AMD Radeon HD 8730M vs NVIDIA Quadro K620M Comparison

The NVIDIA Quadro K620M and AMD Radeon HD 8730M are two legacy mobile graphics solutions that end up in a statistical dead heat. Both sit at the 34th percentile of all GPUs, with an average benchmark score of 5957 and 5955, respectively. The data shows a single OpenCL benchmark where the K620M edges ahead by a razor-thin 0.03% margin, a difference that is functionally negligible. This page dissects where each chip holds a theoretical advantage based on architectural and specification differences, even when the raw performance numbers are nearly identical.

Where Each One Wins

The benchmark score alone does not differentiate these two parts. The real separation comes from the underlying hardware design. The NVIDIA Quadro K620M wins on raw compute throughput. Its peak FP32 performance is 863.2 GFLOPS, which is a substantial 60.5% higher than the AMD Radeon HD 8730M’s 537.6 GFLOPS. This suggests the K620M is better suited for workloads that are heavily arithmetic, such as scientific simulations, financial modeling, or any compute task that relies on shader math. The K620M also holds the clock speed advantage, with a base clock of 1029 MHz and a boost of 1124 MHz, compared to the AMD’s 650 MHz base and 700 MHz boost. This higher frequency directly drives its fillrate advantages, as the K620M achieves 8.992 GPixel/s and 17.98 GTexel/s, outpacing the AMD’s 5.600 GPixel/s and 16.80 GTexel/s. For pixel-heavy rendering or texture sampling, the NVIDIA part is the clear winner on paper.

The AMD Radeon HD 8730M, however, wins on memory bandwidth and bus width. It features a 128-bit memory bus, which is double the width of the K620M’s 64-bit bus. This results in a memory bandwidth of 28.80 GB/s, which is 79.8% higher than the K620M’s 16.02 GB/s. While the NVIDIA card has a higher memory clock (1001 MHz vs 900 MHz), the AMD’s wider bus is the dominant factor. This makes the HD 8730M theoretically better for large data sets that need to be streamed quickly, such as high-resolution texture loads or large framebuffer operations. Furthermore, the AMD card has 24 TMUs and 8 ROPs, while the NVIDIA has 16 TMUs and 8 ROPs. The extra TMUs on the AMD suggest it could handle multi-textured scenes with less bottleneck, even if its overall texture rate is slightly lower due to clock speeds.

The architecture generation is another key differentiator. The NVIDIA Quadro K620M is built on the Maxwell architecture, while the AMD uses GCN 1.0. The data shows the K620M is listed as part of the "Quadro Kepler-M (Kx200M)" generation, which is a naming quirk, but the chip itself is GM108S. The AMD is part of the "Solar System (HD 8700M)" generation. In terms of process node, both are 28 nm, but the NVIDIA chip packs 1,020 million transistors onto a 77 mm² die, while the AMD uses 950 million transistors on the same 77 mm² die. This gives the NVIDIA a higher transistor density of 13.2M / mm² versus 12.3M / mm², indicating a more complex or efficient design per square millimeter.

The Verdict

The data indicates that the choice between these two GPUs depends entirely on the workload, as the single benchmark result shows no meaningful winner. For users prioritizing raw compute throughput and pixel fillrate, the NVIDIA Quadro K620M is the superior choice. The 863.2 GFLOPS FP32 performance is a significant margin over the AMD, and the higher pixel rate of 8.992 GPixel/s means it can drive more pixels per second to the display. This makes it the stronger candidate for CAD, professional visualization, or any application where shading complexity and math-heavy shaders are the bottleneck.

For users dealing with memory-intensive tasks, the AMD Radeon HD 8730M is the better option. The 28.80 GB/s memory bandwidth, enabled by the 128-bit bus, is a massive advantage over the K620M’s 64-bit bus. While the K620M has a higher memory clock, the bus width difference is decisive. This makes the HD 8730M more suitable for workloads that involve large textures, high-resolution frame buffers, or data-parallel tasks that cannot fit into a narrow bus. The extra TMUs (24 vs 16) also give it a theoretical edge in texture-heavy scenarios.

The benchmark results show that these two are equals in the Geekbench OpenCL test, with scores of 5957 and 5955. Therefore, the verdict is not about speed, but about suitability. If the application is shader-bound, pick the NVIDIA. If the application is bandwidth-bound, pick the AMD. The production status for both is end-of-life, so these are legacy parts, but the architecture differences are still instructive for understanding their behavior.

Head-to-Head Benchmarks

The only head-to-head benchmark provided is Geekbench OpenCL. In this specific test, the NVIDIA Quadro K620M scores 5957, while the AMD Radeon HD 8730M scores 5955. The delta percentage is 0.0%, meaning the difference is so small it rounds to zero. This is a statistical tie, and it highlights how the theoretical advantages of each card do not translate into a meaningful OpenCL performance difference.

