AMD Radeon 660M vs NVIDIA GeForce GTX 965M Comparison

The NVIDIA GeForce GTX 965M and AMD Radeon 660M represent two very different approaches to mobile graphics, separated by seven years of silicon evolution. The GTX 965M is a 2015 discrete MXM module built on Maxwell 2.0, while the Radeon 660M is a 2022 integrated GPU on RDNA 2.0. Their average benchmark scores land within 4.3% of each other — 14404 for the NVIDIA part versus 13812 for the AMD part — making this a surprisingly close contest between a dedicated chip and an IGP. The data shows a split decision: the GTX 965M wins OpenCL by 12.7%, while the Radeon 660M counters with a 3% lead in Vulkan. The rest is a study in architectural trade-offs, power envelopes, and feature sets.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA GeForce GTX 965M leads with an average benchmark score of 14404, compared to 13812 for the AMD Radeon 660M. That is a 4.3% gap in favor of the NVIDIA part.

Q: How do they compare in the individual head-to-head tests?

A: The GTX 965M wins Geekbench OpenCL with 14509 against 12876, a 12.7% advantage. The Radeon 660M wins Geekbench Vulkan with 14748 against 14299, a 3% margin.

Q: What are the architectural differences between the two?

A: The GTX 965M uses the GM204 chip on Maxwell 2.0 architecture at 28 nm, with 5,200 million transistors on a 398 mm² die. The Radeon 660M uses the Rembrandt chip on RDNA 2.0 at 6 nm, with 13,100 million transistors on a 208 mm² die.

Q: Which GPU has more shading units and texture units?

A: The GTX 965M has 1024 shading units, 64 TMUs, and 32 ROPs. The Radeon 660M has 384 shading units, 24 TMUs, and 16 ROPs. The NVIDIA part has roughly 2.7x the shading units and TMUs.

Q: Does either GPU support hardware ray tracing?

A: Yes, the AMD Radeon 660M includes 6 ray tracing cores. The NVIDIA GeForce GTX 965M has no ray tracing cores listed.

Q: What is the DirectX support difference?

A: The Radeon 660M supports DirectX 12 Ultimate (12_2), while the GTX 965M supports DirectX 12 (12_1). Both support OpenGL 4.6 and Vulkan 1.4.

Architecture Differences

The GTX 965M is built on NVIDIA's Maxwell 2.0 architecture, using the GM204 chip fabricated on a 28 nm process at TSMC. This discrete GPU packs 5,200 million transistors into a 398 mm² die, yielding a transistor density of 13.1 million per square millimeter. The Radeon 660M, by contrast, uses AMD's RDNA 2.0 architecture with the Rembrandt chip, manufactured on a 6 nm process also at TSMC. It crams 13,100 million transistors into a 208 mm² die, achieving a density of 63.0 million per square millimeter — nearly five times denser than the older NVIDIA part.

The shader configuration differs dramatically. The GTX 965M fields 1024 shading units, 64 texture mapping units, and 32 render output units. The Radeon 660M is far leaner on paper with only 384 shading units, 24 TMUs, and 16 ROPs. Despite this 2.7x disparity in raw shader count, the Radeon 660M compensates with significantly higher clocks: its base clock is 1500 MHz and boost reaches 1900 MHz, versus 924 MHz base and 950 MHz boost for the GTX 965M.

Memory architecture is a fundamental split. The GTX 965M uses 2 GB of dedicated GDDR5 memory on a 128-bit bus, delivering 80.19 GB/s of bandwidth. The Radeon 660M uses system shared memory, with its bus width, size, and bandwidth all listed as system dependent. This means the AMD part's memory performance is entirely contingent on the host laptop's RAM configuration, whereas the NVIDIA part has fixed, dedicated memory resources.

The feature set also diverges. The Radeon 660M includes 6 ray tracing cores and supports DirectX 12 Ultimate (12_2), along with FP16 compute rated at 2.918 TFLOPS (2:1). The GTX 965M has no ray tracing cores, caps at DirectX 12 (12_1), and has no FP16 rating listed. Both GPUs share OpenGL 4.6 and Vulkan 1.4 support. The Radeon 660M is an IGP with a 40 W TDP, while the GTX 965M is an MXM module with no listed TDP. The bus interfaces also differ: PCIe 4.0 x8 for the AMD part versus MXM-B (3.0) for the NVIDIA part.

Head-to-Head Benchmarks

The two GPUs split their two head-to-head benchmark tests, each taking one victory. In Geekbench OpenCL, the GTX 965M scores 14509 against the Radeon 660M's 12876, a 12.7% win for the NVIDIA part. This is the larger margin of the two tests and suggests the GTX 965M's raw shader throughput and dedicated memory give it a clear edge in compute workloads that favor OpenCL.

In Geekbench Vulkan, the tables turn. The Radeon 660M scores 14748, edging out the GTX 965M's 14299 by 3%. This is a notable result because the AMD part achieves this with only 384 shading units versus 1024, relying instead on its higher clock speeds and newer architecture. The 6 nm process and RDNA 2.0 design appear to extract more efficiency per shader in Vulkan workloads.

