GPU Comparison

The NVIDIA GeForce GT 740M and NVIDIA GeForce 920MX represent two distinct approaches to mobile graphics from successive generations, yet the benchmark data reveals a surprisingly close contest. The GT 740M, built on the older Kepler 2.0 architecture, and the 920MX, on Maxwell, each claim one victory in the head-to-head tests, but the magnitude and nature of those wins paint a nuanced picture. The 920MX edges ahead in OpenCL compute, while the GT 740M delivers a dominant performance in Vulkan graphics, leaving the choice heavily dependent on the intended workload.

Where Each One Wins

The split between the two GPUs is cleanly defined by the benchmark type. In the geekbench_opencl test, the 920MX takes the win with a score of 4068 against the GT 740M's 3974, a margin of 2.3%. This indicates that for general-purpose compute tasks leveraging OpenCL, such as video encoding, physics simulations, or certain productivity applications, the 920MX holds a measurable, though modest, advantage. Its newer Maxwell architecture appears better optimized for this type of parallel workload, despite having fewer raw compute units.

Conversely, the geekbench_vulkan test is a decisive victory for the GT 740M, which scores 3459 compared to the 920MX's 2987. This is a substantial 15.8% lead, showing that in graphics-centric APIs like Vulkan, the older card's higher shading unit count and texture rate translate into significantly better performance. For gaming or any application that leverages Vulkan for rendering, the GT 740M is clearly the stronger option, outperforming its rival by a wide margin.

Looking beyond the direct comparison, the average benchmark scores reinforce this split. The GT 740M has an average score of 3717, while the 920MX sits at 3528. This puts the GT 740M in the 22nd percentile of all GPUs, slightly above the 920MX's 21st percentile. The data suggests that while the 920MX excels in one specific compute API, the GT 740M offers a more consistently higher performance across the board, primarily driven by its Vulkan strength.

Architecture Differences

The core architectural divide is generationally significant. The GT 740M is built on Kepler 2.0 architecture with the GK208 chip, while the 920MX utilizes Maxwell architecture with the GM108S chip. Both are manufactured by TSMC on the same 28 nm process node, and both pack the same 1,020 million transistors. The die size, however, differs slightly: the GT 740M has a larger die at 87 mm², while the 920MX is more compact at 77 mm². This results in a higher transistor density for the 920MX at 13.2M / mm² versus the GT 740M's 11.7M / mm², indicating a more efficient packing of transistors in the newer design.

The most consequential difference lies in the compute unit configuration. The GT 740M is equipped with 384 shading units, 32 TMUs, and 8 ROPs. In contrast, the 920MX has fewer of each: 256 shading units, 24 TMUs, and 8 ROPs. This gives the GT 740M a 50% advantage in shading units and a 33% advantage in texture mapping units. Consequently, the GT 740M achieves a higher pixel rate of 8.264 GPixel/s and a texture rate of 33.06 GTexel/s, compared to the 920MX's 7.944 GPixel/s and 23.83 GTexel/s. The FP32 performance follows suit, with the GT 740M delivering 793.3 GFLOPS versus the 920MX's 508.4 GFLOPS.

Despite the architectural differences, the memory subsystems are identical. Both GPUs feature 2 GB of DDR3 memory on a 64-bit bus, providing the same 14.40 GB/s of bandwidth. The clock speeds are also very close, with the GT 740M running at a base of 980 MHz and boost of 1033 MHz, while the 920MX sits slightly lower at 965 MHz base and 993 MHz boost. The memory clock is exactly the same at 900 MHz (or 1800 Mbps effective). The API support also shows a generational shift, with the 920MX supporting Vulkan 1.4 compared to the GT 740M's 1.2.175, though both support DirectX 12 (11_0) and OpenGL 4.6.

FAQ

Q: Which GPU has a higher average benchmark score?

A: The NVIDIA GeForce GT 740M has a higher average benchmark score of 3717, compared to the 920MX's 3528. This places the GT 740M in the 22nd percentile of all GPUs, while the 920MX is in the 21st percentile.

Q: Is the newer GeForce 920MX always faster than the older GT 740M?

A: No. The data shows a split. The 920MX is faster in the geekbench_opencl test (4068 vs 3974), but the GT 740M is significantly faster in the geekbench_vulkan test (3459 vs 2987), a 15.8% lead.

Q: How do their memory configurations compare?

A: They are identical. Both feature 2 GB of DDR3 memory on a 64-bit bus, resulting in the same 14.40 GB/s of memory bandwidth.

Q: What are the TDP differences between the two?

A: The GeForce 920MX has a significantly lower TDP of 16 W, while the GeForce GT 740M has a TDP of 33 W. This makes the 920MX more power-efficient.

