NVIDIA GeForce GTX 870M vs NVIDIA GeForce GTX 960M Comparison

The NVIDIA GeForce GTX 870M and NVIDIA GeForce GTX 960M represent two distinct approaches to mobile graphics within the same family lineage, separated by a year of architectural evolution. The 870M, built on the older Kepler architecture, is a larger, more power-hungry chip that relies on raw compute width, while the 960M, using the Maxwell architecture, is a smaller, more efficient part that achieves its performance through higher clocks and architectural efficiency. The benchmark data shows a narrow overall victory for the 870M, but the story is more complex than a simple aggregate score, as the two GPUs are tuned for different workloads and software environments.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA GeForce GTX 870M has a higher average benchmark score of 9959, compared to the GTX 960M's 9645, a difference of roughly 3.3% based on the aggregate data.

Q: In the specific OpenCL benchmark, which GPU wins and by how much?

A: The GTX 870M wins the Geekbench OpenCL test with a score of 12630, beating the GTX 960M's 11045 by a margin of 14.4%.

Q: What is the difference in their transistor counts and die sizes?

A: The GTX 870M has a substantially larger chip with 3,540 million transistors on a 294 mm² die, while the GTX 960M has 1,870 million transistors on a much smaller 148 mm² die, both fabricated on the same 28 nm TSMC process.

Q: How do their memory configurations differ?

A: The GTX 870M comes with 3 GB of GDDR5 memory on a 192-bit bus, yielding a bandwidth of 120.0 GB/s, whereas the GTX 960M offers 4 GB of GDDR5 on a narrower 128-bit bus, resulting in a lower bandwidth of 80.19 GB/s.

Q: What are the performance percentile ranks for each GPU?

A: The GTX 870M sits at the 48th percentile among all GPUs, while the GTX 960M is just below it at the 46th percentile, indicating they are closely matched in the overall performance distribution.

Q: Which GPU supports the Vulkan API version 1.4?

A: The GTX 960M supports Vulkan 1.4, while the GTX 870M supports an older version, Vulkan 1.2.175, though both offer DirectX 12 (11_0) and OpenGL 4.6.

Architecture Differences

The two GPUs are built on entirely different architectures, which explains most of their performance characteristics. The GTX 870M uses the GK104 chip based on the Kepler architecture, which was designed for high raw throughput. This is reflected in its massive configuration of 1344 shading units, 112 texture mapping units (TMUs), and 24 raster operation units (ROPs). In contrast, the GTX 960M uses the GM107 chip from the Maxwell architecture, which is optimized for efficiency and higher clock speeds per core, but features a leaner setup of only 640 shading units, 40 TMUs, and 16 ROPs.

The process node is the same for both — a 28 nm process from TSMC — but the physical chips are vastly different in scale. The 870M's die size is 294 mm², nearly double the 148 mm² of the 960M, and its transistor count of 3,540 million dwarfs the 960M's 1,870 million. Interestingly, the transistor density is similar, with the 870M at 12.0M / mm² and the 960M at 12.6M / mm², showing that the scale difference is purely about chip size rather than packing efficiency.

Clock speeds tell a different story. The 870M runs at a base clock of 941 MHz with a boost of 967 MHz, while the 960M is clocked significantly higher at 1097 MHz base and 1176 MHz boost. This clock advantage helps the 960M close the gap in compute tasks, but the 870M's sheer number of cores gives it a massive throughput advantage in fill-rate and texture-bound workloads. The memory subsystems also differ: the 870M uses a 192-bit bus, while the 960M is on a 128-bit bus, directly impacting their respective bandwidth figures.

Head-to-Head Benchmarks

The only direct head-to-head benchmark available is Geekbench OpenCL, and it clearly favors the GTX 870M. The 870M scores 12630, while the 960M scores 11045, giving the 870M a 14.4% lead. This is a substantial margin, and it highlights the 870M's advantage in compute-heavy workloads that can leverage its higher number of shading units and texture units. The OpenCL test is often sensitive to raw computational throughput, which is where the 870M's 2.599 TFLOPS of FP32 performance versus the 960M's 1.505 TFLOPS comes into play.

However, the 960M has a benchmark that the 870M cannot match: a Geekbench Vulkan score of 8245. The 870M has no Vulkan benchmark result in the data, which is a notable gap given that the 960M supports the newer Vulkan 1.4 API. This suggests that in modern Vulkan-based titles, the 960M may offer a more future-proof experience, even if its raw compute numbers are lower. The 870M's OpenCL score is its strongest showing, but the lack of a Vulkan result for the 870M means the two cannot be directly compared across all APIs.

