GPU Comparison

The NVIDIA GeForce 920MX and NVIDIA GeForce GT 740 are both end-of-life discrete graphics solutions from NVIDIA, but they serve fundamentally different segments of the market. The 920MX is a mobile-first MXM module built on the Maxwell architecture, while the GT 740 is a desktop single-slot card based on the older Kepler design. Benchmark data shows a split decision: the 920MX wins in OpenCL compute, while the GT 740 dominates in Vulkan performance. Neither card escapes the bottom quartile of all GPUs, both sit at the 21st percentile, but their individual strengths point to very different use cases.

Head-to-Head Benchmarks

The most striking result in the head-to-head comparison is the Vulkan gap, where the GT 740 wins decisively. The GT 740 scores 4083 in geekbench_vulkan, while the 920MX manages only 2987. That is a 26.8 percent advantage for the desktop card, and it is the single largest performance differential between the two in any test. For any workload that leverages the Vulkan API, the GT 740 is the clear choice. The margin is substantial enough to affect real-world gaming or compute tasks that scale with GPU throughput.

However, the 920MX turns the tables in OpenCL compute. The mobile chip scores 4068 in geekbench_opencl, edging out the GT 740's 3847. That is a 5.7 percent win for the 920MX. While the margin is far narrower than the Vulkan gap, it shows that the Maxwell architecture's efficiency translates into competitive compute performance despite a much smaller physical footprint. The 920MX's OpenCL score also exceeds its own average benchmark score of 3528 by roughly 15 percent, indicating that OpenCL is a particularly strong workload for this chip.

Looking at the aggregate benchmarks, the two cards are remarkably close. The 920MX has an average benchmark score of 3528, while the GT 740 averages 3431. That puts the 920MX roughly 2.8 percent ahead on average, but the difference is within the noise of the nearestRivals data. For example, the GT 740's nearest rival is the Quadro 2000M at 3434, a delta of just -0.1 percent, while the 920MX's nearest rival is the GT 545 at 3594, a delta of -1.8 percent. Neither card is far from a cluster of similarly performing GPUs, which suggests that head-to-head results are highly workload-dependent.

The OpenCL results also deserve context against the wider field. The 920MX's 4068 OpenCL score places it ahead of the GT 740's 3847, but both are outperformed by the GT 545's average of 3594 in the rival list, though that number is an average, not a direct OpenCL comparison. The key takeaway is simple: if you need Vulkan performance, the GT 740 wins by a landslide; if you need OpenCL compute, the 920MX has a modest edge.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA GeForce 920MX, with an average benchmark score of 3528 compared to the GT 740's 3431. That is a difference of roughly 2.8 percent, though both sit at the 21st percentile of all GPUs.

Q: How do the two compare in Vulkan performance?

A: The GT 740 is dramatically faster in Vulkan, scoring 4083 versus the 920MX's 2987. That represents a 26.8 percent advantage for the desktop card.

Q: Is the 920MX better for OpenCL workloads?

A: Yes, the 920MX scores 4068 in geekbench_opencl, beating the GT 740's 3847 by 5.7 percent. This is the only benchmark where the mobile chip comes out ahead.

Q: What are the closest rivals to each card?

A: For the 920MX, the nearest rival is the GT 545 with an average score of 3594, a delta of -1.8 percent. For the GT 740, the nearest rival is the Quadro 2000M at 3434, a delta of -0.1 percent.

Q: Which card has a higher pixel rate?

A: Both cards have an identical pixel rate of 7.944 GPixel/s, despite the GT 740 having double the ROPs (16 versus 8).

Q: What is the memory bandwidth difference?

A: The GT 740 has a massive bandwidth advantage, delivering 80.19 GB/s over a 128-bit bus with GDDR5 memory. The 920MX manages only 14.40 GB/s via a 64-bit DDR3 interface.

Architecture Differences

The two cards are built on different NVIDIA architectures, which explains much of their divergent behavior. The 920MX uses the GM108S chip based on Maxwell, fabricated on a 28 nm process at TSMC. The GT 740 uses GK107 based on the older Kepler architecture, also on TSMC's 28 nm node. Despite the same process node, the transistor budgets differ significantly: the GT 740 packs 1,270 million transistors on a 118 mm² die, while the 920MX has 1,020 million transistors on just 77 mm². That gives the 920MX a higher transistor density of 13.2M per mm² versus 10.8M per mm² for the GT 740.

The Maxwell architecture in the 920MX brings a newer feature set, including Vulkan 1.4 support, whereas the GT 740's Kepler design only reaches Vulkan 1.2.175. Both cards support DirectX 12 (11_0) and OpenGL 4.6, so those APIs are not differentiators. The 920MX also has a lower TDP of 16 W, reflecting its mobile design, while the GT 740 draws 64 W and requires a 6-pin power connector with a suggested 250 W PSU.

