AMD Radeon HD 8790M vs NVIDIA GeForce 840M Comparison

The AMD Radeon HD 8790M and NVIDIA GeForce 840M are both end-of-life mobile graphics solutions from the 28 nm era, but they represent fundamentally different architectural approaches. The data shows a complete split in benchmark outcomes, with each GPU winning one of the two available tests by a decisive margin. The AMD part, built on GCN 1.0, takes the Vulkan test with a 30.4% advantage, while the NVIDIA Maxwell part dominates the OpenCL test with a 13% lead. This creates an unusual situation where the average benchmark scores (5691 for AMD, 5322 for NVIDIA) do not reliably predict performance in any given workload, making the architectural differences the primary determinant of which GPU is faster.

Architecture Differences

The AMD Radeon HD 8790M is built on GCN 1.0 architecture using the Mars chip, while the NVIDIA GeForce 840M uses Maxwell architecture with the GM108S chip. Both are manufactured by TSMC on a 28 nm process, and both have the same die size of 77 mm², though NVIDIA packs slightly more transistors into that area: 1,020 million versus AMD's 950 million. This translates to a higher transistor density of 13.2M / mm² for NVIDIA versus 12.3M / mm² for AMD.

The memory subsystems diverge sharply. The AMD Radeon HD 8790M uses 2 GB of GDDR5 memory on a 128-bit bus, delivering 64.00 GB/s of bandwidth. The NVIDIA GeForce 840M also has 2 GB, but uses DDR3 memory on a 64-bit bus, yielding only 16.02 GB/s — exactly one quarter of the AMD's bandwidth. This is the single largest architectural gap between the two, and it directly explains the Vulkan result where AMD wins by 30.4%.

Shader configuration is identical in terms of shading units, with both GPUs featuring 384 shading units. However, the texture mapping units differ: AMD has 24 TMUs versus NVIDIA's 16. Both have 8 ROPs. Clock speeds favor NVIDIA significantly, with a base clock of 1029 MHz and boost of 1124 MHz, compared to AMD's 850 MHz base and 900 MHz boost. Despite this clock advantage, AMD's higher TMU count and wider memory bus allow it to achieve a texture rate of 21.60 GTexel/s versus NVIDIA's 17.98 GTexel/s. NVIDIA does lead in pixel rate at 8.992 GPixel/s versus 7.200 GPixel/s, and in FP32 compute at 863.2 GFLOPS versus 691.2 GFLOPS.

API support shows subtle differences. Both support DirectX 12, but AMD lists version 12 (11_1) while NVIDIA lists 12 (11_0). OpenGL 4.6 is present on both. Vulkan support favors NVIDIA with version 1.4 versus AMD's 1.2.170. The NVIDIA GPU also carries a 33 W TDP, while the AMD part's TDP is not listed in the data.

FAQ

Q: Which GPU has higher raw compute performance in FP32?

A: The NVIDIA GeForce 840M leads in FP32 with 863.2 GFLOPS, which is 24.9% higher than the AMD Radeon HD 8790M's 691.2 GFLOPS.

Q: How do the memory bandwidth figures compare?

A: The AMD Radeon HD 8790M has a 128-bit bus with GDDR5 memory providing 64.00 GB/s, while the NVIDIA GeForce 840M has only a 64-bit bus with DDR3 memory providing 16.02 GB/s — a fourfold difference in bandwidth.

Q: Which GPU wins the Vulkan benchmark, and by how much?

A: The AMD Radeon HD 8790M wins the geekbench_vulkan test with a score of 6365 versus NVIDIA's 4880, a delta of 30.4% in AMD's favor.

Q: What is the NVIDIA GeForce 840M's TDP?

A: The NVIDIA GeForce 840M has a listed TDP of 33 W. The AMD Radeon HD 8790M does not have a TDP value in the data.

Q: How do the two GPUs rank against all other GPUs?

A: The AMD Radeon HD 8790M sits at the 33rd percentile, while the NVIDIA GeForce 840M sits at the 31st percentile. Both are near the lower end of the performance spectrum.

Q: What are the nearest rivals for each GPU?

A: For AMD, the nearest rivals are the Intel Iris Pro Graphics P6300 (5712 avg, -0.4% delta), NVIDIA GeForce GTX 670MX (5721, -0.5%), NVIDIA GeForce GTX 550 Ti (5731, -0.7%), and NVIDIA Quadro M500M (5604, 1.6%). For NVIDIA, the nearest rivals are the NVIDIA GeForce 930A (5317, 0.1%), NVIDIA GeForce GTX 980M (5308, 0.3%), AMD Radeon R7 M445 (5358, -0.7%), and NVIDIA GeForce 940M (5284, 0.7%).