The nearest rivals for the NVIDIA Quadro K620M include the AMD Radeon HD 8730M with a delta of 0%, the AMD Radeon HD 8750M which is 0.2% faster, and the NVIDIA Quadro K4000 which is 0.4% faster. The Intel UHD Graphics 730 is 0.5% slower. For the AMD Radeon HD 8730M, its nearest rivals are the same group, with the NVIDIA Quadro K620M at 0% delta, the HD 8750M 0.3% faster, the Intel UHD 730 0.4% faster, and the Quadro K4000 0.5% faster. This clustering of scores within a 1% range suggests that all these GPUs are essentially equivalent in this OpenCL workload, and the performance hierarchy is not defined by the benchmark itself.

The biggest win for the NVIDIA is not in the benchmark, but in the raw FP32 throughput. The 863.2 GFLOPS versus the 537.6 GFLOPS is a 325.6 GFLOPS advantage, which is a 60.5% lead. This is where the NVIDIA design pulls ahead. The biggest win for the AMD is in memory bandwidth. The 28.80 GB/s versus 16.02 GB/s is a 12.78 GB/s advantage, which is a 79.8% lead. These are the starkest contrasts in the data, and they define the functional separation between the two parts.

FAQ

Q: Which GPU has a higher average benchmark score?

A: The NVIDIA Quadro K620M has an average benchmark score of 5957, which is slightly higher than the AMD Radeon HD 8730M’s 5955. The difference is 2 points, or 0.03%.

Q: Is the AMD Radeon HD 8730M better for memory-heavy tasks?

A: Yes, the data supports this. The AMD has a 128-bit memory bus and 28.80 GB/s bandwidth, while the NVIDIA has a 64-bit bus and 16.02 GB/s bandwidth. The AMD’s bandwidth is 79.8% higher.

Q: What is the FP32 compute difference between the two?

A: The NVIDIA Quadro K620M delivers 863.2 GFLOPS, while the AMD Radeon HD 8730M delivers 537.6 GFLOPS. The NVIDIA part has a 60.5% advantage in this metric.

Q: Are both GPUs on the same manufacturing process?

A: Yes, both the NVIDIA Quadro K620M and the AMD Radeon HD 8730M are fabricated on a 28 nm process at TSMC. The NVIDIA die size is 77 mm², and the AMD die size is also 77 mm².

Q: Which GPU has more texture mapping units?

A: The AMD Radeon HD 8730M has 24 TMUs, while the NVIDIA Quadro K620M has 16 TMUs. This is a 50% advantage for the AMD in TMU count.

Q: What is the memory type and size for both?

A: Both the NVIDIA Quadro K620M and the AMD Radeon HD 8730M come with 2 GB of DDR3 memory. The NVIDIA operates at 1001 MHz (2 Gbps effective), while the AMD operates at 900 MHz (1800 Mbps effective).

Architecture Differences

The NVIDIA Quadro K620M is based on the Maxwell architecture, specifically the GM108S chip. It is listed under the "Quadro Kepler-M (Kx200M)" generation, which is a curious naming anomaly in the data. The AMD Radeon HD 8730M uses the GCN 1.0 architecture with the "Mars" chip, belonging to the "Solar System (HD 8700M)" generation. These are fundamentally different design philosophies. Maxwell focuses on efficiency and high clocks, while GCN 1.0 was designed for compute-heavy parallelism.

The transistor counts differ, with the NVIDIA packing 1,020 million transistors and the AMD using 950 million. Both are on a 77 mm² die, but the NVIDIA has a higher transistor density of 13.2M / mm² versus 12.3M / mm². The shading units are identical at 384, but the ROPs are the same at 8. The core clock speeds are vastly different, with the NVIDIA boosting to 1124 MHz versus the AMD’s 700 MHz. The memory interfaces are the biggest architectural split: NVIDIA uses a 64-bit bus with 16.02 GB/s bandwidth, while AMD uses a 128-bit bus with 28.80 GB/s bandwidth. The bus interface also differs, with NVIDIA using MXM-A (3.0) and AMD using PCIe 3.0 x8.

Specification Differences

The table below highlights the key specification differences between the two GPUs.

| Specification | NVIDIA Quadro K620M | AMD Radeon HD 8730M |

| :--- | :--- | :--- |

| Chip | GM108S | Mars |

| Architecture | Maxwell | GCN 1.0 |

| Transistors | 1,020 million | 950 million |

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

| Base Clock | 1029 MHz | 650 MHz |

| Boost Clock | 1124 MHz | 700 MHz |

| Memory Clock | 1001 MHz (2 Gbps effective) | 900 MHz (1800 Mbps effective) |

| Memory Bus Width | 64 bit | 128 bit |

| Memory Bandwidth | 16.02 GB/s | 28.80 GB/s |

| TMUs | 16 | 24 |

| Pixel Rate | 8.992 GPixel/s | 5.600 GPixel/s |

| Texture Rate | 17.98 GTexel/s | 16.80 GTexel/s |

| FP32 | 863.2 GFLOPS | 537.6 GFLOPS |

| TDP | 30 W | null |

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

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

| Vulkan API | 1.4 | 1.2.170 |

| Release Date | 2015-02-28 | 2013-03-31 |