GPU Comparison

The data presents a clear hierarchy between the NVIDIA GeForce GT 740 and the NVIDIA GeForce 820A, but the story is less about raw speed and more about architectural purpose. The GT 740, a dedicated desktop card, outclasses the 820A, an integrated notebook GPU, in every measurable category. However, the 820A’s existence is justified by its drastically lower power draw and its role in a different class of machine. The benchmarks show a 29% performance gap in favor of the GT 740 in the only shared test, but the underlying specifications reveal a fundamental divide in design philosophy.

The Verdict

The data points to a simple conclusion for most users: the NVIDIA GeForce GT 740 is the superior performer. Its average benchmark score of 3431 places it in the 21st percentile of all GPUs, while the 820A sits at a 19th percentile with an average score of 2983. In the direct Geekbench OpenCL comparison, the GT 740 scores 3847 against the 820A's 2983, a 29% advantage. For any workload involving graphics or compute, the GT 740 is the clear choice.

However, the 820A is not without its purpose. Its 15 W TDP is remarkably low compared to the GT 740's 64 W, and it requires no power connectors, making it suitable for ultra-portable or low-power systems where the GT 740's single-slot size and 6-pin power connector would be prohibitive. The 820A's score of 2983 in Geekbench OpenCL is competitive with its nearest rivals, including the NVIDIA GeForce GTX 860M (deltaPct 0.5) and GTX 750 Ti (deltaPct 1). This suggests that while it is slower than the GT 740, it is not an outlier in its own performance class.

The verdict, strictly from the data, is that the GT 740 is for users who need tangible performance in a desktop system, while the 820A is for those who need a basic GPU solution with minimal power and space requirements. There is no scenario in the benchmark data where the 820A wins, so the choice hinges entirely on system constraints versus performance demands.

Architecture Differences

The two GPUs represent distinct generations of NVIDIA design. The GT 740 is built on the Kepler architecture with the GK107 chip, while the 820A uses the older Fermi 2.0 architecture with the GF117 chip. This is a fundamental difference, as Kepler was a major architectural shift from Fermi, focusing on efficiency and new feature sets. Both are manufactured on the same 28 nm process at TSMC, but the transistor counts tell a different story. The GT 740 packs 1,270 million transistors into a 118 mm² die, achieving a density of 10.8M transistors per mm². The 820A has just 585 million transistors on a 116 mm² die, resulting in a much lower density of 5.0M per mm².

This disparity in transistor count directly impacts the compute resources. The GT 740 has 384 shading units, 32 texture mapping units (TMUs), and 16 render output units (ROPs). The 820A is far more modest, with only 96 shading units, 16 TMUs, and 8 ROPs. The GT 740's memory subsystem is also vastly superior, featuring 1024 MB of GDDR5 on a 128-bit bus, yielding a bandwidth of 80.19 GB/s. The 820A uses 1024 MB of DDR3 on a 64-bit bus, resulting in only 16.02 GB/s of bandwidth. The GT 740's memory clock is 1253 MHz (5 Gbps effective) versus the 820A's 1001 MHz (2 Gbps effective). These are not minor tweaks; they are structural differences that define the performance envelope.

The feature sets also diverge. The GT 740 supports Vulkan 1.2.175, while the 820A has no Vulkan support listed. Both support DirectX 12 (11_0) and OpenGL 4.6. The bus interface differs as well, with the GT 740 using PCIe 3.0 x16 and the 820A using the older PCIe 2.0 x16. The GT 740 is a single-slot card with a 6-pin power connector, while the 820A is an IGP (integrated graphics processor) with no power connectors, relying on the system's existing power delivery.

Head-to-Head Benchmarks

The only direct comparison available is the Geekbench OpenCL test, and it is decisive. The GT 740 scores 3847, while the 820A scores 2983. This yields a deltaPct of 29%, meaning the GT 740 is 29% faster. This is a substantial margin, almost a third more performance. The GT 740's nearest rivals in the overall database include the NVIDIA GeForce 920MX (avg score 3528, deltaPct -2.7) and the Intel HD Graphics 530 (avg score 3332, deltaPct 3). The 820A, in contrast, sits near the NVIDIA GeForce GTX 860M (avg score 2967, deltaPct 0.5) and the GTX 750 Ti (avg score 2953, deltaPct 1). The data suggests the GT 740 is competing in a higher tier of performance than the 820A.

