GPU Comparison

# AMD Radeon 660M vs AMD Radeon 740M

The AMD Radeon 660M and AMD Radeon 740M are both integrated graphics solutions from AMD, but they target different performance profiles and come from different architectural generations. The 660M is a Rembrandt-class IGP built on RDNA 2.0, while the 740M is a Phoenix2-based IGP using RDNA 3.0. Benchmark data shows a split decision: the 660M wins in OpenCL compute workloads with a 26.6% margin, while the 740M counters in Vulkan performance with a 5.3% lead. The average benchmark scores are remarkably close, 13,812 for the 660M versus 12,870 for the 740M, placing both in the mid-range of GPU performance overall.

Head-to-Head Benchmarks

The Geekbench OpenCL test shows a decisive win for the AMD Radeon 660M. It scores 12,876 against the 740M's 10,172, a delta of 26.6%. This is not a marginal difference; it represents a substantial advantage in compute-heavy tasks that rely on OpenCL. The 660M's higher shading unit count, 384 versus 256, likely drives this result, as does its higher base clock of 1500 MHz compared to the 740M's 800 MHz. In raw FP32 throughput, the 660M delivers 1,459.2 GFLOPS, while the 740M reaches 2,867 TFLOPS, yet the benchmark shows the older architecture outperforming in this specific workload. This suggests that OpenCL efficiency favors the RDNA 2.0 design's wider execution resources.

The Vulkan test flips the script. Here, the AMD Radeon 740M scores 15,568, edging out the 660M's 14,748 by 5.3%. This is a smaller margin than the OpenCL gap, but it indicates that the 740M's RDNA 3.0 architecture handles modern graphics APIs more effectively. The 740M's boost clock of 2800 MHz is substantially higher than the 660M's 1900 MHz, which helps compensate for its fewer cores. The Vulkan result positions the 740M as the better choice for gaming and graphics-heavy applications that leverage this API, even though the overall average benchmark score favors the 660M slightly.

Looking at overall performance, the 660M's average benchmark score of 13,812 places it at the 55th percentile of all GPUs, just 0.1% behind the NVIDIA RTX A2000 Mobile and 0.5% ahead of the AMD Radeon RX 7900 XT. The 740M's average of 12,870 sits at the 53rd percentile, nearly matching the AMD FirePro W5100 (0.2% ahead) and the AMD Radeon Pro 455 (0.3% ahead). These rival comparisons show both IGPs are competitive with dedicated desktop and mobile GPUs from several generations ago, though neither breaks into the top tier of performance.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The AMD Radeon 660M leads with an average benchmark score of 13,812, compared to the 740M's 12,870. This places the 660M at the 55th percentile versus the 740M's 53rd percentile.

Q: How do they compare in Vulkan performance?

A: The AMD Radeon 740M wins the Geekbench Vulkan test with a score of 15,568, beating the 660M's 14,748 by 5.3%. This suggests the 740M is better suited for modern graphics APIs.

Q: What is the biggest performance gap between them?

A: The largest difference appears in the Geekbench OpenCL test, where the 660M scores 12,876 against 10,172, a 26.6% advantage. This is the only test where the margin exceeds 10%.

Q: Are these GPUs similar to any dedicated graphics cards?

A: The 660M's average score is within 0.1% of the NVIDIA RTX A2000 Mobile and 0.5% ahead of the AMD Radeon RX 7900 XT. The 740M is 0.2% ahead of the AMD FirePro W5100 and 0.3% behind the AMD Radeon RX 580.

Q: Which GPU has better raw compute throughput?

A: The 740M has higher FP32 performance at 2.867 TFLOPS, compared to the 660M's 1,459.2 GFLOPS. However, the 660M wins the OpenCL benchmark, indicating that raw throughput doesn't always translate to benchmark wins.

Q: What is the production status of each GPU?

A: The AMD Radeon 660M is marked as end-of-life, while the AMD Radeon 740M is listed as active and currently in production.

Architecture Differences

The AMD Radeon 660M is built on the Rembrandt chip using RDNA 2.0 architecture, manufactured on TSMC's 6 nm process. It packs 13,100 million transistors into a 208 mm² die, giving a transistor density of 63.0M per mm². The 740M, in contrast, uses the Phoenix2 chip with RDNA 3.0 architecture on a 4 nm process. It contains 20,900 million transistors on a smaller 137 mm² die, resulting in a much higher density of 152.6M per mm². This newer process allows the 740M to fit nearly 60% more transistors in a significantly smaller area.

