GPU Comparison

NVIDIA’s GeForce 920M and GeForce 820A are both end-of-life mobile graphics solutions, but they represent different generations of the company’s design philosophy. The 920M is a Kepler 2.0 part aimed at the GeForce 900M series, while the 820A is a Fermi 2.0 chip from the GeForce 800A line. Benchmark results show a clear, though not overwhelming, performance gap between the two, with the 920M taking the only head-to-head victory.

Head-to-Head Benchmarks

The sole direct comparison available is the Geekbench OpenCL test, and the results are unambiguous. The GeForce 920M scores 3,725 points, while the GeForce 820A manages 2,983 points. This gives the 920M a decisive 24.9% advantage in this compute-oriented workload. That is a substantial margin, nearly a quarter more performance, which aligns with the 920M’s newer architecture and higher specifications.

Looking at the broader context, the 920M’s average benchmark score of 3,287 places it in the 20th percentile of all GPUs. Its nearest rivals include the GeForce GT 730M (3,316, -0.9%), Intel HD Graphics 530 (3,332, -1.4%), and the GeForce GT 640 (3,210, +2.4%). This means the 920M sits in a tight cluster where small changes in workload can flip the ranking. It is nearly tied with the GT 730M, slightly behind the HD 530, and slightly ahead of the GT 640 in average terms.

The 820A, by contrast, has an average score of 2,983, placing it in the 19th percentile. Its nearest rivals are surprisingly high-end parts: the GeForce GTX 860M (2,967, +0.5%), GeForce GTX 750 Ti (2,953, +1%), Quadro P600 (2,923, +2.1%), and even the GeForce RTX 4060 Ti 8 GB (2,913, +2.4%). This is an odd grouping, the 820A’s raw score is consistently a few points higher than these much more powerful desktop and mobile GPUs in this specific test, but the deltas are tiny (under 2.5%). The practical takeaway is that the 820A’s OpenCL result is an outlier relative to its class; it performs roughly on par with those high-end parts in this single benchmark, but its overall capability is far more limited.

The data shows one clear winner in the head-to-head: the 920M. The 24.9% delta in OpenCL is the only direct comparison, and it favors the newer part without ambiguity.

Where Each One Wins

The 920M wins the only benchmark where both are present. In Geekbench OpenCL, its 3,725 points versus 2,983 points represents a 24.9% lead. This suggests the 920M is the better choice for general-purpose compute tasks that leverage OpenCL, such as video encoding acceleration or certain physics simulations. Its higher average benchmark score (3,287 vs. 2,983) reinforces that it is the stronger overall performer in the aggregate.

The 820A has no benchmark victories in the data. There is no test where it outperforms the 920M. However, its lower TDP of 15 W versus 33 W makes it a more power-efficient part on paper. For a laptop where battery life is paramount and the GPU is only used occasionally, the 820A’s lower power draw could be a deciding factor, even if it loses on raw performance. The 820A also supports a PCIe 2.0 x16 interface, which is older but wider than the 920M’s PCIe 3.0 x8 link; for low-bandwidth tasks this difference is unlikely to matter, but it is a technical distinction.

In practical terms, the 920M is the pick for anyone who needs the extra compute headroom. The 820A is for scenarios where power consumption is the primary constraint and the workload is light enough that the 24.9% performance gap is irrelevant.

Architecture Differences

The two GPUs are built on different architectures, which explains much of the performance gap. The 920M uses Kepler 2.0, the same design found in NVIDIA’s 900M series, while the 820A is based on the older Fermi 2.0 architecture from the 800A series. Both are fabricated on a 28 nm process at TSMC, but that is where the similarities end.

The 920M’s chip, codenamed GK208B, packs 1,020 million transistors on an 87 mm² die, yielding a transistor density of 11.7 million per mm². The 820A’s GF117 chip is physically larger at 116 mm² but contains only 585 million transistors, resulting in a much lower density of 5.0 million per mm². This is a clear sign of architectural inefficiency: the Fermi design uses more silicon area for fewer components.

The compute resources differ sharply. The 920M has 384 shading units, 32 texture mapping units (TMUs), and 8 raster operation units (ROPs). The 820A has just 96 shaders, 16 TMUs, and 8 ROPs. That is a 4x advantage in shader count for the 920M, and a 2x advantage in TMUs, with equal ROP counts. The pixel rate reflects this: the 920M outputs 7.632 GPixel/s versus 2.500 GPixel/s for the 820A. Texture rate is 30.53 GTexel/s versus 10.00 GTexel/s.

The 920M also has a massive lead in floating-point performance. Its FP32 throughput is 732.7 GFLOPS, more than three times the 820A’s 240.0 GFLOPS. Neither part supports FP16, so that field is null for both.

