AMD Radeon 660M vs NVIDIA GeForce RTX 3050 OEM Comparison

Head-to-Head Benchmarks

The recorded data shows a decisive, though lopsided, comparison between the NVIDIA GeForce RTX 3050 OEM and the AMD Radeon 660M. Across the two benchmark tests shared in the database, the NVIDIA part wins both, and the margins are substantial.

In Geekbench OpenCL, the RTX 3050 OEM scores 60,740 against the Radeon 660M's 12,876. That is a delta of 371.7% in favor of the NVIDIA card. This is not a close contest; the NVIDIA part delivers roughly 4.7 times the raw compute throughput in this API. The Geekbench Vulkan result follows the same pattern: the RTX 3050 OEM posts 57,103 while the Radeon 660M manages 14,748, a 287.2% advantage for NVIDIA. In both cases, the NVIDIA GPU is the clear winner, with the AMD integrated part trailing by a wide margin.

For context, the RTX 3050 OEM's average benchmark score sits at 15,199, which places it in the 57th percentile of all GPUs in the database. Its nearest rivals include the AMD Radeon RX 7600 at 15,171 (a 0.2% difference), the AMD Radeon 680M at 15,270 (0.5% higher), the NVIDIA GeForce GTX 580 at 15,283 (0.5% higher), and the NVIDIA GeForce RTX 2060 at 15,290 (0.6% higher). The RTX 3050 OEM is effectively in a dead heat with these parts, all within a single percentage point of each other. This means the NVIDIA OEM card is not an outlier in its class; it sits right in the middle of a tightly packed performance band.

The Radeon 660M, by contrast, has an average benchmark score of 13,812, placing it in the 55th percentile of all GPUs. Its nearest rivals are the NVIDIA RTX A2000 Mobile at 13,821 (0.1% higher), the AMD Radeon RX 570X at 13,871 (0.4% higher), the AMD Radeon RX 7900 XT at 13,745 (0.5% lower), and the NVIDIA Tesla K10 at 14,029 (1.5% higher). Interestingly, the Radeon 660M's average score is only about 9% lower than the RTX 3050 OEM's average, despite the massive gap in the two Geekbench tests. This discrepancy suggests that the 660M's performance profile is uneven across different workloads, while the RTX 3050 OEM is more consistently strong.

Looking at the individual benchmark suite, the RTX 3050 OEM shows a broad range of results. In Passmark, it scores 137 in DirectX 9, 86 in DirectX 11, 61 in DirectX 10, and 58 in DirectX 12. Its Passmark G3D score is 11,857, and its GPU compute score is 5,779. The 2D score is 973. These numbers indicate a part that handles older DirectX titles reasonably well but also has the architecture to support modern features. The Radeon 660M, however, has no Passmark results recorded in the database, so no comparison is possible for those specific workloads.

The wins tally is simple: the RTX 3050 OEM wins 2 out of 2 head-to-head tests, while the Radeon 660M wins none. The data does not show a single test where the AMD integrated GPU comes out ahead. That said, the average scores tell a slightly different story: the gap between the two parts' overall averages is much smaller than the Geekbench deltas would suggest. The RTX 3050 OEM's average of 15,199 versus the 660M's 13,812 is a 10% difference, which is still meaningful but nowhere near the 287% to 371% margins seen in the individual Geekbench tests.

What this means in practice is that the RTX 3050 OEM is a dedicated discrete GPU with a consistent performance level across the board, while the Radeon 660M is an integrated part that can be competitive in certain scenarios but falls far behind in compute-heavy API workloads. The Geekbench OpenCL and Vulkan tests are particularly demanding on raw shader throughput and memory bandwidth, areas where the NVIDIA card's dedicated GDDR6 memory and higher transistor count provide a clear edge.

The Verdict

The data points to two very different products for two very different purposes. The NVIDIA GeForce RTX 3050 OEM is the clear winner in raw performance, winning both head-to-head tests by margins of 287.2% and 371.7%. Its average benchmark score of 15,199 also beats the Radeon 660M's 13,812 by about 10%. For anyone who needs maximum compute performance in OpenCL or Vulkan workloads, the RTX 3050 OEM is the obvious choice based solely on these numbers.