The 29% lead in OpenCL is likely a direct consequence of the architectural advantages. The GT 740 has four times the shading units, double the TMUs and ROPs, and five times the memory bandwidth. The Fermi architecture of the 820A is older and less efficient per transistor compared to Kepler. The GT 740's memory bandwidth of 80.19 GB/s is particularly critical for compute tasks, as it prevents data starvation for the 384 shading units. The 820A's 16.02 GB/s bandwidth is a severe bottleneck for its 96 shading units, limiting its ability to execute complex shaders and compute kernels. The data shows a clear, linear relationship between these specifications and the observed benchmark scores.

FAQ

Q: Which GPU has a higher average benchmark score?

A: The NVIDIA GeForce GT 740 has an average benchmark score of 3431, which is higher than the NVIDIA GeForce 820A's average score of 2983. The GT 740 also sits in the 21st percentile of all GPUs, compared to the 820A's 19th percentile.

Q: How much faster is the GT 740 in the shared Geekbench OpenCL test?

A: The GT 740 scores 3847, while the 820A scores 2983. This gives the GT 740 a 29% lead in performance, as indicated by the deltaPct of 29 in the head-to-head data.

Q: What are the key memory differences between the two cards?

A: The GT 740 uses 1024 MB of GDDR5 memory on a 128-bit bus, providing 80.19 GB/s of bandwidth. The 820A uses 1024 MB of DDR3 memory on a 64-bit bus, providing only 16.02 GB/s of bandwidth. The GT 740's memory clock is 1253 MHz (5 Gbps effective) versus the 820A's 1001 MHz (2 Gbps effective).

Q: Does the 820A support Vulkan?

A: No, the FACT PACK lists no Vulkan support for the NVIDIA GeForce 820A. In contrast, the GT 740 supports Vulkan version 1.2.175.

Q: What is the power consumption difference?

A: The GT 740 has a TDP of 64 W and requires a 6-pin power connector, while the 820A has a TDP of only 15 W and requires no power connectors. The 820A is also listed as an IGP, whereas the GT 740 is a single-slot card.

Q: Which GPU is closer to the performance of the GeForce GTX 750 Ti?

A: The 820A is much closer, with an average score of 2983 compared to the GTX 750 Ti's 2953, a deltaPct of 1. The GT 740's score of 3431 is far above the GTX 750 Ti's score.

Where Each One Wins

The GT 740 wins in every performance scenario defined by the data. It is 29% faster in OpenCL, has higher raw compute throughput (762.6 GFLOPS vs 240.0 GFLOPS), and a significantly higher texture rate (31.78 GTexel/s vs 10.00 GTexel/s) and pixel rate (7.944 GPixel/s vs 2.500 GPixel/s). For any application that is GPU-bound, such as gaming, video editing, or GPU-accelerated compute, the GT 740 is the only logical choice. Its support for Vulkan also makes it more future-proof for modern graphics APIs.

The 820A wins in the domain of system integration and power efficiency. Its 15 W TDP is a fraction of the GT 740's 64 W, making it ideal for thin-and-light laptops or small form factor systems where heat and power are critical constraints. It requires no additional power connectors and is an IGP, meaning it is soldered onto the motherboard, saving space. The 820A's performance, while lower, is still competitive with discrete mobile GPUs like the GTX 860M (deltaPct 0.5), suggesting it can handle basic tasks. The choice is not about which is faster, the GT 740 wins that outright, but about whether the system can physically accommodate the GT 740's power and size requirements.

Specification Differences

| Specification | NVIDIA GeForce GT 740 | NVIDIA GeForce 820A |

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

| Architecture | Kepler | Fermi 2.0 |

| Chip | GK107 | GF117 |

| Process Node | 28 nm | 28 nm |

| Transistors | 1,270 million | 585 million |

| Die Size | 118 mm² | 116 mm² |

| Transistor Density | 10.8M / mm² | 5.0M / mm² |

| Memory Clock | 1253 MHz (5 Gbps effective) | 1001 MHz (2 Gbps effective) |

| Memory Type | GDDR5 | DDR3 |

| Memory Bus Width | 128 bit | 64 bit |

| Memory Bandwidth | 80.19 GB/s | 16.02 GB/s |

| Shading Units | 384 | 96 |

| TMUs | 32 | 16 |

| ROPs | 16 | 8 |

| Pixel Rate | 7.944 GPixel/s | 2.500 GPixel/s |

| Texture Rate | 31.78 GTexel/s | 10.00 GTexel/s |

| FP32 Performance | 762.6 GFLOPS | 240.0 GFLOPS |

| TDP | 64 W | 15 W |

| Slot Width | Single-slot | IGP |

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

| Suggested PSU | 250 W | None |

| Bus Interface | PCIe 3.0 x16 | PCIe 2.0 x16 |

| Vulkan Support | 1.2.175 | None |

| Release Date | 2014-05-28 | 2014-03-16 |

| Launch MSRP | 89 USD | None |