AMD Radeon 660M vs NVIDIA GeForce GTX 960A Comparison

Head-to-Head Benchmarks

The only direct benchmark comparison recorded in the database is the Geekbench OpenCL test, and the AMD Radeon 660M takes the win. The AMD part scores 12,876 points against 11,998 for the NVIDIA GeForce GTX 960A, a delta of 7.3%. That is a clear, if not overwhelming, margin: roughly three-quarters of a standard performance step in the database's distribution.

Context from the nearest rivals helps clarify what that 7.3% gap means. The Radeon 660M's average benchmark score of 13,812 places it within 0.1% of the NVIDIA RTX A2000 Mobile (13,821) and 0.4% of the AMD Radeon RX 570X (13,871). It actually sits 0.5% above the AMD Radeon RX 7900 XT in this particular ranking, though that delta is within noise. The Radeon 660M is a 55th percentile GPU overall. The GTX 960A, with an average score of 11,998, lands at the 51st percentile and sits 0.3% above the GeForce GTX 1080 (11,960), 1.3% above the Radeon RX 6500 XT (11,842), 2.7% above the GeForce GTX 1660 (11,680), and 3.2% above the Radeon RX 7800 XT (11,627). So the 960A is not a weak GPU in absolute terms; it is simply one step lower in the performance hierarchy.

The single OpenCL result is the only head-to-head data point, so the verdict on overall performance rests on that one test. The 660M wins the only recorded comparison, and the database shows one win for AMD and zero for NVIDIA. There is no Vulkan result for the GTX 960A, despite the 660M posting a 14,748 Geekbench Vulkan score. That absence means the OpenCL test carries the full weight of the comparison. The 7.3% advantage in OpenCL is consistent with the 660M's higher percentile ranking and its proximity to much newer discrete GPUs in the rival list.

Notably, the 660M's average benchmark score (13,812) is 15.1% higher than the 960A's average (11,998). That gap is larger than the single OpenCL delta because the 660M's average includes its Vulkan result, which is substantially higher than its OpenCL score. The 960A has no equivalent secondary test in the database, so its average is identical to its OpenCL score. The data indicates the 660M is the faster part in every measured dimension.

The Verdict

The benchmark data points to a straightforward choice for most use cases. The AMD Radeon 660M wins the only direct comparison, posts a higher average score, and holds a higher percentile rank. Anyone selecting between these two for compute workloads measured by Geekbench OpenCL should pick the 660M. The 7.3% lead is not enormous, but it is consistent and backed by a second data point (Vulkan) where the 660M performs even better.

The GTX 960A is not without merit. Its 11,998 OpenCL score is respectable, and it outperforms several modern discrete GPUs in the database's rival list, including the GeForce GTX 1080 and Radeon RX 6500 XT by small margins. But in a direct comparison, it trails the 660M, and it lacks the additional API coverage that the 660M demonstrates. The 960A's average score is 1,814 points lower than the 660M's average, a 13.1% deficit.

For users constrained to the GTX 960A's platform, such as legacy MXM-based systems, the data shows it remains a capable performer relative to other entries in its neighborhood. But given a free choice, the 660M is the stronger option based on recorded measurements. The 660M also offers a more modern feature set, which is discussed below, and that reinforces the performance data.

Architecture Differences

The two GPUs come from different architectural generations and different design philosophies. The AMD Radeon 660M is built on RDNA 2.0, specifically the Rembrandt chip, and belongs to the Navi II IGP generation for Rembrandt Mobile. It is an integrated graphics processor on a 6 nm TSMC process. The GTX 960A uses NVIDIA's Maxwell architecture, built around the GM107 chip, and belongs to the GeForce 900A generation. It is fabricated on a 28 nm TSMC process.

The process node difference is stark: 6 nm versus 28 nm. That translates into a massive transistor density gap. The 660M packs 13,100 million transistors into a 208 mm² die, yielding 63.0 million transistors per square millimeter. The 960A contains 1,870 million transistors on a 148 mm² die, for 12.6 million per square millimeter. The 660M has roughly seven times the transistor density, which reflects the generational jump in manufacturing technology.

Core counts move in opposite directions. The 660M has 384 shading units, 24 texture mapping units, and 16 render output units. It also includes 6 ray tracing cores, a feature entirely absent from the 960A. The GTX 960A has 640 shading units and 40 texture mapping units, but also 16 ROPs. So the NVIDIA part fields more raw shader and texture hardware, yet the AMD part still wins the benchmark. The reason likely lies in clock speeds and memory architecture, discussed below.

The 660M is a RDNA 2.0 part with DirectX 12 Ultimate (12_2) support, while the 960A is Maxwell with DirectX 12 (11_0) support. Both support OpenGL 4.6 and Vulkan 1.4, so API coverage is similar outside of DirectX. The 660M also has a Vulkan benchmark score in the database (14,748), while the 960A has none recorded. That suggests the 660M's driver stack and hardware are better positioned for modern compute APIs.

