NVIDIA GeForce GTX 980

The NVIDIA GeForce GTX 980, built on the GM204 chip with Maxwell 2.0 architecture, was released on September 18, 2014. Fabricated on TSMC's 28 nm process, it packs 5,200 million transistors into a 398 mm² die, achieving a transistor density of 13.1M per mm². The card operates at a base clock of 1127 MHz and a boost clock of 1216 MHz, with 2048 shading units, 128 texture mapping units, and 64 ROPs. It carries 4 GB of GDDR5 memory on a 256-bit bus, yielding 224.4 GB/s of bandwidth, and is rated at a TDP of 165 W. With a launch MSRP of 549 USD, the GTX 980 has since reached end-of-life status, succeeded by the GeForce 10 series.

Benchmark Performance

The GTX 980's average benchmark score stands at 9477, placing it in the 45th percentile of all GPUs in the database. This position indicates that the card sits slightly below the median, meaning the majority of modern GPUs outperform it. However, its nearest rivals reveal a tight performance cluster. Against the NVIDIA GeForce GTX 850M, which scores 9490, the GTX 980 is 0.1% slower—a negligible difference that puts the two within statistical noise. The GTX TITAN BLACK, with an average score of 9385, is 1% slower than the 980, making the 980 the faster card in that pairing. Conversely, the GTX 465 scores 9600, giving it a 1.3% advantage over the 980, and the Tesla M10 at 9634 is 1.6% faster. These deltas, all within 1.6%, demonstrate that the GTX 980 occupies a performance tier shared by a diverse set of cards from different generations and market segments, from a mobile GPU (850M) to a compute-oriented Tesla product.

Delving into individual benchmark results, the PassMark G3D score of 11095 provides a broad measure of 3D rendering capability, while the G2D score of 792 covers 2D operations. The GPU compute score of 4753 reflects general-purpose compute throughput. In Geekbench, the card achieves 15163 in Metal, 33457 in OpenCL, and 38172 in Vulkan—the Vulkan result being notably higher than the OpenCL and Metal figures, suggesting a strong affinity for that API. The 3DMark Steel Nomad DX12 score of 474 is comparatively low, aligning with the PassMark DX12 result of 46, which is the weakest of the four DirectX versions tested. PassMark scores for DX9, DX11, and DX10 are 164, 83, and 53 respectively, showing a clear regression as the API advances. This pattern indicates that the Maxwell 2.0 architecture delivers progressively lower relative performance under newer API workloads, with DX12 being the most challenging. The GTX 980's 4.981 TFLOPS of FP32 compute and 155.6 GTexel/s texture rate provide the raw throughput, but the API-level inefficiencies become apparent in modern titles that leverage DX12's low-level features.

Memory Subsystem

The GTX 980 is equipped with 4 GB of GDDR5 memory operating at 1753 MHz (7 Gbps effective) across a 256-bit bus, resulting in a memory bandwidth of 224.4 GB/s. This configuration was standard for high-end cards of its era, but the 4 GB capacity is a notable limitation for high-resolution rendering. The card's pixel rate of 77.82 GPixel/s and texture rate of 155.6 GTexel/s are directly tied to the available bandwidth; the 224.4 GB/s figure is sufficient to sustain these rates under typical workloads, but scenarios that exceed 4 GB of texture and geometry data will force memory swapping or compression, degrading performance. The 64 ROPs handle final pixel output, and their efficiency is contingent on memory bandwidth. For users targeting high resolutions with demanding texture packs, the 4 GB ceiling may become the primary bottleneck, though the benchmark data does not isolate memory-limited cases. The 256-bit bus width provides a balanced channel count for the GDDR5 modules, and the effective 7 Gbps speed is moderate by today's standards, but it was competitive at launch.

Power and Cooling

The GTX 980 has a TDP of 165 W, a figure that dictates its power and cooling requirements. The card is dual-slot and requires two 6-pin power connectors, with a suggested power supply rating of 450 W. This PSU recommendation provides a comfortable margin for the card's peak draw, and the dual-slot cooler design—measuring 267 mm in length, 111 mm in height, and 40 mm in width—is typical for a high-performance part of its generation. The 2x6-pin configuration is straightforward and compatible with most power supplies, avoiding the need for high-wattage 8-pin or 12-pin connectors. The 165 W TDP is moderate; it allows for a single-fan or dual-fan cooler, and the card's dimensions fit into most mid-tower cases, though the 267 mm length may require careful case selection. The PCIe 3.0 x16 interface provides adequate bandwidth for the card's data transfer needs, and the display outputs include 1x DVI, 1x HDMI 2.0, and 3x DisplayPort 1.2, covering a range of monitor configurations.

Who Should Consider It

Given its 45th percentile placement and average benchmark score of 9477, the GTX 980 is a mid-range performer. It is well-suited for users who primarily play games built around DirectX 9, 10, or 11, as evidenced by the PassMark scores of 164, 53, and 83 respectively. The card's DX11 score of 83 is its strongest among the modern APIs, suggesting that it can handle DX11 titles at high settings in 1080p or 1440p resolutions, though no specific resolution data is provided. However, its DX12 performance (PassMark 46, 3DMark Steel Nomad 474) is markedly weaker, indicating that users who play the latest DX12-only titles will see reduced frame rates relative to other cards in its score range. The GTX 980 is also a candidate for compute workloads that leverage OpenCL or Vulkan, given its strong Geekbench Vulkan score of 38172 and OpenCL score of 33457. But for those seeking to run modern, API-heavy games or applications that require more than 4 GB of video memory, the card's limitations become apparent. The 45th percentile ranking means that a majority of GPUs in the database outperform it, so it is best considered for legacy gaming or as a budget secondary card, not for cutting-edge performance.

Ray Tracing and Feature Set

The GTX 980 does not include dedicated ray tracing (RT) cores or tensor cores; both fields are null in the specifications. This means any ray tracing or AI-based features must be executed on the general-purpose shading units, which deliver 4.981 TFLOPS of FP32 compute. In practice, ray tracing workloads would run extremely slowly, and the card lacks the specialized hardware for accelerated ray traversal or denoising. On the API front, the GTX 980 supports DirectX 12 (feature level 12_1), OpenGL 4.6, and Vulkan 1.4. The DX12 support is limited to feature level 12_1, which omits some of the more advanced features found in DX12 Ultimate, such as mesh shaders and variable rate shading. OpenGL 4.6 and Vulkan 1.4 provide modern cross-platform capabilities, and the Vulkan score of 38172 suggests the card is reasonably proficient in that API. For users interested in ray tracing, the absence of RT cores is a decisive drawback; the card cannot accelerate these effects, and any implementation would rely on compute-based fallbacks that are impractical for real-time use. The feature set is otherwise solid for its time, with support for a wide range of display outputs and API standards, but the lack of dedicated AI and ray tracing hardware places it firmly in the pre-ray-tracing era of graphics cards.