Q: Which GPU has more shading units?

A: The NVIDIA GeForce GT 740M has more shading units, with 384 compared to the 920MX's 256. This contributes to its higher texture rate and FP32 performance.

Q: What is the production status of these GPUs?

A: Both GPUs are marked as end-of-life. The GT 740M was released in 2013, while the 920MX was released later in 2016.

Specification Differences

The specification sheet highlights several key differences between the two mobile GPUs. The most prominent is the shading units count, where the GT 740M holds a clear edge with 384 units versus the 920MX's 256. This leads to a corresponding advantage in TMUs (32 vs 24) and a significant lead in FP32 performance (793.3 GFLOPS vs 508.4 GFLOPS). The texture rate also favors the GT 740M at 33.06 GTexel/s compared to 23.83 GTexel/s, and the pixel rate is slightly higher at 8.264 GPixel/s versus 7.944 GPixel/s.

Architecturally, they differ in architecture (Kepler 2.0 vs Maxwell), chip (GK208 vs GM108S), and generation (GeForce 700M vs GeForce 900M). The die size is different (87 mm² vs 77 mm²), which, combined with the same transistor count, yields a different transistor density (11.7M / mm² vs 13.2M / mm²). The base and boost clocks are slightly different, with the GT 740M running at 980 MHz / 1033 MHz and the 920MX at 965 MHz / 993 MHz. The TDP is a major differentiator, with the 920MX drawing only 16 W compared to the GT 740M's 33 W. Finally, the Vulkan API support differs, with the 920MX supporting version 1.4 versus the GT 740M's 1.2.175. All other fields, including memory size, type, bus width, bandwidth, ROPs, and bus interface, are identical.

Head-to-Head Benchmarks

The direct comparison in the head-to-head benchmarks provides the clearest picture of their relative strengths. In the geekbench_opencl test, the GeForce 920MX wins with a score of 4068 against the GT 740M's 3974. The deltaPct of -2.3% (from the perspective of the GT 740M) quantifies this narrow defeat. This result is interesting because the GT 740M has significantly higher theoretical FP32 throughput, yet the 920MX still manages to outperform it in this specific OpenCL workload. This suggests that the Maxwell architecture's efficiency and scheduling are better suited to the types of operations in this test, or that the benchmark is not purely limited by raw shader count.

The geekbench_vulkan test tells a completely different story. Here, the GT 740M wins decisively with a score of 3459, leaving the 920MX far behind at 2987. The deltaPct of 15.8% is a substantial margin, indicating a dominant performance. This result aligns perfectly with the hardware specifications: the GT 740M's 50% more shading units and 33% more TMUs provide a massive advantage in a graphics API that can fully utilize these resources. The 920MX's lower clock speeds and reduced compute configuration simply cannot keep pace in this rendering-focused workload.

The overall win count is one apiece, but the data suggests the wins are not of equal weight. The GT 740M's Vulkan victory is a landslide, while the 920MX's OpenCL win is a narrow margin. This asymmetry implies that for any application that leans on graphics rendering, the GT 740M is the superior choice by a wide margin, whereas the 920MX's advantage is limited to a specific subset of compute tasks and is much less pronounced. The average benchmark scores corroborate this, as the GT 740M's overall average of 3717 is higher than the 920MX's 3528, reflecting the impact of its dominant Vulkan performance.

The Verdict

The data directs different users to different cards. For those primarily concerned with graphics performance, particularly in Vulkan-based applications or games, the NVIDIA GeForce GT 740M is the clear winner. Its 15.8% lead in the Vulkan benchmark is a decisive margin, and its higher texture and pixel rates provide a tangible benefit in rendering workloads. The GT 740M's higher average benchmark score and superior shading unit count make it the more capable graphics processor.

For users prioritizing compute performance in OpenCL environments, the NVIDIA GeForce 920MX is the better option. Its narrow 2.3% win in the OpenCL benchmark, while modest, indicates better optimization for this API. Furthermore, its significantly lower TDP of 16 W versus 33 W makes it a far more power-efficient choice for thin-and-light laptops where battery life and thermals are critical. The 920MX also offers newer Vulkan API support (1.4 vs 1.2.175), which may be relevant for future compatibility.

Ultimately, the choice hinges on the primary use case. If the laptop is intended for gaming or graphics-intensive tasks, the GT 740M's raw rendering power makes it the superior pick despite being older. If the priority is efficient compute acceleration or maximum battery life, the 920MX's architecture and low power draw make it the more sensible option. The benchmark data does not declare an overall winner, but rather presents a clear trade-off between graphics throughput and compute efficiency.