Memory architecture differs fundamentally. The 660M uses system-shared memory, with type, bus width, and bandwidth all dependent on the host system. The 960A has dedicated 2 GB of GDDR5 on a 128-bit bus, delivering 80.19 GB/s of bandwidth. The 660M's memory bandwidth is listed as "System Dependent," meaning it varies by platform. In practice, shared memory can be a bottleneck, but the 660M's higher clocks and newer architecture compensate in the recorded benchmarks.

Pixel and texture rates reflect the clock and core differences. The 660M achieves 30.40 GPixel/s and 45.60 GTexel/s. The 960A manages 18.82 GPixel/s and 47.04 GTexel/s. The 660M leads in pixel fill rate by a wide margin, while the 960A has a slight edge in texture fill rate. The 660M's FP32 throughput is 1,459.2 GFLOPS, close to the 960A's 1.505 TFLOPS (1,505 GFLOPS). The 660M also supports FP16 at 2.918 TFLOPS (2:1 rate), which the 960A does not list at all.

Specification Differences

The two parts differ across nearly every specification field. The process nodes are 6 nm for the 660M and 28 nm for the 960A, both from TSMC. Transistor counts are 13,100 million versus 1,870 million. Die sizes are 208 mm² versus 148 mm². Transistor density is 63.0M / mm² versus 12.6M / mm².

Clock speeds differ. The 660M has a base clock of 1500 MHz and a boost clock of 1900 MHz. The 960A has a base clock of 1097 MHz and a boost clock of 1176 MHz. Memory clocks are not directly comparable: the 660M uses system-shared memory, while the 960A runs its GDDR5 at 1253 MHz (5 Gbps effective).

Memory specifications diverge completely. The 660M has system-shared memory size, type, bus width, and bandwidth. The 960A has 2 GB of GDDR5, a 128-bit bus, and 80.19 GB/s of bandwidth. The 660M's bandwidth is system dependent.

Shading units: 384 versus 640. TMUs: 24 versus 40. ROPs: 16 versus 16, a tie. Ray tracing cores: 6 on the 660M, none on the 960A. Pixel rates: 30.40 GPixel/s versus 18.82 GPixel/s. Texture rates: 45.60 GTexel/s versus 47.04 GTexel/s. FP32: 1,459.2 GFLOPS versus 1.505 TFLOPS. FP16: 2.918 TFLOPS on the 660M, not listed for the 960A.

Thermal design power differs: 40 W for the 660M versus 75 W for the 960A. The 660M is an IGP with no power connectors, while the 960A is an MXM module with no power connectors. Bus interfaces: PCIe 4.0 x8 for the 660M, MXM-B (3.0) for the 960A. Both have portable-device-dependent display outputs.

Release dates are far apart: the 660M launched January 3, 2022, while the 960A launched March 12, 2015. The 660M's predecessor is the Vega II IGP and its successor is the Navi III IGP. The 960A's predecessor is the GeForce 800A and it has no recorded successor. Both are end-of-life production status. Neither has a launch MSRP in the database.

FAQ

Q: Which GPU is faster in the recorded benchmarks?

A: The AMD Radeon 660M wins the only head-to-head test, Geekbench OpenCL, with 12,876 points versus 11,998 for the GTX 960A, a 7.3% advantage. Its average benchmark score is 13,812 versus 11,998.

Q: Does the GTX 960A win any benchmark comparison?

A: No. The database records one head-to-head benchmark, and the 660M wins it. The win tally is 1 for the 660M and 0 for the 960A.

Q: How does each GPU compare to its nearest rivals?

A: The 660M is within 0.1% of the RTX A2000 Mobile (13,821), 0.4% of the RX 570X (13,871), and 0.5% above the RX 7900 XT (13,745). The 960A sits 0.3% above the GTX 1080 (11,960), 1.3% above the RX 6500 XT (11,842), 2.7% above the GTX 1660 (11,680), and 3.2% above the RX 7800 XT (11,627).

Q: What are the biggest architectural differences?

A: The 660M uses RDNA 2.0 on a 6 nm process with 13,100 million transistors and 6 ray tracing cores. The 960A uses Maxwell on a 28 nm process with 1,870 million transistors and no ray tracing cores. The 660M supports DirectX 12 Ultimate (12_2), while the 960A supports DirectX 12 (11_0).

Q: How do the memory configurations compare?

A: The 660M uses system-shared memory with bandwidth that is system dependent. The 960A has 2 GB of GDDR5 on a 128-bit bus with 80.19 GB/s of bandwidth.

Q: Which GPU has higher power consumption?

A: The GTX 960A has a TDP of 75 W, while the Radeon 660M has a TDP of 40 W. The 660M is an integrated GPU, while the 960A is an MXM module.