Looking at the broader context, the GTX 965M's nearest rivals include the AMD Radeon RX Vega 11 at 14385 (0.1% delta), the NVIDIA GeForce GTX TITAN at 14373 (0.2% delta), and the AMD Radeon Vega 11 at 14352 (0.4% delta). The Radeon 660M's nearest rivals include the NVIDIA RTX A2000 Mobile at 13821 (-0.1% delta), the AMD Radeon RX 570X at 13871 (-0.4% delta), and the AMD Radeon RX 7900 XT at 13745 (0.5% delta). This places the GTX 965M in a cluster of older high-end parts, while the Radeon 660M sits near modern mobile and desktop options.

The percentile rankings are nearly identical: the GTX 965M sits at the 56th percentile of all GPUs, and the Radeon 660M at the 55th. This confirms that despite their architectural differences, both land in the same performance tier. The wins are evenly split at 1-1, making the head-to-head verdict a matter of workload preference rather than overall superiority.

Specification Differences

The two GPUs differ on nearly every specification field that matters. The process node is a 28 nm for the GTX 965M versus 6 nm for the Radeon 660M. Transistor count is 5,200 million versus 13,100 million, and die size is 398 mm² versus 208 mm². The transistor density is 13.1M per mm² for NVIDIA and 63.0M per mm² for AMD.

Clock speeds favor the AMD part substantially: the GTX 965M runs at 924 MHz base and 950 MHz boost, while the Radeon 660M runs at 1500 MHz base and 1900 MHz boost. The memory situation is inverted — the GTX 965M has 2 GB of GDDR5 on a 128-bit bus with 80.19 GB/s bandwidth, while the Radeon 660M uses system shared memory with system dependent bandwidth.

Compute resources differ sharply: 1024 shading units, 64 TMUs, and 32 ROPs for NVIDIA versus 384 shading units, 24 TMUs, and 16 ROPs for AMD. The pixel rates are identical at 30.40 GPixel/s, but the texture rates differ — 60.80 GTexel/s for the GTX 965M versus 45.60 GTexel/s for the Radeon 660M. FP32 performance is 1.946 TFLOPS for NVIDIA versus 1,459.2 GFLOPS for AMD.

The Radeon 660M has 6 ray tracing cores and FP16 capability at 2.918 TFLOPS (2:1), neither of which the GTX 965M offers. TDP is 40 W for the AMD part, with no TDP listed for NVIDIA. The GTX 965M is an MXM module with MXM-B (3.0) interface, while the Radeon 660M is an IGP with PCIe 4.0 x8. Both share the same display outputs (portable device dependent), power connectors (none), and production status (end-of-life). Release dates are 2015-01-08 for NVIDIA and 2022-01-03 for AMD.

The Verdict

The benchmark data presents a nuanced picture. The GTX 965M holds a 4.3% average score advantage (14404 versus 13812) and a decisive 12.7% win in OpenCL, but the Radeon 660M counters with a 3% win in Vulkan and brings modern features to the table. The GTX 965M's dedicated 2 GB GDDR5 memory and 80.19 GB/s bandwidth give it a fixed performance baseline, while the Radeon 660M's system shared memory makes its performance variable depending on the host system.

For buyers who prioritize raw compute throughput in OpenCL-heavy workloads, the GTX 965M is the data-backed choice. Its 1024 shading units and 64 TMUs provide a substantial resource advantage that shows up in the 12.7% OpenCL lead. The identical pixel rates (30.40 GPixel/s) suggest that rasterization-bound scenarios may be closer than the spec sheets imply.

For those who favor Vulkan workloads or want modern feature support, the Radeon 660M is the pick. Its 3% Vulkan win demonstrates that the newer RDNA 2.0 architecture and 1900 MHz boost clock can overcome the shader count deficit. The 6 ray tracing cores and DirectX 12 Ultimate support add future-proofing that the GTX 965M simply cannot match, despite the NVIDIA part's higher average score.

The production status of both parts is end-of-life, so neither is a forward-looking purchase. But within their respective eras, the data shows the GTX 965M as the higher-performing GPU overall, with the Radeon 660M as the more feature-complete and efficient option. The choice hinges on whether you need the NVIDIA part's compute muscle or the AMD part's modern API and ray tracing capabilities.

Where Each One Wins

NVIDIA GeForce GTX 965M wins in OpenCL compute performance, taking a 12.7% lead over the Radeon 660M in that specific benchmark. Its higher average benchmark score (14404 versus 13812) and larger shading unit count (1024 versus 384) make it the better choice for applications that leverage raw shader throughput. The dedicated 2 GB GDDR5 memory with 80.19 GB/s bandwidth ensures consistent memory performance without relying on system RAM. The higher texture rate (60.80 GTexel/s versus 45.60 GTexel/s) also favors this part in texture-heavy workloads.

AMD Radeon 660M wins in Vulkan performance, posting a 3% higher score (14748 versus 14299) in that test. Its 1900 MHz boost clock and modern RDNA 2.0 architecture deliver better per-shader efficiency, compensating for the lower shader count. The 6 ray tracing cores provide hardware acceleration that the GTX 965M lacks entirely, and DirectX 12 Ultimate support enables the latest graphics features. The 40 W TDP makes it a far more power-efficient option, suitable for thinner laptops without discrete graphics. The 6 nm process and 63.0M per mm² transistor density indicate a much more modern and dense design, which translates to better performance per watt in supported workloads.