NVIDIA GeForce GT 645M vs NVIDIA GeForce GT 740M Comparison

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA GeForce GT 645M records an average benchmark score of 4411, placing it in the 26th percentile of all GPUs. The GeForce GT 740M scores 3717 on average, which lands in the 22nd percentile.

Q: How do the two compare in OpenCL performance?

A: The GeForce GT 740M wins the OpenCL test with a score of 3974 versus 2680 for the GT 645M. That represents a 32.6% advantage for the GT 740M in this workload.

Q: Which GPU wins the Vulkan benchmark?

A: The GeForce GT 645M takes the Vulkan test with a score of 4875, while the GT 740M scores 3459. The GT 645M leads by 40.9% in this API.

Q: What are the core counts for each GPU?

A: Both GPUs feature 384 shading units and 32 texture mapping units. The GT 645M has 16 ROPs, while the GT 740M has 8 ROPs.

Q: What is the memory configuration difference?

A: Both have 2 GB of DDR3 memory, but the GT 645M uses a 128-bit bus with 28.80 GB/s bandwidth. The GT 740M uses a 64-bit bus with 14.40 GB/s bandwidth.

Q: What is the transistor density difference?

A: The GT 645M uses the GK107 chip with 1,270 million transistors on a 118 mm² die, giving a density of 10.8M per mm². The GT 740M uses the GK208 chip with 1,020 million transistors on an 87 mm² die, giving 11.7M per mm².

Architecture Differences

The GeForce GT 645M is built on the GK107 chip using the original Kepler architecture, while the GeForce GT 740M uses the GK208 chip with the Kepler 2.0 architecture. Both are fabricated by TSMC on a 28 nm process node, but the underlying chip designs diverge significantly.

The GT 645M's GK107 die measures 118 mm² and packs 1,270 million transistors, resulting in a density of 10.8 million transistors per square millimeter. The GT 740M's GK208 die is smaller at 87 mm², holding 1,020 million transistors for a higher density of 11.7 million per mm². This means the GT 740M achieves greater transistor packing efficiency despite having fewer total transistors.

Clock speeds tell a clear story. The GT 645M runs at a base clock of 709 MHz with a boost of 780 MHz. The GT 740M operates substantially higher, with a 980 MHz base and 1033 MHz boost. This clock advantage directly feeds into compute rates: the GT 740M delivers 793.3 GFLOPS of FP32 throughput versus 599.0 GFLOPS for the GT 645M, a 32.4% gap in raw floating-point capability.

Texture and pixel processing also differ. The GT 645M achieves a texture rate of 24.96 GTexel/s and a pixel rate of 6.240 GPixel/s. The GT 740M, despite having the same 32 TMUs, reaches 33.06 GTexel/s due to its higher clocks. However, with only 8 ROPs compared to 16, the GT 740M's pixel rate of 8.264 GPixel/s benefits from the clock speed but loses parallelism.

Memory architecture presents a trade-off. Both use 2 GB of DDR3 with the same 900 MHz memory clock and 1800 Mbps effective data rate. The GT 645M uses a 128-bit bus for 28.80 GB/s bandwidth, double the GT 740M's 64-bit bus and 14.40 GB/s. This means the GT 645M has a significant memory bandwidth edge, which can matter in bandwidth-sensitive workloads.

The GT 645M is an integrated graphics processor (IGP) with a 32 W TDP, while the GT 740M is a modular MXM design rated at 33 W. Both use PCIe 3.0, but the GT 645M connects via x16 lanes while the GT 740M uses x8. Display outputs are portable-device dependent for both, and neither requires external power connectors.

Head-to-Head Benchmarks

The recorded data shows a split decision between these two mobile GPUs. Each wins one of the two shared benchmark tests, and the margins are substantial in both directions.

In Geekbench OpenCL, the GeForce GT 740M dominates with a score of 3974 against the GT 645M's 2680. The delta of 32.6% is the largest single-test gap in this comparison. This result aligns with the GT 740M's higher clock speeds and greater FP32 throughput. OpenCL workloads that scale with compute units and clock frequency will favor the GT 740M, and the data confirms this expectation.

The Geekbench Vulkan test reverses the outcome. The GT 645M scores 4875, while the GT 740M manages 3459. The GT 645M wins by 40.9%, an even larger margin than the GT 740M's OpenCL victory. Vulkan performance often depends on memory bandwidth and ROP throughput, where the GT 645M holds clear advantages with its 128-bit bus and 16 ROPs. The GT 645M's wider memory interface appears to provide a decisive edge in this API.

Looking at the broader context, the GT 645M's average score of 4411 places it just 0.5% above the NVIDIA GeForce 930M (4388) and 1.2% above the Intel Iris Pro Graphics 5200 (4360). It trails the AMD Radeon R7 M260 (4499) by 1.9%, and sits 1.8% ahead of the NVIDIA GeForce RTX 4070 GDDR6 (4335), a surprising comparison that reflects the benchmark's weighting.

