NVIDIA Quadro P620 Mobile

The NVIDIA Quadro P620 Mobile is a Pascal-generation mobile workstation GPU built for professional laptops, pairing a compact 14 nm GP107 chip with 4 GB of GDDR5 memory. It occupies a specific niche: a low-power, end-of-life mobile part with no direct rivals listed in the current data, yet its specifications and architectural profile make it a clean baseline for evaluating professional mobile graphics of its era.

Power and Cooling

The Quadro P620 Mobile carries a 40 W TDP, a figure that places it firmly in the low-power segment of mobile workstation GPUs. This modest thermal envelope means the card can be integrated into thin-and-light professional laptops without the heavy cooling solutions required by higher-tier parts. The data shows no suggested PSU rating and no power connectors, the card draws entirely from the PCIe 3.0 x16 slot, which is typical for IGP (integrated graphics processor) mobile designs. The slot width is listed as "IGP," confirming that this is not a discrete add-in card but a soldered or embedded solution for portable devices.

For system integrators, the absence of supplementary power connectors simplifies laptop motherboard design, as no external 6-pin or 8-pin cables are needed. The 40 W TDP also implies that a capable air cooler, such as a thin heat pipe assembly with a single fan, is sufficient for sustained operation. The production status is end-of-life, so new laptop designs should account for the lack of ongoing supply, but the low power draw remains a relevant data point for existing deployments. Benchmark results indicate that the power efficiency of the Pascal architecture at 14 nm (with Samsung as the foundry) allows this GPU to deliver usable professional performance without taxing battery life or chassis thermals.

Memory Subsystem

Memory capacity is 4 GB of GDDR5, paired with a 128-bit bus width. The memory clock is 1502 MHz, which translates to 6 Gbps effective data rate. Bandwidth is 96.13 GB/s. This combination is modest by contemporary standards, but for the target workload, mobile professional CAD, 3D modeling, and light GPU-accelerated tasks, the capacity and bandwidth are balanced. The 128-bit bus is a direct consequence of the GP107 chip's small die size (132 mm²), which houses 3,300 million transistors at a density of 25.0M per mm².

At high resolutions, the 96.13 GB/s bandwidth becomes a limiting factor. For 4K displays or multi-monitor setups, texture-heavy scenes and large frame buffers will saturate the memory bus. The 4 GB capacity is sufficient for moderate workloads, but data shows that scenes exceeding this limit will force texture swapping, which degrades performance. The GDDR5 type is older than GDDR6, but the effective 6 Gbps speed is typical for a 2018-era product. The pixel rate of 23.09 GPixel/s and texture rate of 46.18 GTexel/s are closely tied to the memory bandwidth; when the bus is saturated, both rates drop proportionally. For professional applications that rely on large datasets, such as point clouds or high-resolution textures, the memory subsystem will be the bottleneck before the compute units are fully utilized.

Benchmark Performance

The benchmark data for this GPU is sparse: the average benchmark score is 0, and the percentile versus all GPUs is 50. The 50th percentile indicates that this card sits exactly at the median of all GPUs in the database, meaning half of all GPUs perform better and half perform worse. This is a neutral position, not a performance leader, but not a bottom-tier part either. The FP32 compute is 1,477.6 GFLOPS, which is a straightforward measure of single-precision throughput. The FP16 rate is 23.09 GFLOPS, with a 1:64 ratio to FP32, indicating that half-precision compute is heavily de-emphasized, a common trait for Pascal architecture, which did not prioritize FP16 for gaming or professional workloads.

The shading units number 512, with 32 texture mapping units (TMUs) and 16 render output units (ROPs). These ratios are consistent with a small, efficiency-focused chip. The texture rate of 46.18 GTexel/s and pixel rate of 23.09 GPixel/s are derived from the core clocks (1177 MHz base, 1443 MHz boost). In practical terms, this GPU can handle entry-level professional tasks, but it will struggle with high-polygon scenes or complex shaders. The clock speeds are relatively high for a 40 W part, meaning the chip is designed to run at near-maximum frequency under load, but the limited core count (512 shading units) caps absolute throughput. No nearest rivals are provided in the data, so direct percentage comparisons are not possible. However, the 50th percentile rank suggests that in a mixed workload of gaming and professional GPUs, this card performs on par with the median.

How It Compares

Without nearestRivals data, direct head-to-head comparisons are unavailable from the fact pack. The predecessor is Quadro Maxwell-M, and the successor is Quadro Turing-M. Relative to the predecessor, the Pascal architecture in the P620 Mobile offers a newer process node (14 nm vs. the older Maxwell process) and likely higher efficiency per watt, though no specific benchmark deltas are listed. The data shows that the GP107 chip is a small, low-transistor-count design (3,300 million) compared to higher-end Pascal parts, which means the P620 Mobile is positioned as an entry-level mobile workstation GPU. The successor, Quadro Turing-M, introduces Turing architecture with dedicated RT and tensor cores, features that the Pascal-based P620 Mobile lacks entirely (rtCores and tensorCores are null in the data). This generational gap is significant: Turing adds hardware ray tracing and AI acceleration, which the P620 Mobile cannot match. However, for legacy software that relies on OpenGL 4.6 or DirectX 12 (12_1), the Pascal part remains serviceable.

The bus interface is PCIe 3.0 x16, which is fully compatible with modern laptops, though the bandwidth of the interface is unlikely to be a bottleneck given the GPU's modest compute and memory bandwidth. The display outputs are listed as "Portable Device Dependent," meaning the actual connectors vary by laptop model, no fixed DP or HDMI ports are specified. This is typical for mobile IGP designs where the laptop vendor controls the output configuration.

Ray Tracing and Feature Set

The Quadro P620 Mobile has no dedicated ray tracing cores and no tensor cores, both fields are null in the data. This is a fundamental architectural limitation, as the Pascal generation predates the RTX family's hardware acceleration for ray tracing. The FP16 performance is 23.09 GFLOPS, a 1:64 ratio to FP32, which means there is no meaningful half-precision acceleration for AI or machine learning tasks. The feature set relies entirely on the base Pascal capabilities: 512 shading units for conventional rasterization, 32 TMUs for texture filtering, and 16 ROPs for pixel output.

API support is robust for its era: DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4 are all listed. This ensures compatibility with modern professional applications that use these APIs. The DirectX 12_1 feature level includes support for conservative rasterization and rasterizer-ordered views, which are relevant for certain CAD and visualization workloads. Vulkan 1.4 support is notably recent, suggesting driver updates have kept the card compatible with newer low-level APIs. However, the absence of ray tracing and tensor cores means that any workload requiring hardware-accelerated RT or AI denoising will fall back to compute shaders on the 512 shading units, which is far slower. The pixel rate of 23.09 GPixel/s and texture rate of 46.18 GTexel/s are the practical ceilings for rasterization, these are modest figures that will limit fill-rate-heavy effects like high-resolution shadows or post-processing filters. The card is end-of-life, so no further feature additions are expected, but the existing API support covers the majority of legacy professional software.