The memory subsystems are completely different in design. The 920MX uses 2 GB of DDR3 on a 64-bit bus, yielding 14.40 GB/s of bandwidth. The GT 740 uses 1 GB of GDDR5 on a 128-bit bus, delivering 80.19 GB/s, nearly six times the bandwidth. This explains why the GT 740 excels in Vulkan, which often benefits from higher memory throughput. The GT 740 also has more shading units (384 versus 256), more TMUs (32 versus 24), and double the ROPs (16 versus 8).

Specification Differences

The most consequential specification difference is memory bandwidth: the GT 740's 80.19 GB/s dwarfs the 920MX's 14.40 GB/s. The GT 740 also has a wider 128-bit bus versus 64-bit, and uses GDDR5 instead of DDR3. Memory capacity differs as well, the 920MX has 2 GB, while the GT 740 has 1024 MB.

Compute resources are another clear differentiator. The GT 740 has 384 shading units, 32 TMUs, and 16 ROPs, versus the 920MX's 256 shading units, 24 TMUs, and 8 ROPs. This results in higher texture rate for the GT 740 (31.78 GTexel/s versus 23.83 GTexel/s) and higher FP32 throughput (762.6 GFLOPS versus 508.4 GFLOPS). The pixel rate is identical at 7.944 GPixel/s despite the ROP difference.

Power and physical design differ sharply. The 920MX is an MXM module with a 16 W TDP and no power connectors, while the GT 740 is a single-slot card with a 64 W TDP, one 6-pin connector, and a suggested 250 W PSU. The GT 740 measures 145 mm (5.7 inches) in length. The bus interface also differs: the 920MX uses PCIe 3.0 x8, while the GT 740 uses PCIe 3.0 x16. Display outputs vary, with the GT 740 offering 2x DVI and 1x mini-HDMI 1.4a, while the 920MX is portable-device dependent.

Where Each One Wins

The GT 740 is the clear winner for any workload that relies on Vulkan. Its 4083 Vulkan score versus the 920MX's 2987 makes it the superior choice for gaming or compute applications built on that API. The 26.8 percent advantage is the largest margin in any benchmark, and it aligns with the GT 740's massive memory bandwidth advantage (80.19 GB/s versus 14.40 GB/s) and higher raw compute resources. The GT 740 also wins on raw FP32 performance (762.6 GFLOPS versus 508.4 GFLOPS) and texture rate (31.78 GTexel/s versus 23.83 GTexel/s), making it the better pick for fill-rate-bound scenarios.

The 920MX wins in OpenCL compute, scoring 4068 versus 3847. This is a narrower victory, but it suggests that Maxwell's architecture is more efficient for certain compute kernels. The 920MX also has a higher average benchmark score (3528 versus 3431), a higher transistor density (13.2M per mm² versus 10.8M), and a significantly lower TDP of 16 W versus 64 W. For environments where power draw is critical, such as thin-and-light laptops, the 920MX is the only viable option, as the GT 740 is a desktop card with a 6-pin power requirement.

Memory capacity favors the 920MX at 2 GB versus 1 GB, which can matter for holding larger textures or datasets in VRAM, though the GT 740's much faster GDDR5 mitigates that advantage in practice. The 920MX also supports Vulkan 1.4, a newer API version than the GT 740's 1.2.175.

The Verdict

The data points to a simple conclusion: the GT 740 is the stronger performer for graphics-heavy tasks, while the 920MX has a niche in low-power compute. If Vulkan performance is the priority, the GT 740 wins outright, a 26.8 percent lead is decisive and reflects its superior memory bandwidth, shading units, and texture rate. The GT 740 is also the better choice for any workload that benefits from raw FP32 throughput or higher fill rates, given its 762.6 GFLOPS and 31.78 GTexel/s.

The 920MX, however, is not a card to dismiss. Its OpenCL score of 4068 beats the GT 740's 3847, and its average benchmark score of 3528 is higher than the GT 740's 3431. For compute-heavy applications that lean on OpenCL, the 920MX offers competitive performance at a fraction of the power draw, 16 W versus 64 W. This makes it the only sensible pick for mobile or power-constrained systems.

Users should pick the GT 740 if they need maximum graphics throughput, particularly in Vulkan-based games or applications, and can accommodate a desktop card with a 6-pin connector. The 920MX is the choice for portable systems where power efficiency is paramount, or for OpenCL compute tasks where the newer Maxwell architecture gives it a slight edge. Neither card is a high performer, both sit at the 21st percentile of all GPUs, but within this head-to-head, the GT 740 delivers the more compelling overall package for traditional GPU workloads.