NVIDIA GeForce GT 745M

The NVIDIA GeForce GT 745M is a Kepler-architecture part built on TSMC's 28 nm process, packing 1,270 million transistors into a 118 mm² die with a transistor density of 10.8 million per square millimeter. It belongs to the GeForce 700M generation and is now end-of-life, having been released in March 2013. In the Geekbench OpenCL benchmark, it scores 3537 points, placing it at the 20th percentile of all GPUs in the database—meaning it outperforms only a fifth of the field. This score is nearly identical to several modern and professional parts, as detailed in the nearest rival comparisons below.

Memory Subsystem

The GT 745M is equipped with 2 GB of GDDR5 memory on a 128-bit bus, yielding a memory bandwidth of 64.00 GB/s. The memory clock is listed at 1000 MHz, which translates to 4 Gbps effective data rate. For a GPU of this era and class, the 128-bit interface and 64 GB/s bandwidth are modest figures. At high resolutions, the limited bandwidth becomes a bottleneck because each frame requires more texture fetches and framebuffer operations than at lower resolutions. The 2 GB capacity also restricts the amount of texture data and geometry that can be resident, making it unsuitable for modern high-resolution asset packs. Benchmark data shows that the GPU's overall performance is low, and the memory subsystem is a contributing factor—the pixel rate of 4.392 GPixel/s and texture rate of 17.57 GTexel/s are consistent with a part that will saturate its memory bus quickly when pushing many pixels. While the GDDR5 type is a step above DDR3, the narrow bus and relatively low bandwidth mean that users should expect significant performance degradation when moving from 720p-class workloads to 1440p or 4K, even if the GPU's compute capacity were sufficient.

Ray Tracing and Feature Set

The GT 745M does not include any dedicated ray tracing cores or tensor cores; both fields are null in the specification. This is expected for a Kepler-generation product, as hardware-accelerated ray tracing was not introduced until much later. The API support includes DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. The DirectX 12 support is at the 11_0 feature level, meaning the GPU can run DirectX 12 applications but with the feature set of the older 11_0 specification. This limits its ability to leverage newer DirectX 12 features such as mesh shaders or variable rate shading. Vulkan 1.2.175 provides a modern low-level API path, but without dedicated RT or tensor hardware, any ray-traced effects must be implemented via compute shaders, which will be extremely slow given the FP32 performance of 421.6 GFLOPS. The GPU's shading units total 384, with 32 texture mapping units and 16 ROPs, which are modest counts for a part that targets entry-level mobile graphics. The lack of tensor cores also precludes any AI-accelerated features like DLSS, which rely on dedicated hardware.

Power and Cooling

The GT 745M has a thermal design power (TDP) of 45 W, which is low by desktop standards but typical for a mobile integrated graphics solution. The slot width is listed as "IGP," indicating that the GPU is designed to be integrated directly into a motherboard or laptop logic board rather than installed as a separate card. No power connector is specified, and no suggested PSU is provided, which aligns with an integrated design that draws power from the system's main power delivery. The 45 W TDP means that cooling solutions can be modest—likely a small fan or heatpipe in a thin laptop chassis. This low power envelope makes the GT 745M suitable for systems where battery life and thermal management are priorities over raw performance. However, the 45 W figure also constrains clock speeds and shader activity, and the benchmark score reflects that constraint. Users should not expect to overclock or push this GPU beyond its factory settings, as the integrated nature and lack of discrete power connectors limit headroom.

Who Should Consider It

Given its 20th percentile standing and 2 GB VRAM, the GT 745M is best suited for light, older games and productivity tasks at lower resolutions. The memory bandwidth of 64 GB/s and pixel rate of 4.392 GPixel/s suggest that the GPU can handle 720p gaming with moderate settings in titles from its release era. For 1080p, the GPU will likely struggle with modern games, especially those with high-resolution textures or complex shading. The FP32 throughput of 421.6 GFLOPS is insufficient for heavy compute workloads, so users should not consider this part for machine learning or 3D rendering. The API support for Vulkan 1.2.175 and OpenGL 4.6 means that some current lightweight applications will run, but performance will be limited. In short, this is a legacy integrated GPU that is appropriate for basic web browsing, video playback, and very old or esports-level games at low settings. It is not a candidate for high-refresh-rate gaming or content creation.

