NVIDIA GeForce GT 740 vs NVIDIA GeForce GT 745M Comparison

Head-to-Head Benchmarks

The recorded data splits the benchmark results between these two mobile graphics parts in a clear pattern. The GT 745M wins two of the three head-to-head tests, while the GT 740 takes one. The largest single win for the GT 745M comes in the Vulkan test, where it scores 5502 against the GT 740's 4083, a 34.8% advantage. The OpenCL test is the one win for the GT 740, with a score of 3847 versus 3580 for the 745M, a delta of -6.9% in favor of the 740. The Metal test sits in between: the 745M scores 2777, the 740 scores 2363, giving the 745M a 17.5% lead.

Looking at the average benchmark score across all three tests, the GT 745M records 3953, while the GT 740 records 3431. That is an overall average lead of about 15% for the 745M. The percentile data reinforces this ordering: the 745M sits at the 23rd percentile of all GPUs, while the 740 sits at the 21st percentile. In short, the 745M is the faster card in two of three API tests, and it holds a consistent single-digit to mid-double-digit advantage in raw performance.

Architecture Differences

Both chips are built on the same GK107 silicon and the Kepler architecture, which simplifies the comparison. The 745M has a transistor count of 1,270 million on a 118 mm² die, while the 740 has 1,270 million transistors on the same 118 mm² die. The transistor density is therefore effectively identical, at 10.8M per mm² for both. Neither chip has a full list of the higher-end features; there are no ray tracing cores, no tensor cores, and no additional graphics features beyond the base Kepler feature set.

The memory configuration is the first major split. The 745M is equipped with 2 GB of GDDR5 on a 128-bit bus, yielding a bandwidth of 64.00 GB/s. The 740 has 1024 MB of GDDR5, also on a 128-bit bus, producing 80.19 GB/s of bandwidth. The 745M has double the capacity but lower bandwidth; the 740 has a narrower memory pool but a higher data rate. Memory clocks also differ: the 745M runs at 1000 MHz (4 Gbps effective), while the 740 runs at 1253 MHz (5 Gbps effective).

The compute configuration shifts as well. The 745M carries 384 shading units, 32 texture mapping units, and 16 ROPs. The 740 has 384 shading units, 32 TMUs, and 16 ROPs. So the core math pipeline is identical in width on both cards. The 745M records a pixel rate of 4.392 GPixel/s and a texture rate of 17.57 GTexel/s. The 740 reaches 7.944 GPixel/s and 31.78 GTexel/s, which are higher on both fronts. However, the 745M counters with a higher FP32 throughput: 421.6 GFLOPS versus 762.6 GFLOPS for the 740.

Both cards use the same PCIe 3.0 x16 bus interface, and both are single-slot designs. The 745M draws power from a single 6-pin connector, the 740 has no additional power connectors and relies on the slot alone. The 745M has a TDP of 45 W, the 740 has a TDP of 64 W. The 745M is therefore the more power-efficient part. The 740 supports up to a 250 W power supply, the 745M is not listed with a specific PSU recommendation in the database.

Where Each One Wins

Looking strictly at the data, the use cases split by API. In the Vulkan test, the GT 745M is the clear winner: it scores 5502 against 4083, a 34.8% lead. If a workload is heavily Vulkan-based, the GT 745M is the stronger choice. In the OpenCL test, the situation flips: the GT 740 wins, 3847 versus 3580, a 6.9% advantage. The OpenCL result is the only head-to-head win for the 740, and it is a narrow one. Compute-oriented workflows on OpenCL will favor the GT 740.

In the Metal benchmark, the 745M again wins, with 2777 against 2363, a 17.5% margin. For users on Windows platforms where Metal is the primary API, the 745M has the better score. The two Geekbench results are not uniform; each card leads in a specific API. The average however favors the 745M because its wins are larger than the 740's single win.

The 745M's higher FP32 throughput and larger memory capacity also suggest it will hold up better in sustained compute workloads. The 740's higher pixel rate and texture rate point to it being a better fit for display-bound tasks. The 740's bandwidth advantage is useful in memory-intensive scenes, but the 745M's larger memory pool compensates in most real configurations.

The Verdict

The data points are consistent. The GT 745M is the faster card in Vulkan and Metal, and the GT 740 is the faster in OpenCL. If a user's software stack is built around Vulkan or Metal, the 745M has a clear performance edge. If the software uses OpenCL, the 740 has a modest edge. The benchmark data does not support the 740 as a general winner; it only wins in one API. The 745M wins in two, and by larger margins.

The all-around percentile is also close: both around the 21st and 23rd percentile of all GPUs. Neither card is a high-end part, both are positioned in the low-mid range of the database. Users looking for a mobile Kepler-class card with a specific API priority should choose based on the benchmark results above. The 745M is the better pick for Vulkan and Metal performance. The 740 is the pick for OpenCL. For general mobile use, the 745M's higher average score (3953 versus 3431, a 15% advantage) makes it the more balanced choice, especially when the API varies.

FAQ

Q: Which card wins in the Vulkan benchmark?

A: The NVIDIA GeForce GT 745M wins with a score of 5502, a 34.8% lead over the GT 740's 4083.

Q: Is the GT 740 faster in any test?

A: Yes, in the OpenCL test it scores 3847 versus 3580, a 6.9% advantage.

Q: Do both cards have the same memory bus width?

A: Yes, both use a 128-bit bus, but the 745M has 2 GB of GDDR5, while the 740 has 1024 MB.

Q: Which card has a higher pixel rate?

A: The GT 740 has a higher pixel rate: 7.944 GPixel/s versus 4.392 GPixel/s for the GT 745M.

Q: Which card uses less power?

A: The GT 745M has a TDP of 45 W, the GT 740 has 64 W.

Q: What is the API support?

A: Both support DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175.

Specification Differences

| Specification | NVIDIA GeForce GT 745M | NVIDIA GeForce GT 740 |

| Bus interface | PCIe 3.0 x8 | PCIe 3.0 x8 |

| Memory | 2 GB GDDR5 | 1024 MB GDDR5 |

| Memory bus | 128 bit | 128 bit |

| Bandwidth | 64.00 GB/s | 80.19 GB/s |

| Shading units | 384 | 384 |

| TMUs | 32 | 32 |

| ROPs | 16 | 16 |

| Pixel rate | 4.392 GPixel/s | 7.944 GPixel/s |

| Texture rate | 17.57 GTexel/s | 31.78 GTexel/s |

| FP32 | 421.6 GFLOPS | 762.6 GFLOPS |

| TDP | 45 W | 64 W |

| Power connectors | 1x 6-pin | None |

The two cards share the same GK107 chip, Kepler architecture, and 28 nm process at TSMC. The transistor density is effectively identical (10.8M / mm²). The 745M has twice the memory capacity, but the 740 has a higher bandwidth and a higher memory clock. All other compute resources are identical: 384 shading units, 32 TMUs, and 16 ROPs. The 740 has a higher pixel rate and texture rate, while the 745M has a higher FP32 throughput. The 745M requires a 6-pin power connector, while the 740 is slot powered only.