Core configuration differs substantially. The 660M has 384 shading units, 24 texture mapping units, 16 raster operation units, and 6 ray tracing cores. The 740M scales down to 256 shading units, 16 TMUs, 8 ROPs, and 4 ray tracing cores. Despite having fewer cores, the 740M achieves higher clock speeds, boost clock of 2800 MHz versus 1900 MHz, which helps it close the performance gap in certain workloads. The 740M also supports FP16 at a 1:1 ratio with FP32, whereas the 660M uses a 2:1 ratio, meaning the 740M can handle half-precision calculations more efficiently.

Both GPUs support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, so API compatibility is identical. Memory architecture is also the same: both use system-shared memory with type, bus width, and bandwidth all listed as system-dependent. This means actual memory performance varies based on the host system's RAM configuration, not the GPU itself.

Specification Differences

The most obvious specification gap is in process technology: the 660M uses 6 nm, while the 740M uses 4 nm. This translates to a significant difference in transistor density, 63.0M per mm² versus 152.6M per mm². The 660M's die size is 208 mm², while the 740M's is 137 mm², and transistor counts are 13,100 million and 20,900 million respectively.

Clock speeds differ dramatically. The 660M has a base clock of 1500 MHz and a boost clock of 1900 MHz. The 740M starts lower at 800 MHz base but boosts much higher to 2800 MHz. This 900 MHz boost advantage explains how the 740M can compete despite having fewer cores. Pixel rate favors the 660M at 30.40 GPixel/s versus 22.40 GPixel/s, while texture rates are closer: 45.60 GTexel/s for the 660M and 44.80 GTexel/s for the 740M. FP32 performance goes to the 740M at 2.867 TFLOPS, while the 660M delivers 1,459.2 GFLOPS.

Thermal design power differs slightly: the 660M is rated at 40 W, while the 740M is rated at 45 W. Both are IGP slot width, require no power connectors, and use a PCIe 4.0 x8 bus interface. Display outputs are portable device dependent for the 660M and motherboard dependent for the 740M. Release dates are nearly two years apart: January 2022 for the 660M and January 2024 for the 740M. The 660M's predecessor is the Vega II IGP and its successor is the Navi III IGP; the 740M's predecessor is the Navi II IGP and its successor is also Navi III IGP.

The Verdict

The benchmark data presents a nuanced picture. The AMD Radeon 660M wins the OpenCL test decisively by 26.6%, while the 740M takes the Vulkan test by 5.3%. The 660M has a higher average benchmark score (13,812 versus 12,870) and a better percentile ranking (55th versus 53rd). However, the 740M is the newer design with RDNA 3.0 architecture, a more advanced 4 nm process, and significantly higher boost clocks.

For users prioritizing compute workloads that leverage OpenCL, the 660M is the clear choice. Its 26.6% lead in that benchmark is substantial and unlikely to be overcome by the 740M's architectural advantages. The 660M also holds a slight edge in overall average score, making it the safer pick for mixed workloads. On the other hand, the 740M's Vulkan win suggests it handles modern graphics APIs better, which could translate to better gaming performance in titles that favor Vulkan.

The 740M's higher TDP (45 W versus 40 W) and newer release date (2024 versus 2022) indicate it is the more forward-looking option, but the data shows the 660M remains competitive or superior in key metrics. The 660M's production status is end-of-life, while the 740M is active, meaning the latter will be easier to source in new systems. Ultimately, the choice depends on workload: OpenCL-heavy tasks favor the 660M, while Vulkan-based graphics favor the 740M.

Where Each One Wins

AMD Radeon 660M wins in: OpenCL compute workloads, delivering a 26.6% higher score (12,876 versus 10,172). It also has a higher average benchmark score (13,812 versus 12,870) and better percentile ranking (55th versus 53rd). Its pixel rate of 30.40 GPixel/s is higher than the 740M's 22.40 GPixel/s, and its base clock of 1500 MHz is nearly double the 740M's 800 MHz. For users running legacy compute applications or OpenCL-dependent software, the 660M is the stronger performer.

AMD Radeon 740M wins in: Vulkan graphics workloads, with a 5.3% higher score (15,568 versus 14,748). It also has superior FP32 performance at 2.867 TFLOPS versus 1,459.2 GFLOPS, and a much higher boost clock of 2800 MHz compared to 1900 MHz. The 740M's newer RDNA 3.0 architecture and 4 nm process make it more power-efficient per transistor, and its higher TDP of 45 W allows for sustained performance. For gaming or graphics applications that use Vulkan, the 740M is the better choice, despite its lower core count.