Intel UHD Graphics 730 vs NVIDIA Quadro K620M Comparison

The NVIDIA Quadro K620M and Intel UHD Graphics 730 occupy nearly the same performance tier, but the data shows a narrow edge for the Intel part. In the only directly comparable benchmark, Geekbench OpenCL, the Intel UHD Graphics 730 scores 5988, while the Quadro K620M scores 5957. That is a delta of just 0.5% in favor of Intel. For practical purposes, the two are functionally equivalent in raw compute throughput.

Looking at the broader rival landscape reinforces this parity. The Quadro K620M’s average benchmark score of 5957 places it 0.2% behind the AMD Radeon HD 8750M (5970) and 0.4% behind the NVIDIA Quadro K4000 (5982). Its nearest rival, the AMD Radeon HD 8730M, scores 5955, which is a 0% delta. The Intel UHD Graphics 730, with an average score of 5929, sits 0.4% behind the HD 8730M, 0.5% behind the Quadro K620M, 0.7% behind the HD 8750M, and 0.9% behind the Quadro K4000. The entire cluster of five GPUs spans a range of roughly 1%, meaning these parts are statistically indistinguishable in most workloads.

One notable divergence appears in API support. The Intel part supports DirectX 12 (12_1), while the Quadro K620M is limited to DirectX 12 (11_0). This is a meaningful difference for modern titles that leverage tier-1 features. However, the Quadro K620M counters with a higher shading unit count: 384 versus 192 for Intel. That doubling of shading units does not translate into a compute win, as the FP32 figures show—863.2 GFLOPS for NVIDIA versus 499.2 GFLOPS for Intel. The Intel part’s higher boost clock of 1300 MHz versus 1124 MHz helps close the gap, but the raw throughput advantage still belongs to NVIDIA.

Pixel fill rate favors Intel at 10.40 GPixel/s versus 8.992 GPixel/s for the Quadro K620M. Texture fill rate, however, favors NVIDIA at 17.98 GTexel/s versus 15.60 GTexel/s. So the two split the fill-rate categories. The Intel part’s 8 ROPs match the Quadro’s 8 ROPs, but Intel only has 12 TMUs against NVIDIA’s 16.

FAQ

Q: Which GPU wins the Geekbench OpenCL benchmark?

A: The Intel UHD Graphics 730 wins with a score of 5988, edging out the NVIDIA Quadro K620M’s 5957 by 0.5%.

Q: How close are these two GPUs in average benchmark score?

A: The Quadro K620M averages 5957, while the Intel UHD Graphics 730 averages 5929. That is a 0.5% delta favoring NVIDIA in the aggregate.

Q: Does the Quadro K620M have more shading units?

A: Yes, the Quadro K620M has 384 shading units, exactly double the 192 found in the Intel UHD Graphics 730.

Q: Which GPU has a higher pixel fill rate?

A: The Intel UHD Graphics 730 reaches 10.40 GPixel/s, while the Quadro K620M delivers 8.992 GPixel/s.

Q: What DirectX version does each GPU support?

A: The Intel UHD Graphics 730 supports DirectX 12 (12_1), whereas the Quadro K620M supports DirectX 12 (11_0).

Q: Is there a texture fill rate difference?

A: Yes, the Quadro K620M produces 17.98 GTexel/s, which is higher than the Intel UHD Graphics 730’s 15.60 GTexel/s.

Where Each One Wins

The Intel UHD Graphics 730 takes the only head-to-head victory in the Geekbench OpenCL test. Its higher boost clock of 1300 MHz and pixel rate of 10.40 GPixel/s give it an edge in scenarios that scale with clock speed and ROP throughput. The DirectX 12 (12_1) support also makes it the more future-proof option for applications that rely on newer API features, even if the raw compute delta is negligible.

The Quadro K620M wins on shading unit count and FP32 throughput, posting 863.2 GFLOPS against Intel’s 499.2 GFLOPS. Its texture rate of 17.98 GTexel/s also exceeds the Intel part’s 15.60 GTexel/s. For workloads that are heavily shader-bound or texture-bound, the NVIDIA part should show a measurable advantage, despite the near-identical OpenCL scores. The dedicated 2 GB DDR3 memory with 16.02 GB/s bandwidth also provides a fixed resource, whereas the Intel UHD Graphics 730 relies on system shared memory, making its effective bandwidth system dependent.