Benchmark Performance

The sole benchmark result in the database is a Geekbench OpenCL score of 3537. This places the GT 745M at the 20th percentile, meaning it is slower than 80% of all GPUs tracked. The nearest rivals are revealing: the NVIDIA GeForce RTX 5060 Ti 16 GB scores 3577, a delta of -1.1% relative to the GT 745M—so the GT 745M is just 1.1% slower than a modern high-end card. Similarly, the RTX 5000 Mobile Ada Generation scores 3596, putting the GT 745M 1.6% behind. The GeForce GT 545, a desktop part from an older generation, scores 3643, which is 2.9% higher than the GT 745M. On the other side, the Quadro 2000M scores 3429, and the GT 745M is 3.1% ahead of it. These deltas are remarkably small, indicating that the GT 745M's OpenCL compute performance is roughly on par with these disparate cards—despite massive differences in architecture, memory, and power. This suggests that the OpenCL workload is not memory-bound or shader-bound in a way that differentiates these GPUs; instead, it may be limited by driver overhead or specific compute patterns. The near-identical scores also imply that the GT 745M, despite its age, can handle basic OpenCL compute tasks at a level comparable to much newer parts, though its gaming and graphics capabilities are far more limited due to its older feature set and lower bandwidth.

FAQ

Q: What is the memory configuration of the GT 745M?

A: It has 2 GB of GDDR5 memory on a 128-bit bus, providing 64.00 GB/s of bandwidth and a 4 Gbps effective memory clock.

Q: Does the GT 745M support hardware ray tracing?

A: No. The specification lists no ray tracing cores or tensor cores, so ray tracing is not hardware-accelerated.

Q: What is the TDP of this GPU?

A: The TDP is 45 W, and the slot width is listed as IGP, indicating an integrated design with no external power connector.

Q: How does it compare to the GeForce GT 545?

A: The GT 745M scores 3537 in OpenCL, which is 2.9% lower than the GT 545's score of 3643.

Q: What DirectX version does it support?

A: It supports DirectX 12 at the 11_0 feature level, along with OpenGL 4.6 and Vulkan 1.2.175.

Q: Is the GT 745M still in production?

A: No, its production status is end-of-life, and it was released in March 2013.

How It Compares

NVIDIA GeForce RTX 5060 Ti 16 GB

The RTX 5060 Ti scores 3577, just 1.1% higher than the GT 745M. Despite the massive generational and architectural gap, the OpenCL compute result is nearly identical. This suggests that the GT 745M's compute throughput, while low in absolute terms, is still competitive for certain parallel workloads that do not rely on memory bandwidth or modern instruction sets.

NVIDIA RTX 5000 Mobile Ada Generation

With a score of 3596, this professional mobile GPU is 1.6% ahead of the GT 745M. The Ada part features dedicated RT and tensor cores, but those are not exercised in the OpenCL benchmark. The close margin indicates that the GT 745M's Kepler shaders can hold their own in raw compute, though it lacks the specialized hardware that makes the RTX 5000 far more capable in modern applications.

NVIDIA GeForce GT 545

The GT 545 scores 3643, which is 2.9% higher than the GT 745M. Both are older Kepler-generation parts, but the GT 545 is a desktop card with a higher power envelope. The small delta implies that the GT 745M's mobile design does not severely hamper its OpenCL performance, though the desktop part still edges ahead.

NVIDIA Quadro 2000M

The Quadro 2000M scores 3429, and the GT 745M is 3.1% faster. This is a direct win for the GT 745M, showing that it outperforms a professional mobile GPU from the same era in compute tasks. The advantage is modest but consistent with the GT 745M's higher shader count and newer architecture.