NVIDIA GeForce MX450 30.5W 8Gbps

The NVIDIA GeForce MX450 30.5W 8Gbps is an end-of-life Turing part built around the TU117S chip and manufactured on TSMC's 12 nm process. The die contains 4,700 million transistors on a 200 mm² footprint, giving a transistor density of 23.5M per mm². It belongs to the GeForce MX (4xx) generation, with a base clock of 1035 MHz and a boost clock of 1275 MHz. The memory runs at 2000 MHz with 8 Gbps effective, and the listed TDP is 31 W while the SKU name carries a 30.5W designation. The release date is 2020-08-24, and the database lists no benchmark entries and no nearest rivals for this product.

Memory Subsystem

The memory subsystem is 2 GB of GDDR5 on a 64-bit memory bus. Memory clock is 2000 MHz, which the data sheet records as 8 Gbps effective, producing a bandwidth of 64.00 GB/s. That 64-bit bus width is the defining constraint of this part: a narrow interface limits how much data can move between the GPU and memory per clock, independent of the shader array speed. The pixel rate is 40.80 GPixel/s and the texture rate is 71.40 GTexel/s, which means the render outputs and texture units can generate substantial work. The memory bus must feed those ROPs and TMUs while also streaming textures, geometry, and render targets.

For high-resolution workloads, both the capacity and the bandwidth are limiting factors. A 2 GB frame buffer can hold a small set of textures and buffers; once the working set exceeds that, performance degrades as data movement dominates. The 64.00 GB/s bandwidth is a modest figure, so high-resolution rendering with large textures will run into the memory path quickly. Lower resolutions with smaller texture budgets are a better fit for the memory subsystem. The data suggests that the GPU is designed around restrained memory pressure, not around moving large high-resolution frames.

Ray Tracing and Feature Set

The specification lists no RT cores and no tensor cores. Even though the architecture is Turing, this particular implementation omits the dedicated hardware commonly associated with that family. Hardware-accelerated ray tracing is not part of the feature set according to the data. Any ray tracing effects would have to run using the 896 shading units, which would divert compute throughput away from general rendering work. There is no tensor-accelerated compute path listed either, so features that depend on tensor cores are not available from this specification.

API support is one area where the data shows a modern baseline. The GPU supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. That means the software driver surface is current even if the dedicated hardware features are not present. DirectX feature level 12_1 is not the highest feature level available, and with no RT cores or tensor cores in the record, the practical feature ceiling is set by conventional shader-based rendering. The RT and tensor fields are null, and the data sheet does not provide any alternate acceleration path.

Power and Cooling

The TDP is listed as 31 W. This is a low-power part, and the power connector field explicitly reads "None", which means no auxiliary PCIe power cables are needed. The only listed interface is PCIe 4.0 x4, which also carries the data connection. No suggested PSU figure appears in the record, so the database provides no PSU wattage recommendation to repeat here. The absence of power connectors and the 31 W TDP are the complete power delivery story from the data.

Physical cooling specifications are largely absent. Slot width is null, and dimensions for length, height, and width are not provided. Display outputs are listed as portable device dependent, so the video output implementation is determined by the host device rather than the GPU board itself. That makes this part inherently different from a fixed desktop add-in card; its thermal profile and connector layout are tied to the portable system it is installed in. With no dimensions and no slot width, this record cannot be used to evaluate heatsink clearance or fan placement.

How It Compares

The nearestRivals array is empty in the data. There are no rival names, no comparison scores, and no deltaPct values to anchor a head-to-head assessment. Because the benchmark list is also empty, there is no measured score to place above or below any competitor. This SKU cannot be positioned against specific GPUs from the data in this record.

The only global rank available is percentileVsAllGpus, which is 50. That rank places the part at the midpoint of the database's GPU distribution, but with no underlying benchmark entries, it cannot be interpreted as a measured competitive result. The absence of rivals is itself notable: for a product with no nearestRivals data, the database has no direct comparison frame. A paragraph-per-rival structure cannot be constructed when the record contains no rival entries at all.

Benchmark Performance

The benchmark array is empty, and the average benchmark score is listed as 0. That should be read as "no aggregated runs exist", not as a claim of zero actual performance. The percentileVsAllGpus value is 50, which is the exact midpoint of all GPUs in the database. With no individual scores feeding that percentile, the percentile has no workload-derived basis in this record. There are no exact percentage deltas to report against rivals because no rival benchmark scores are present.

The data sheet does include raw throughput figures. FP32 compute is 2.285 TFLOPS, and FP16 compute is 4.570 TFLOPS at a 2:1 ratio. The base clock is 1035 MHz and the boost clock is 1275 MHz, with 896 shading units, 56 TMUs, and 32 ROPs. Pixel throughput is 40.80 GPixel/s and texture throughput is 71.40 GTexel/s. These figures give a measurable computational envelope even without game-level benchmark scores. The boost clock of 1275 MHz is the highest frequency listed, so the FP32 and FP16 figures represent throughput at that specified boost state.

The 2:1 FP16 ratio is explicit in the data: FP16 performance is double FP32 performance. That is useful for workloads that can use half-precision math. The pixel and texture rates, combined with the 64-bit memory bus, paint a picture of a small shader-capable GPU whose output is throttled by memory bandwidth in demanding scenes. Without benchmark scores, the percentile value of 50 cannot be validated against actual frame rates.

Who Should Consider It

This GPU suits a portable system with modest graphics requirements. The 31 W TDP, the lack of auxiliary power connectors, and the portable-device-dependent display outputs all point toward compact hosts rather than large desktop builds. The 2 GB memory capacity and 64.00 GB/s bandwidth set a practical ceiling: lower resolutions and moderate settings are appropriate, while high-resolution textures will stress both capacity and bandwidth. With no RT cores or tensor cores, users should not expect hardware-accelerated ray tracing or tensor-core features. The API list — DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4 — provides broad software compatibility for existing drivers and applications.

This is an end-of-life product. For users already maintaining a system that uses this GPU, the data offers a clear set of constraints: keep the memory working set small, avoid high-resolution rendering demands, and use the API features that do not require dedicated RT or tensor hardware. The 896 shading units and 2.285 TFLOPS FP32 throughput are enough for lighter 3D workloads, but the 2 GB frame buffer is the hard boundary. Anyone expecting ray tracing, high-resolution performance, or large textures will need a different GPU. The record shows a low-power, portable-oriented part whose practical role is confined to moderate graphics tasks.