The Radeon 660M, however, is not without its merits. It is an integrated GPU (IGP) with a 40 W TDP, compared to the RTX 3050 OEM's 130 W TDP. It requires no power connectors and is designed for portable devices, as indicated by its display outputs being "Portable Device Dependent." The RTX 3050 OEM, on the other hand, is a dual-slot card that requires a 1x 8-pin power connector and a 300 W suggested PSU, with a length of 242 mm (9.5 inches) and a height of 112 mm (4.4 inches). These physical and power requirements make the NVIDIA card unsuitable for compact or battery-powered systems.

Who should pick which? Strictly from the data, a user building a desktop with room for a discrete GPU and a sufficient power supply should choose the RTX 3050 OEM. It offers more than double the performance in the shared benchmarks and has a higher average score. A user constrained to a laptop or a small form-factor system with no room for a discrete card should choose the Radeon 660M, as it is the only option that fits that form factor. The 660M's 6 nm process node and lower power draw make it the sensible choice for mobile devices, even though its performance is significantly lower.

The percentile rankings reinforce this split: the RTX 3050 OEM is at the 57th percentile of all GPUs, while the Radeon 660M is at the 55th. Both are near the middle of the pack, but the NVIDIA card is slightly higher. The nearest rivals for each part also tell a story: the RTX 3050 OEM is flanked by the RX 7600, the 680M, the GTX 580, and the RTX 2060, all within 0.6% of its average score. The Radeon 660M is flanked by the RTX A2000 Mobile, RX 570X, RX 7900 XT, and Tesla K10, all within 1.5%. Neither part is an extreme outlier, but the RTX 3050 OEM is the stronger performer in absolute terms.

Architecture Differences

The architectural gap between these two GPUs is wide. The NVIDIA GeForce RTX 3050 OEM is built on the GA106 chip using the Ampere architecture, manufactured on an 8 nm process at Samsung. It packs 12,000 million transistors on a die size of 276 mm², giving a transistor density of 43.5 million per mm². The AMD Radeon 660M, by contrast, uses the Rembrandt chip with RDNA 2.0 architecture, built on a 6 nm process at TSMC. It has 13,100 million transistors on a smaller die of 208 mm², resulting in a higher transistor density of 63.0 million per mm². The AMD part is denser, but that does not translate into higher performance in the benchmarks.

The clock speeds differ as well. The RTX 3050 OEM has a base clock of 1515 MHz and a boost clock of 1755 MHz. The Radeon 660M has a base clock of 1500 MHz and a boost clock of 1900 MHz. The AMD part boosts higher, but the NVIDIA part has far more hardware resources to feed those clocks. The RTX 3050 OEM features 2304 shading units, 72 TMUs, 72 tensor cores, 18 ray tracing cores, and 32 ROPs. The Radeon 660M has 384 shading units, 24 TMUs, 16 ROPs, 6 ray tracing cores, and no tensor cores. The NVIDIA card has 6 times the shading units and 3 times the TMUs, which explains its dominance in texture-heavy and compute-heavy workloads.

Memory configurations are fundamentally different. The RTX 3050 OEM uses 8 GB of GDDR6 memory on a 128-bit bus, with a bandwidth of 224.0 GB/s and a memory clock of 1750 MHz (14 Gbps effective). The Radeon 660M uses system shared memory, with system-dependent bandwidth and bus width. In the Geekbench OpenCL test, the NVIDIA card's 224.0 GB/s dedicated bandwidth versus the 660M's system-dependent memory is likely a major factor in the 371.7% delta. The NVIDIA part also has a pixel rate of 56.6 GPixel/s and a texture rate of 126.4 GTexel/s, versus the 660M's 30.40 GPixel/s and 45.60 GTexel/s.

Compute output is similarly lopsided. The RTX 3050 OEM delivers 8.087 TFLOPS FP32 and 8.087 TFLOPS FP16 (1:1 ratio), while the Radeon 660M offers 1,459.2 GFLOPS FP32 (about 1.46 TFLOPS) and 2.918 TFLOPS FP16 (2:1 ratio). The NVIDIA card has more than 5 times the FP32 throughput. Both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, so feature-level API support is identical. The RTX 3050 OEM has 18 ray tracing cores, while the Radeon 660M has 6, but neither part shows a ray tracing benchmark in the database.