The GT 740M's average of 3717 puts it essentially tied with the NVIDIA Quadro 3000M (3718, 0% delta). It is 0.6% ahead of the GeForce 825M (3694) and 1.9% ahead of the AMD Radeon HD 6770 (3649), but 0.6% behind the GeForce GT 635M (3740). These narrow margins place the GT 740M in a tightly contested performance tier.

The head-to-head results show that neither GPU is universally superior. The GT 740M wins compute-heavy OpenCL tasks, while the GT 645M wins Vulkan workloads. Users should consider which API and application types matter most for their specific use case.

Specification Differences

| Specification | GeForce GT 645M | GeForce GT 740M |

|---|---|---|

| Chip | GK107 | GK208 |

| Architecture | Kepler | Kepler 2.0 |

| Generation | GeForce 600M | GeForce 700M |

| Transistors | 1,270 million | 1,020 million |

| Die Size | 118 mm² | 87 mm² |

| Transistor Density | 10.8M / mm² | 11.7M / mm² |

| Base Clock | 709 MHz | 980 MHz |

| Boost Clock | 780 MHz | 1033 MHz |

| Memory Bus Width | 128 bit | 64 bit |

| Memory Bandwidth | 28.80 GB/s | 14.40 GB/s |

| ROPs | 16 | 8 |

| Pixel Rate | 6.240 GPixel/s | 8.264 GPixel/s |

| Texture Rate | 24.96 GTexel/s | 33.06 GTexel/s |

| FP32 | 599.0 GFLOPS | 793.3 GFLOPS |

| TDP | 32 W | 33 W |

| Slot Width | IGP | MXM Module |

| Bus Interface | PCIe 3.0 x16 | PCIe 3.0 x8 |

| Release Date | 2012-09-30 | 2013-06-19 |

| Predecessor | GeForce 500M | GeForce 600M |

| Successor | GeForce 700M | GeForce 800M |

Identical specifications include the 28 nm process node, TSMC foundry, 2 GB DDR3 memory, 384 shading units, 32 TMUs, 900 MHz memory clock, 1800 Mbps effective memory rate, DirectX 12 (11_0), OpenGL 4.6, Vulkan 1.2.175, and the end-of-life production status.

The Verdict

The benchmark data presents a clear but nuanced choice. The GeForce GT 645M holds the higher average score at 4411 versus 3717, a 18.7% difference in overall performance. It also wins the Vulkan test by a decisive 40.9% margin. The GT 740M counters with a 32.6% OpenCL victory and superior raw compute metrics, including higher clocks, greater FP32 throughput, and faster texture fill.

For users prioritizing average benchmark performance and Vulkan-based applications, the GT 645M is the stronger option. Its wider memory bus and double ROP count give it advantages that show up in the recorded Vulkan score. The 26th percentile ranking versus the GT 740M's 22nd percentile further supports this conclusion.

However, the GT 740M's OpenCL performance is substantial. The 3974 score approaches the GT 645M's Vulkan result, and the 793.3 GFLOPS FP32 throughput indicates strong compute capability. Applications that leverage OpenCL and benefit from higher clock speeds will run better on the GT 740M.

The release timeline also matters. The GT 645M launched on 2012-09-30, while the GT 740M arrived on 2013-06-19. The GT 740M represents a newer generation with a different chip design, but the data shows this does not translate into overall superiority.

Where Each One Wins

NVIDIA GeForce GT 645M wins in:

• Overall average benchmark score: 4411 versus 3717, a 18.7% lead
• Vulkan performance: 4875 versus 3459, a 40.9% advantage
• Memory bandwidth: 28.80 GB/s versus 14.40 GB/s, double the bandwidth
• ROP count: 16 versus 8, providing better pixel processing parallelism
• Bus interface width: PCIe 3.0 x16 versus x8
• GPU percentile ranking: 26th versus 22nd

NVIDIA GeForce GT 740M wins in:

• OpenCL performance: 3974 versus 2680, a 32.6% advantage
• Clock speeds: 980 MHz base and 1033 MHz boost versus 709 MHz and 780 MHz
• FP32 throughput: 793.3 GFLOPS versus 599.0 GFLOPS
• Texture rate: 33.06 GTexel/s versus 24.96 GTexel/s
• Pixel rate: 8.264 GPixel/s versus 6.240 GPixel/s
• Transistor density: 11.7M per mm² versus 10.8M per mm²

The use-case split is straightforward. The GT 645M suits workloads that rely on memory bandwidth, such as Vulkan-based games and applications where texture data must move quickly. The GT 740M fits compute-oriented tasks that scale with clock speed and FP32 throughput, particularly OpenCL workloads. Users should match their primary applications to the GPU that wins the relevant benchmark.