NVIDIA GeForce GTX 950

The NVIDIA GeForce GTX 950 is a Maxwell 2.0 architecture part, built on the GM206 chip using a 28 nm process at TSMC. It carries 2,940 million transistors on a 228 mm² die, yielding a transistor density of 12.9 million per square millimeter. The card is positioned as an end-of-life product, released in August 2015, and sits in the 53rd percentile of all GPUs according to aggregate benchmark data. Its average benchmark score of 13,549 places it in a tightly contested performance tier, where it trades blows with several AMD and NVIDIA offerings.

Memory Subsystem

The GTX 950 is equipped with 2 GB of GDDR5 memory, operating across a 128-bit bus. The memory clock runs at 1653 MHz, which translates to 6.6 Gbps effective and yields a total bandwidth of 105.8 GB/s. This configuration is a notable constraint for modern workloads, as the 2 GB capacity and 128-bit interface are modest by contemporary standards. Pixel rate sits at 38.02 GPixel/s, while texture rate reaches 57.02 GTexel/s.

For high-resolution gaming, the data indicates this memory subsystem is a limiting factor. At 1080p, the 105.8 GB/s bandwidth can support a reasonable frame buffer, but the 2 GB VRAM will fill quickly with high-resolution textures and modern asset streaming. At 1440p or above, the combination of small capacity and narrow bus width will likely induce texture swapping and stutter in memory-intensive scenes. The benchmark results do not isolate memory-bound scenarios, but the raw figures suggest that the card is best suited for lower resolutions or reduced texture quality settings. The 128-bit bus width, in particular, means that the memory controller must work harder to sustain data throughput compared to wider interfaces, which is reflected in the modest bandwidth figure.

Ray Tracing and Feature Set

The GTX 950 does not include dedicated ray tracing cores or tensor cores, as these are absent from the FACT PACK specifications. It relies on 768 shading units, 48 texture mapping units, and 32 ROPs to handle rendering tasks. The architecture supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4, providing a baseline of modern API compatibility. However, without RT cores, any ray-traced workloads would fall back to compute shaders, which is not a practical path given the card’s 1.825 TFLOPS of FP32 performance.

The feature set is rooted in the Maxwell 2.0 generation, which introduced improved color compression and tessellation performance over its predecessor. The card supports PCIe 3.0 x16 for host connectivity and offers display outputs of 1x DVI, 1x HDMI 2.0, and 3x DisplayPort 1.2, enabling multi-monitor setups. The lack of tensor cores also means that any AI-accelerated features, such as deep learning super sampling, are unsupported. Benchmark scores for Geekbench Metal, OpenCL, and Vulkan are 7,172, 16,740, and 16,734, respectively, indicating that compute performance is consistent across different API environments, but none of these figures suggest hardware-accelerated ray tracing capability.

Benchmark Performance

The aggregate benchmark score for the GTX 950 is 13,549, which places it in a narrow band of performance where differences of a few percent separate it from its nearest rivals. Against the AMD Radeon RX 580, the GTX 950 scores 0.1% lower, a negligible margin that falls within run-to-run variance. Similarly, the NVIDIA GeForce GTX 570 shows a 0.1% lower delta, meaning the GTX 950 is effectively tied with it. The AMD Radeon Pro 455 trails by 1%, and the AMD Radeon Pro 555 trails by 1.1%, indicating the GTX 950 holds a slight but consistent edge over these mobile-class parts.

Breaking down the individual tests, the Geekbench OpenCL score of 16,740 is the highest among the three benchmarks, while the Vulkan score of 16,734 is nearly identical, differing by only 6 points. The Metal score of 7,172 is substantially lower, reflecting the platform-specific nature of that API rather than a general performance deficiency. The average of these three scores, 13,549, is heavily influenced by the two higher compute results, which suggests the card is more capable in general-purpose compute workloads than in graphics-specific APIs like Metal. The 53rd percentile ranking indicates it outperforms slightly more than half of all GPUs in the database, a middling position that aligns with its age and specifications.

Who Should Consider It

The GTX 950 is a card for gamers who prioritize low-resolution esports titles or older AAA games at 1080p with reduced settings. The 2 GB VRAM and 105.8 GB/s bandwidth are adequate for games that do not demand high-resolution texture packs, and the 1.825 TFLOPS FP32 performance can handle light shading workloads. Benchmark data shows the card is competitive with the Radeon RX 580, which is surprising given the latter’s newer architecture, but the GTX 950’s limitations in memory capacity become apparent in modern titles.

For users targeting 1080p with medium to high settings in games released around the card’s launch era, the GTX 950 provides a playable experience. At higher resolutions, such as 1440p, the memory subsystem will become a bottleneck, and the card is not recommended for such scenarios. The 90 W TDP and 250 W suggested PSU make it a low-power option for compact builds, and the dual-slot cooler with a 1x 6-pin connector is straightforward to install. However, the end-of-life production status means it is not a forward-looking purchase, and users should consider it only for legacy systems or as a temporary stopgap.

How It Compares

AMD Radeon RX 580: The GTX 950 trails the RX 580 by 0.1% in average benchmark score, making the two effectively equivalent in aggregate performance. This is a surprising result, as the RX 580 is a later-generation card, but the data shows no meaningful performance gap between them. In practice, the RX 580 likely has a larger memory allocation, but that spec is not cited here, so the comparison rests solely on the near-identical scores.

NVIDIA GeForce GTX 570: The GTX 950 scores 0.1% lower than the GTX 570, again a negligible difference. The GTX 570 is from an older generation, yet it holds its own against the Maxwell part. This suggests that architectural efficiency gains in the GTX 950 are offset by other factors, though the numbers do not reveal the cause. For practical purposes, these two cards are interchangeable in performance.

AMD Radeon Pro 455: The GTX 950 leads the Pro 455 by 1%, a modest but consistent margin. The Pro 455 is a mobile workstation GPU, so the GTX 950’s advantage in compute workloads aligns with its desktop-class power envelope. The delta is small enough that real-world differences would only appear in sustained workloads.

AMD Radeon Pro 555: The GTX 950 outperforms the Pro 555 by 1.1%, the largest margin among its nearest rivals. This indicates a slightly more decisive edge over this mobile part, though still within a tight performance band. The GTX 950’s higher thermal and power headroom likely contributes to this advantage, but the benchmark data alone supports the 1.1% figure.