In practical terms, the Intel part is better suited for lightweight desktop workloads where the 15 W TDP and integrated design keep power draw minimal. The Quadro K620M, with a 30 W TDP and MXM module form factor, targets mobile workstations that need sustained compute without relying on system memory. If the workload leverages DirectX 12_1 features, Intel wins. If the workload is shader-heavy and can use the dedicated memory, NVIDIA wins. For everything else, the 0.5% delta in OpenCL suggests users will not notice a difference.

Specification Differences

| Specification | NVIDIA Quadro K620M | Intel UHD Graphics 730 |

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

| Chip | GM108S | Rocket Lake |

| Architecture | Maxwell | Generation 12.1 |

| Process Node | 28 nm | 14 nm+++ |

| Foundry | TSMC | Intel |

| Base Clock | 1029 MHz | 300 MHz |

| Boost Clock | 1124 MHz | 1300 MHz |

| Memory Size | 2 GB | System Shared |

| Memory Type | DDR3 | System Shared |

| Memory Bus Width | 64 bit | System Shared |

| Memory Bandwidth | 16.02 GB/s | System Dependent |

| Shading Units | 384 | 192 |

| TMUs | 16 | 12 |

| ROPs | 8 | 8 |

| Pixel Rate | 8.992 GPixel/s | 10.40 GPixel/s |

| Texture Rate | 17.98 GTexel/s | 15.60 GTexel/s |

| FP32 | 863.2 GFLOPS | 499.2 GFLOPS |

| FP16 | Not listed | 998.4 GFLOPS (2:1) |

| TDP | 30 W | 15 W |

| Slot Width | MXM Module | IGP |

| Bus Interface | MXM-A (3.0) | Ring Bus |

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

| Transistors | 1,020 million | Not listed |

| Die Size | 77 mm² | Not listed |

Architecture Differences

The two GPUs come from fundamentally different design philosophies. The Quadro K620M is built on NVIDIA’s Maxwell architecture using a 28 nm process at TSMC, packing 1,020 million transistors into a 77 mm² die. That yields a transistor density of 13.2M per mm². The Intel UHD Graphics 730, by contrast, uses Intel’s Generation 12.1 architecture on a 14 nm+++ process, with no transistor or die size figures listed. The Intel part is an integrated graphics processor (IGP) connected via the Ring Bus, while the NVIDIA part is a discrete MXM-A (3.0) module.

The Quadro K620M’s Maxwell architecture is older, reflected in its DirectX 12 (11_0) support, but the dedicated 2 GB DDR3 memory gives it a fixed 64-bit bus and 16.02 GB/s of bandwidth. The Intel part has no dedicated memory; everything is system shared, making its bandwidth system dependent. The Intel architecture compensates with a higher boost clock and support for DirectX 12 (12_1), plus FP16 compute at 998.4 GFLOPS, a feature the Maxwell part does not list.

Shading and texture resources also differ sharply. The Quadro K620M has 384 shading units, 16 TMUs, and 8 ROPs. The Intel part has 192 shading units, 12 TMUs, and 8 ROPs. Despite half the shading units, Intel manages to nearly match the OpenCL score, likely due to its higher clock speed and architectural efficiency. The FP32 peak of 863.2 GFLOPS for NVIDIA versus 499.2 GFLOPS for Intel shows that raw ALU throughput does not directly translate to benchmark parity.

Power and form factor are another major split. The Quadro K620M consumes 30 W and uses an MXM module, indicating a replaceable, mobile workstation component. The Intel UHD Graphics 730 is an IGP with a 15 W TDP, embedded in the processor and dependent on the motherboard for display outputs. The NVIDIA part is end-of-life with a release date of 2015, while the Intel part is also end-of-life but released in 2021. The production status for both is listed as end-of-life, and neither has a launch MSRP.