The aggregate average benchmark scores reflect a closer contest: the 870M's 9959 average is only 3.3% higher than the 960M's 9645. This narrower gap, compared to the 14.4% OpenCL delta, indicates that the 960M is more competitive in other, unlisted workloads, likely due to its higher clock speeds and architectural efficiency. The data implies that the 870M is the better choice for pure compute throughput, but the 960M is not far behind on average.

Specification Differences

The specification sheets for these two GPUs show clear divergences. The 870M has a base clock of 941 MHz and a boost clock of 967 MHz, while the 960M runs at 1097 MHz base and 1176 MHz boost, making the 960M roughly 16% faster in raw clock speed. Memory size favors the 960M at 4 GB versus the 870M's 3 GB, but the 870M counters with a wider 192-bit bus and higher bandwidth of 120.0 GB/s, compared to the 960M's 128-bit bus and 80.19 GB/s bandwidth.

Shader and texture resources are heavily skewed toward the 870M: 1344 shading units versus 640, 112 TMUs versus 40, and 24 ROPs versus 16. This leads to significant differences in pixel rate (27.08 GPixel/s for the 870M versus 18.82 GPixel/s for the 960M) and texture rate (108.3 GTexel/s versus 47.04 GTexel/s). The FP32 compute is also lopsided, with the 870M delivering 2.599 TFLOPS versus the 960M's 1.505 TFLOPS.

Power consumption is a major point of differentiation. The 870M has a 100 W TDP, while the 960M is rated at 75 W, making the 960M the more thermally efficient option. Both use MXM-B (3.0) interfaces and have no dedicated power connectors, but the 870M's higher TDP suggests it will generate more heat in a laptop chassis. The 960M also supports Vulkan 1.4, whereas the 870M is limited to Vulkan 1.2.175, a notable software feature divergence.

Where Each One Wins

The GTX 870M is the clear winner in compute-heavy scenarios. Its 14.4% OpenCL lead, combined with 72.8% higher FP32 throughput (2.599 TFLOPS versus 1.505 TFLOPS) and a 130.2% higher texture rate (108.3 GTexel/s versus 47.04 GTexel/s), makes it superior for GPU-accelerated tasks like rendering, video encoding, or any workload that can saturate its many cores. The wider 192-bit memory bus and higher 120.0 GB/s bandwidth also give it an advantage in memory-intensive applications where data throughput is the bottleneck.

The GTX 960M wins in efficiency and modern API support. Its 75 W TDP is 25 W lower than the 870M's 100 W, which is critical for thin-and-light laptops where cooling and battery life are paramount. The higher clock speeds (1097 MHz base versus 941 MHz) mean that for lightly-threaded or latency-sensitive tasks, the 960M can feel more responsive. The Vulkan 1.4 support is a forward-looking advantage, as it allows the 960M to run newer games and applications that may rely on Vulkan features unavailable on the 870M's older 1.2.175 implementation.

In terms of memory capacity, the 960M's 4 GB frame buffer is better suited for modern games with high-resolution textures, even if the bandwidth is lower. The 870M's 3 GB may become a limitation in such scenarios, despite its faster bus. The 960M's 4 GB versus the 870M's 3 GB is a practical win for gaming, where VRAM capacity often matters more than theoretical bandwidth.

The Verdict

The data presents a nuanced picture. The GTX 870M is the stronger performer in raw compute benchmarks, as evidenced by its 14.4% lead in OpenCL and its higher average score of 9959 versus 9645. Users who need maximum GPU compute power for tasks like 3D rendering or scientific computing, and who are not concerned about power consumption, should opt for the 870M. Its 1344 shading units and 2.599 TFLOPS of FP32 performance are simply unmatched by the 960M.

However, the GTX 960M is the more balanced choice for laptop buyers who prioritize efficiency and modern software compatibility. Its 75 W TDP makes it easier to cool and more battery-friendly, and its Vulkan 1.4 support ensures better compatibility with future titles. The 4 GB VRAM is also a practical advantage for modern gaming. The 960M's average score of 9645, which is only 3.3% lower than the 870M, shows that it sacrifices little in overall performance while offering significant gains in efficiency.

For gamers, the 960M is likely the safer pick due to its higher clocks, lower power draw, and larger memory pool. For those who need a mobile workstation GPU for compute-heavy tasks, the 870M's superior throughput in OpenCL benchmarks makes it the logical choice, provided the system can handle its 100 W thermal envelope. The data does not support a universal winner; it supports a decision based on workload priorities.