NVIDIA GeForce GTX 960

The NVIDIA GeForce GTX 960, built on the Maxwell 2.0 architecture with the GM206 chip, occupies a specific historical position in the GeForce 900 generation. This analysis draws exclusively on the provided data to evaluate its power characteristics, memory subsystem, feature set, and competitive standing.

Power and Cooling

The GTX 960 carries a thermal design power of 120 W, which places it in a modest power envelope for a desktop graphics card of its era. This figure is a defining characteristic of the card, as it dictates the thermal and electrical requirements for system integration. The board is designed as a dual-slot solution, indicating that it will occupy two expansion slots in a chassis, a common configuration for cards with a 120 W TDP to accommodate a substantial heatsink and fan assembly.

For power delivery, the card requires a single 6-pin PCIe power connector. This is a minimal requirement, and the data indicates a suggested power supply rating of 300 W for the overall system. This combination of a 120 W TDP and a single 6-pin connector suggests that the GTX 960 was engineered to be a drop-in upgrade for a wide range of existing desktop systems, avoiding the need for high-capacity power supplies or multiple auxiliary power connections. The card interfaces with the motherboard via a PCIe 3.0 x16 bus interface, the standard for discrete graphics of its generation. Physically, the card has a length of 241 mm or 9.5 inches, a dimension that should be verified against chassis clearance before installation. The production status is listed as end-of-life, and the card was released on January 21, 2015, succeeding the GeForce 700 series and preceding the GeForce 10 series.

Memory Subsystem

The GTX 960 is equipped with 2 GB of GDDR5 memory, a capacity that was standard for its performance class at launch. The memory interface is a 128-bit bus, which is a narrower configuration compared to higher-tier cards of the same generation. This bus width, combined with a memory clock of 1753 MHz (7 Gbps effective), yields a total memory bandwidth of 112.2 GB/s. This bandwidth figure is a critical metric for understanding the card's performance envelope, particularly at higher resolutions and with texture-heavy workloads. The 2 GB capacity, in particular, becomes a limiting factor for modern games that require larger framebuffers; at high resolutions with maximum texture detail, the card may be forced to use slower system memory as a fallback, which can cause stuttering and reduced frame rates. The 112.2 GB/s bandwidth is sufficient for the card's target performance level at 1080p, but the combination of capacity and bandwidth defines its boundary conditions for more demanding scenarios. The pixel rate is 37.70 GPixel/s, and the texture rate is 75.39 GTexel/s, which are derived from the 32 ROPs and 64 TMUs working at the core clock speeds. These rates provide a baseline for the card's fill-rate capabilities.

Ray Tracing and Feature Set

The GTX 960 does not have dedicated ray tracing cores or tensor cores. Its feature set is defined by the Maxwell 2.0 architecture, which provides support for DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The DirectX 12 (12_1) feature level indicates support for certain advanced rendering features but predates the hardware-accelerated ray tracing introduced in later architectures. Consequently, any ray tracing workloads would be handled through software or compute shaders, which is not a practical approach for real-time performance on this hardware. The API support, however, ensures broad compatibility with modern applications and games that use these low-level graphics APIs, allowing the card to function in contemporary software environments, though without the performance benefits of dedicated RT or tensor cores. The display outputs include 1x DVI, 1x HDMI 2.0, and 3x DisplayPort 1.2, providing flexible connectivity options for multiple monitors. The card has 1024 shading units, 64 texture mapping units, and 32 raster operation pipelines, which are the core computational resources for traditional rasterization workloads. Its FP32 performance is rated at 2.413 TFLOPS.

How It Compares

The GTX 960's position in the market can be understood by examining its average benchmark score relative to its nearest rivals. The data shows a tight cluster of performance around the 12,000-point mark, indicating that these cards are direct competitors.

NVIDIA GeForce GTX 670: The GTX 960 has an average score of 12045, which is 0.3% higher than the GTX 670's 12014. This is a statistically insignificant difference, effectively placing the two cards on par in overall performance. The GTX 670 is a card from a previous generation, so this result shows the GTX 960 delivering comparable performance to a higher-tier card from the GeForce 700 era.

NVIDIA GeForce GTX 960A: The GTX 960 leads the GTX 960A by 0.4% (12045 vs 12002). The GTX 960A appears to be a mobile or specialized variant of the same chip, and the slight performance advantage of the desktop GTX 960 is consistent with expectations for a full-power desktop implementation.

AMD Radeon Pro 5500M: The GTX 960 holds a 0.5% performance advantage over the Radeon Pro 5500M (12045 vs 11986). This is a mobile workstation GPU, and the margin is minimal. The comparison highlights that the GTX 960, despite being an older desktop part, can compete with a more modern mobile solution in aggregate benchmark scores.

AMD Radeon RX 6500 XT: The GTX 960 outperforms the Radeon RX 6500 XT by 1.2% (12045 vs 11897). This is the largest lead in the rival group, but it is still a narrow margin. The RX 6500 XT is a much newer card, yet the data shows the GTX 960 providing comparable, and slightly better, average performance in this benchmark suite.

Benchmark Performance

The GTX 960 achieves an average benchmark score of 12045 across all tests, placing it at the 50th percentile of all GPUs in the database. This median standing confirms its historical position as a mainstream performer. The individual benchmark results show variance across different testing methodologies. In the 3dmark_3dmark_steel_nomad_dx12 test, the card scores 162, which is a demanding, modern DirectX 12 workload. This low score reflects the card's age and its limitations with contemporary rendering techniques. Conversely, in compute-oriented tests, the card performs more strongly. It scores 18751 in Geekbench OpenCL and 20495 in Geekbench Vulkan. The highest score is in the Vulkan test, suggesting that the card's compute capabilities are relatively robust for its era. The Geekbench Metal score is 8773.

When interpreting these scores against its rivals, the data indicates a remarkably tight competitive field. The GTX 960's 0.3% lead over the GTX 670 is the closest margin, effectively a tie. The deltaPct values against all rivals are within 1.2%, which is within the typical run-to-run variance of benchmark testing. This suggests that in aggregate performance, these four cards are functionally interchangeable. However, the individual test scores reveal a more nuanced picture. For instance, while the GTX 960 trails in the specific 3dmark Steel Nomad test (scoring 162), its compute scores in Geekbench are substantial. This pattern indicates that the card maintains strong raw compute throughput, as evidenced by its 2.413 TFLOPS FP32 rate, but may be limited by its memory subsystem or driver optimizations in specific gaming workloads. The 50th percentile ranking is a direct consequence of this balanced but not exceptional performance profile. The data shows that the GTX 960 is a card that, despite its age and end-of-life status, still holds a competitive, if marginal, edge over its identified nearest rivals in overall average benchmark performance.