Memory is another differentiator. The 920M comes with 2 GB of DDR3 on a 64-bit bus, running at 900 MHz (1,800 Mbps effective), yielding 14.40 GB/s of bandwidth. The 820A has 1,024 MB (1 GB) of DDR3 on the same 64-bit bus, but at a higher clock of 1,001 MHz (2 Gbps effective), giving it 16.02 GB/s of bandwidth. Despite the higher memory clock, the 820A’s smaller capacity and lower shader count mean it cannot exploit that bandwidth as effectively.

Feature support also diverges. The 920M supports Vulkan 1.2.175, while the 820A lists no Vulkan support. Both support DirectX 12 (11_0) and OpenGL 4.6. The 920M’s bus interface is PCIe 3.0 x8, whereas the 820A uses PCIe 2.0 x16. The 920M’s TDP is 33 W, more than double the 820A’s 15 W.

FAQ

Q: Which GPU is faster in the Geekbench OpenCL test?

A: The GeForce 920M scores 3,725 points versus the 820A’s 2,983 points, a 24.9% advantage for the 920M.

Q: How do the two compare in terms of memory capacity?

A: The 920M has 2 GB of DDR3, while the 820A has 1,024 MB (1 GB). Both use a 64-bit bus, but the 820A’s memory runs at a higher effective speed of 2 Gbps versus 1,800 Mbps on the 920M, giving the 820A slightly higher bandwidth (16.02 GB/s vs. 14.40 GB/s).

Q: Does the 820A support Vulkan?

A: No. The 820A has no Vulkan support listed, while the 920M supports Vulkan 1.2.175.

Q: What is the power draw difference?

A: The 920M has a TDP of 33 W, while the 820A is rated at 15 W. The 820A is significantly more power-efficient.

Q: Which GPU has more shading units?

A: The 920M has 384 shading units, compared to 96 on the 820A, a 4x difference.

Q: Are both GPUs based on the same architecture?

A: No. The 920M uses Kepler 2.0 (chip GK208B), while the 820A uses Fermi 2.0 (chip GF117). Both are built on a 28 nm TSMC process.

Specification Differences

| Specification | NVIDIA GeForce 920M | NVIDIA GeForce 820A |

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

| Architecture | Kepler 2.0 | Fermi 2.0 |

| Chip | GK208B | GF117 |

| Process Node | 28 nm | 28 nm |

| Transistors | 1,020 million | 585 million |

| Die Size | 87 mm² | 116 mm² |

| Transistor Density | 11.7M / mm² | 5.0M / mm² |

| Base Clock | 954 MHz | Not specified |

| Boost Clock | 954 MHz | Not specified |

| Memory Clock | 1,800 Mbps effective | 2 Gbps effective |

| Memory Size | 2 GB | 1,024 MB |

| Memory Type | DDR3 | DDR3 |

| Memory Bus | 64 bit | 64 bit |

| Memory Bandwidth | 14.40 GB/s | 16.02 GB/s |

| Shading Units | 384 | 96 |

| TMUs | 32 | 16 |

| ROPs | 8 | 8 |

| Pixel Rate | 7.632 GPixel/s | 2.500 GPixel/s |

| Texture Rate | 30.53 GTexel/s | 10.00 GTexel/s |

| FP32 Performance | 732.7 GFLOPS | 240.0 GFLOPS |

| TDP | 33 W | 15 W |

| Bus Interface | PCIe 3.0 x8 | PCIe 2.0 x16 |

| Vulkan API | 1.2.175 | Not specified |

| OpenCL Benchmark | 3,725 | 2,983 |

| Average Benchmark | 3,287 | 2,983 |

| Percentile (All GPUs) | 20th | 19th |

| Release Date | 2015-03-12 | 2014-03-16 |

The Verdict

Based strictly on the data, the GeForce 920M is the superior performer. It wins the only head-to-head benchmark by 24.9%, has a higher average score (3,287 vs. 2,983), and offers a 4x advantage in shading units and a 3x advantage in FP32 throughput. Its Kepler 2.0 architecture is more efficient, packing more transistors into a smaller die while delivering more than double the pixel and texture rates. If you need a GPU for OpenCL compute tasks or any graphics workload that scales with shader count, the 920M is the clear choice.

The 820A’s only advantages are its lower TDP (15 W vs. 33 W) and its slightly higher memory bandwidth (16.02 GB/s vs. 14.40 GB/s). The higher bandwidth is irrelevant given its weaker compute core; the lower power draw is meaningful only in ultra-portable or battery-sensitive designs where the 24.9% performance loss is an acceptable trade-off for longer runtime. Its PCIe 2.0 x16 interface is older but wider, which may not matter for the light workloads it can handle.

For a builder or buyer choosing between these two end-of-life parts, the decision hinges on power versus performance. The 920M is the practical recommendation for almost any real workload, it is faster, supports Vulkan, and has double the memory. The 820A is only preferable in a system where the 33 W TDP of the 920M is unacceptable and the tasks are trivial enough that the 820A’s 96 shaders suffice. The data favors the 920M without qualification; the 820A is a fallback for power-constrained niche cases.