The interface and connectivity also differ. Both use PCIe 4.0 x8. The RTX 3050 OEM offers display outputs of 1x HDMI 2.1 and 3x DisplayPort 1.4a, while the Radeon 660M's outputs are portable-device dependent. The NVIDIA card is dual-slot with a 130 W TDP, 1x 8-pin power connector, and a 300 W suggested PSU. The AMD part is an IGP with 40 W TDP and no power connectors. The NVIDIA card is 242 mm long and 112 mm tall; the AMD part has no listed dimensions, consistent with its integrated nature.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA GeForce RTX 3050 OEM has an average benchmark score of 15,199, while the AMD Radeon 660M has an average score of 13,812. The NVIDIA part is about 10% higher.

Q: How big is the performance gap in Geekbench OpenCL?

A: The RTX 3050 OEM scores 60,740 in Geekbench OpenCL, while the Radeon 660M scores 12,876. That is a 371.7% advantage for NVIDIA.

Q: Does the Radeon 660M win any head-to-head benchmark?

A: No. In the database, the RTX 3050 OEM wins 2 out of 2 head-to-head tests, and the Radeon 660M wins 0.

Q: What are the power requirements for each GPU?

A: The RTX 3050 OEM has a 130 W TDP, requires a 1x 8-pin power connector, and has a suggested PSU of 300 W. The Radeon 660M has a 40 W TDP and no power connectors.

Q: Do both GPUs support the same APIs?

A: Yes. Both support DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4.

Q: What memory configurations do they use?

A: The RTX 3050 OEM uses 8 GB of GDDR6 on a 128-bit bus with 224.0 GB/s bandwidth. The Radeon 660M uses system shared memory with system-dependent bandwidth.

Where Each One Wins

The RTX 3050 OEM wins in every benchmark category where both parts have recorded results. In Geekbench OpenCL, it beats the Radeon 660M by 371.7%. In Geekbench Vulkan, it wins by 287.2%. Its average score of 15,199 is higher than the 660M's 13,812, and it sits at the 57th percentile versus the 660M's 55th. The NVIDIA card also has the advantage in raw specifications: more shading units (2304 versus 384), more TMUs (72 versus 24), more ROPs (32 versus 16), more ray tracing cores (18 versus 6), and 8.087 TFLOPS FP32 versus 1,459.2 GFLOPS. It has dedicated 8 GB GDDR6 memory with 224.0 GB/s bandwidth, while the 660M relies on system shared memory.

Where does the Radeon 660M win? It wins on efficiency and form factor. The 660M has a 40 W TDP versus 130 W for the RTX 3050 OEM. It requires no power connectors, while the NVIDIA card needs a 1x 8-pin connector and a 300 W PSU. The 660M is an IGP, so it fits in portable devices with no expansion slot, while the RTX 3050 OEM is a dual-slot card measuring 242 mm long and 112 mm tall. The 660M also uses a smaller die (208 mm² versus 276 mm²) on a more advanced 6 nm process, with a higher transistor density of 63.0M per mm² versus 43.5M per mm². It boosts to 1900 MHz, higher than the NVIDIA card's 1755 MHz boost, though it has far fewer cores to feed.

For use cases, the split is clear. The RTX 3050 OEM is the pick for desktop users who need maximum compute performance in OpenCL or Vulkan, have space for a dual-slot card, and can supply 300 W of power. The Radeon 660M is the pick for laptop or portable-device users who need a GPU with no dedicated power connector, no expansion slot, and minimal power draw. The data does not support any scenario where the 660M outperforms the RTX 3050 OEM in raw speed, but the 660M's integrated nature makes it the only viable option for ultra-portable systems. The RTX 3050 OEM is end-of-life, as is the Radeon 660M, and neither part has a listed launch MSRP. Both were released on the same date, 2022-01-03. The RTX 3050 OEM's predecessor is GeForce 20 and its successor is GeForce 40; the Radeon 660M's predecessor is Vega II IGP and its successor is Navi III IGP.