NVIDIA Quadro M620 Mobile

The NVIDIA Quadro M620 Mobile is an end-of-life mobile workstation GPU built around the GM107 chip on TSMC's 28 nm Maxwell process. The fact pack records 1,870 million transistors on a 148 mm² die, 512 shading units, 32 TMUs, and 16 ROPs. It is packaged as an MXM Module with no power connectors, no listed suggested PSU, and a 30 W TDP. The benchmark array is empty, the average benchmark score is 0, and nearestRivals is empty, leaving the 50th percentile vs all GPUs as the only aggregate placement signal.

How It Compares

The nearestRivals array in the fact pack is empty, so there are no rival names, no rival scores, and no deltaPct values to interpret. That means the database currently supplies no direct comparative positioning against any specific GPU; any statement such as "ahead of X" or "behind Y" would be unsupported by the data. The only position-related signal is percentileVsAllGpus: 50, which places this part at the midpoint of the tracked all-GPU distribution. Because no deltaPct values are attached, the magnitude of separation from neighboring GPUs is unknown.

The product lineage fields are also part of the comparison context. The predecessor is listed as Quadro Kepler-M, and the successor is listed as Quadro Pascal-M. Those labels place the M620 between two generations in NVIDIA's mobile Quadro line, but the fact pack gives no scores for either predecessor or successor. Without measured values in those fields, no percentage gain or loss can be stated. The comparison story is therefore incomplete: the specifications can be read, but the measured rival deltas are absent.

Power and Cooling

The TDP is 30 W, a low-power figure for a GPU module. The power connectors field is None, and the suggested PSU field is null, so the fact pack contains no PSU wattage recommendation. The slot width is MXM Module, and the bus interface is MXM-A (3.0), indicating that power delivery and signaling are handled through the module connector rather than auxiliary power cables. Display outputs are marked Portable Device Dependent, meaning the host system determines the physical video output path.

Because the suggested PSU field is absent, any external power supply figure would have to come from outside the data and is not provided here. The 30 W TDP implies a modest thermal envelope, but the fact pack does not specify a cooler or chassis thermal solution. The lack of auxiliary power connectors means the host platform is expected to supply the module's power. The MXM form factor also means the cooling solution belongs to the host laptop, not to the add-in card itself. No connector requirements beyond the MXM-A interface are recorded.

Who Should Consider It

The fact pack contains no benchmark scores, so a conventional resolution-and-settings recommendation cannot be grounded in measured results. However, the memory and pipeline numbers define an envelope. The frame buffer is 2 GB GDDR5 on a 128-bit bus with 80.19 GB/s bandwidth. The memory clock is 1253 MHz, with 5 Gbps effective signaling. The pixel rate is 16.29 GPixel/s, and the texture rate is 32.58 GTexel/s.

For workloads that fit within 2 GB, these figures suggest a part aimed at portable workstations rather than high-resolution desktop-class rendering. The 512 shading units and 1,042.4 GFLOPS FP32 throughput provide compute headroom only when the memory system can feed it. At high resolutions, the 2 GB capacity and 80.19 GB/s bandwidth are the most obvious constraints. The 16 ROPs produce 16.29 GPixel/s, which is the upper bound on pixel output; the 32 TMUs produce 32.58 GTexel/s for texture-bound work.

Without scores, the data cannot assign a specific resolution and settings tier. The theoretical peak rates are specification-derived limits, not measured frame-rate results. A user whose workload is bounded by FP32 math would look at 1,042.4 GFLOPS; a user whose workload is bounded by textures would look at 32.58 GTexel/s; a user whose workload is bound by frame-buffer capacity would look at 2 GB and 80.19 GB/s. The fact pack does not say how those peaks translate into playable settings.

FAQ

Q: What TDP is listed for the Quadro M620 Mobile?

A: The TDP is 30 W.

Q: What is the memory configuration?

A: The GPU has 2 GB GDDR5 over a 128-bit bus, with 80.19 GB/s bandwidth. The memory clock is 1253 MHz, with 5 Gbps effective signaling.

Q: Which APIs are supported?

A: The fact pack lists DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.4.

Q: Does the fact pack show ray tracing or tensor cores?

A: The rtCores and tensorCores fields are both null, so no hardware ray tracing core counts or tensor core counts are recorded.

Q: What are the clock speeds?

A: The base clock is 756 MHz, and the boost clock is 1018 MHz.

Q: What is the production status?

A: The production status is End-of-life, with a release date of 2017-01-10.

Ray Tracing and Feature Set

The ray tracing and AI acceleration data is absent: rtCores is null, and tensorCores is null. Consequently, the fact pack cannot substantiate hardware ray tracing or tensor-core acceleration. The feature set instead rests on API compatibility: DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.4. The DirectX 12 entry carries the 11_0 feature level, which is the exact value recorded in the fact pack.

The underlying architecture is Maxwell, implemented with the GM107 chip at TSMC on a 28 nm process. The die contains 1,870 million transistors across 148 mm², giving a transistor density of 12.6M / mm². The shading resources are 512 shading units, 32 TMUs, and 16 ROPs. Display outputs are portable-device dependent. The process node and transistor budget are the physical basis of the feature set, but the fact pack lists no dedicated RT or tensor hardware. This is a feature set defined by API reach and Maxwell-generation compute resources rather than by dedicated ray tracing or AI blocks.

Memory Subsystem

The memory subsystem is 2 GB GDDR5 on a 128-bit bus, with 80.19 GB/s of bandwidth. The memory clock is 1253 MHz, translating to 5 Gbps effective. For a mobile MXM part with a 30 W TDP, the bandwidth is enough for modest frame buffers but not for high-resolution texture loads. The 128-bit bus is the interface width over which the GDDR5 data travels; the 80.19 GB/s figure is the resulting sustained transfer rate.

The 16 ROPs produce a pixel rate of 16.29 GPixel/s, which is the upper bound on pixel output per second. The 32 TMUs produce 32.58 GTexel/s, the corresponding texture processing limit. At 2 GB, the capacity ceiling will be reached before the compute hardware is exhausted in many high-resolution scenes. The 128-bit interface means memory throughput is tied to the relatively narrow bus width, and the 80.19 GB/s bandwidth is the ceiling for data movement between the frame buffer and the shading units. No rival memory bandwidth figures are present in the nearestRivals data, so no comparison of memory throughput against other GPUs is possible.

Benchmark Performance

The benchmark data is effectively absent. The benchmarks list is empty, avgBenchmarkScore is 0, and nearestRivals contains no entries. The only all-GPU signal is percentileVsAllGpus: 50, placing the M620 at the midpoint of the database's tracked GPU distribution. Because there are no deltaPct values, no exact percentage advantage or disadvantage relative to any rival can be stated. The 50th percentile is a relative rank, not a score; it does not show how far the M620 sits from neighboring GPUs.

The theoretical peak rates are the only performance numbers available. FP32 compute is 1,042.4 GFLOPS, pixel rate is 16.29 GPixel/s, and texture rate is 32.58 GTexel/s. The clocks are 756 MHz base and 1018 MHz boost. These are specification-derived limits, not measured benchmark results. The zero average score likely indicates that no recorded runs exist in the fact pack rather than a literal zero-performance result. The empty benchmarks array means there are no individual workload scores for any specific application in the data. In the absence of direct rivals, the 50th percentile should be read as a provisional placement based on the database's aggregate ranking, not as a verified rival comparison.