Where Each One Wins

The AMD Radeon HD 8790M is the clear winner in scenarios that benefit from memory bandwidth and texture throughput. Its 64.00 GB/s bandwidth — four times that of the NVIDIA — makes it the stronger choice for workloads that stream large amounts of texture data or require frequent memory access. The 24 TMUs versus NVIDIA's 16 also give AMD a 20.1% advantage in texture fill rate. The Vulkan benchmark result of 6365 versus 4880 demonstrates that AMD's architecture handles this modern API significantly better, likely due to its GCN 1.0 design's lower-level hardware scheduling. For users running Vulkan-based applications or games, the AMD part is the superior option by a substantial margin.

The NVIDIA GeForce 840M wins in compute-heavy workloads that rely on raw shader throughput. Its FP32 output of 863.2 GFLOPS exceeds AMD's by 24.9%, and its pixel rate of 8.992 GPixel/s is 24.9% higher than AMD's 7.200 GPixel/s. The higher clock speeds (1029 MHz base, 1124 MHz boost) give NVIDIA an advantage in latency-sensitive tasks where memory bandwidth is not the bottleneck. The OpenCL benchmark result of 5764 versus 5017 confirms that NVIDIA's Maxwell architecture extracts better performance from OpenCL workloads. Additionally, NVIDIA lists a 33 W TDP, which may indicate better power efficiency per unit of compute, though AMD's TDP is not available for direct comparison.

Head-to-Head Benchmarks

The two GPUs split the benchmark results exactly one win each, making this a genuinely balanced comparison. In the geekbench_opencl test, the NVIDIA GeForce 840M scores 5764 against AMD's 5017, a delta of -13% from AMD's perspective. This means NVIDIA outperforms AMD by 13% in OpenCL. The margin is substantial and reflects NVIDIA's higher clock speeds and FP32 compute capabilities. In practical terms, this suggests that OpenCL-based applications such as video encoding, physics simulation, or general-purpose GPU compute will run noticeably faster on the NVIDIA part.

The geekbench_vulkan test tells the opposite story. Here, the AMD Radeon HD 8790M scores 6365 while the NVIDIA GeForce 840M manages only 4880, giving AMD a commanding 30.4% advantage. This is a much larger margin than NVIDIA's OpenCL lead, indicating that AMD's GCN architecture has a significant edge in Vulkan workloads. The memory bandwidth difference likely plays a decisive role, as Vulkan's low-level nature allows developers to explicitly manage memory and take full advantage of the 64.00 GB/s available on AMD versus the 16.02 GB/s on NVIDIA. The average benchmark scores reflect this split: AMD's average is 5691, which is 6.9% higher than NVIDIA's 5322, though this average masks the divergent performance profile.

Specification Differences

| Specification | AMD Radeon HD 8790M | NVIDIA GeForce 840M |

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

| Architecture | GCN 1.0 | Maxwell |

| Chip | Mars | GM108S |

| Transistors | 950 million | 1,020 million |

| Transistor Density | 12.3M / mm² | 13.2M / mm² |

| Base Clock | 850 MHz | 1029 MHz |

| Boost Clock | 900 MHz | 1124 MHz |

| Memory Type | GDDR5 | DDR3 |

| Memory Bus Width | 128 bit | 64 bit |

| Memory Bandwidth | 64.00 GB/s | 16.02 GB/s |

| Memory Clock | 1000 MHz / 4 Gbps effective | 1001 MHz / 2 Gbps effective |

| TMUs | 24 | 16 |

| Pixel Rate | 7.200 GPixel/s | 8.992 GPixel/s |

| Texture Rate | 21.60 GTexel/s | 17.98 GTexel/s |

| FP32 | 691.2 GFLOPS | 863.2 GFLOPS |

| TDP | Not listed | 33 W |

| Slot Width | MXM Module | IGP |

| Bus Interface | MXM-A (3.0) | PCIe 3.0 x8 |

| Vulkan Version | 1.2.170 | 1.4 |

| DirectX Version | 12 (11_1) | 12 (11_0) |

| Release Date | 2013-03-31 | 2014-03-11 |

| Predecessor | London | GeForce 700M |

| Successor | Gem System | GeForce 900M |

| Percentile | 33 | 31 |

| Average Benchmark Score | 5691 | 5322 |