for compute workloads such as video encoding, machine learning inference, or physics simulations that leverage OpenCL. The 48.9% Vulkan lead is relevant for games and applications that use Vulkan, which includes many recent AAA titles and emulators.

The AMD Radeon 660M wins in power efficiency. Its 40 W TDP is 10 W lower than the 680M, which translates to less heat and longer battery life in portable devices. For users who rarely game or run heavy compute tasks, the 660M's lower power draw may be the deciding factor, especially in ultra-thin laptops with limited cooling.

The 660M also holds a release date advantage in one sense: it launched in January 2022, a year before the 680M, so it may be available in older, more affordable systems. However, its End-of-life status means it will not appear in new designs going forward.

In summary, the 680M is the performance leader and the safer future-proof choice. The 660M is a lower-power alternative that trades significant compute capability for a 10 W reduction in TDP. For any workload where GPU performance matters, the data favors the 680M without ambiguity.