GPU Comparison

Head-to-Head Benchmarks

The only direct benchmark comparison available is Geekbench OpenCL, and it shows a narrow victory for the NVIDIA GeForce 920MX. The 920MX scores 4068 points against the Quadro K2000D’s 3919 points, a delta of 3.8%. That margin is modest in absolute terms, but it flips the expected hierarchy given the two cards’ positioning. The 920MX is a mobile-oriented part, while the K2000D is a professional workstation card, yet in raw compute throughput the consumer chip edges ahead.

Context from the nearest rivals reinforces this result. The 920MX’s average benchmark score is 3528, placing it at the 21st percentile of all GPUs. Its closest competitor by average score is the GeForce GT 545 at 3594, which trails by 1.8%, meaning the 920MX is slightly below that rival. Against the Quadro 2000M, the 920MX leads by 2.7%. The K2000D, meanwhile, averages 3919, sitting at the 23rd percentile. Its nearest rival, the Quadro 2000D, scores 3930, a 0.3% gap in favor of the rival. The K2000D also edges the Quadro 2000 (3898) by 0.5% and the GeForce GT 745M (3953) by 0.9%. So while the 920MX wins the head-to-head OpenCL test, the K2000D’s lone benchmark score is actually higher than the 920MX’s average across all tests, 3919 versus 3528. That discrepancy suggests the 920MX’s Vulkan result drags down its average, while the K2000D has no Vulkan score recorded.

Looking at the wins tally, the 920MX takes 1 win and the K2000D takes 0. But that single data point should be weighed carefully. The 3.8% delta is well within the noise of typical benchmark variance, and the K2000D’s average score indicates it is generally competitive with the 920MX’s best result. In short, the head-to-head is essentially a tie in practical terms, with a slight statistical edge to the 920MX.

Architecture Differences

The underlying silicon tells a clear generational story. The 920MX uses the GM108S chip built on the Maxwell architecture, manufactured by TSMC on a 28 nm process. It packs 1,020 million transistors into a 77 mm² die, yielding a transistor density of 13.2 million per square millimeter. The K2000D uses the GK107 chip on the older Kepler architecture, also 28 nm from TSMC, but with 1,270 million transistors on a larger 118 mm² die, for a density of 10.8 million per square millimeter. The Maxwell part is denser and more modern, while the Kepler part has more raw silicon area.

Compute resources differ significantly. The 920MX has 256 shading units, 24 texture mapping units, and 8 raster output pipelines. The K2000D has 384 shading units, 32 TMUs, and 16 ROPs. That is 50% more shaders, 33% more TMUs, and double the ROPs on the K2000D. Yet the 920MX still wins the OpenCL test, which points to architectural efficiency gains in Maxwell. The 920MX’s peak FP32 throughput is 508.4 GFLOPS, while the K2000D reaches 732.7 GFLOPS, a 44% advantage for the K2000D on paper. Pixel rate is nearly identical: 7.944 GPixel/s for the 920MX versus 7.632 GPixel/s for the K2000D. Texture rate favors the K2000D at 30.53 GTexel/s against 23.83 GTexel/s.

Memory architecture is where the gap becomes stark. The 920MX uses 2 GB of DDR3 on a 64-bit bus, with a memory clock of 900 MHz (1800 Mbps effective) and bandwidth of 14.40 GB/s. The K2000D uses 2 GB of GDDR5 on a 128-bit bus, with a memory clock of 1000 MHz (4 Gbps effective) and bandwidth of 64.00 GB/s. That is over 4.4 times the memory bandwidth on the K2000D. For memory-bound workloads, this is a decisive advantage, even if the Maxwell chip’s efficiency compensates in compute tests.

Power and physical differences are notable. The 920MX is rated at 16 W TDP and uses an MXM module slot width, with no power connectors. The K2000D draws 51 W, is single-slot, also has no power connectors, but recommends a 250 W power supply. The 920MX connects via PCIe 3.0 x8, while the K2000D uses PCIe 2.0 x16. Display outputs differ: the 920MX is “Portable Device Dependent,” meaning it relies on the host laptop’s outputs, while the K2000D offers 2x DVI and 1x mini-DisplayPort 1.2, with physical dimensions of 202 mm length and 111 mm height.

API support is similar but not identical. Both support DirectX 12 (11_0) and OpenGL 4.6. The 920MX supports Vulkan 1.4, while the K2000D supports Vulkan 1.2.175. Neither has ray tracing or tensor cores, as both predate those features. The K2000D’s launch MSRP was 599 USD; the 920MX has no listed MSRP.

The Verdict

The data points to a nuanced conclusion. If the sole criterion is the Geekbench OpenCL score, the GeForce 920MX wins by 3.8%. That is a real but slim margin. However, the K2000D’s average benchmark score of 3919 is higher than the 920MX’s average of 3528, and the K2000D sits at the 23rd percentile versus the 920MX’s 21st. On average performance across all recorded tests, the K2000D is the better card.

For compute tasks that stress raw FP32 throughput and memory bandwidth, the K2000D’s specifications are superior: 732.7 GFLOPS versus 508.4 GFLOPS, and 64.00 GB/s versus 14.40 GB/s. The 50% more shading units and double the ROPs also favor the K2000D for geometry-heavy or bandwidth-intensive workloads. Yet the 920MX’s Maxwell architecture delivers better efficiency per transistor, as evidenced by its OpenCL win despite lower theoretical specs.

The practical verdict: the Quadro K2000D is the stronger workstation card for professional applications where memory bandwidth and shader count matter, and its 23rd percentile ranking supports that. The GeForce 920MX is a competent mobile chip for light compute, but its 21st percentile and lower average score indicate it is the weaker overall performer. The K2000D also offers dedicated display outputs and a fixed physical form factor, making it a drop-in desktop solution, whereas the 920MX is dependent on the host device.

Specification Differences

| Specification | GeForce 920MX | Quadro K2000D |

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

| Architecture | Maxwell | Kepler |

| Chip | GM108S | GK107 |

| Transistors | 1,020 million | 1,270 million |

| Die Size | 77 mm² | 118 mm² |

| Transistor Density | 13.2M / mm² | 10.8M / mm² |

| Base Clock | 965 MHz | Not listed |

| Boost Clock | 993 MHz | Not listed |

| Memory Clock | 900 MHz / 1800 Mbps effective | 1000 MHz / 4 Gbps effective |

| Memory Type | DDR3 | GDDR5 |

| Memory Bus Width | 64 bit | 128 bit |

| Memory Bandwidth | 14.40 GB/s | 64.00 GB/s |

| Shading Units | 256 | 384 |

| TMUs | 24 | 32 |

| ROPs | 8 | 16 |

| Pixel Rate | 7.944 GPixel/s | 7.632 GPixel/s |

| Texture Rate | 23.83 GTexel/s | 30.53 GTexel/s |

| FP32 Compute | 508.4 GFLOPS | 732.7 GFLOPS |

| TDP | 16 W | 51 W |

| Slot Width | MXM Module | Single-slot |

| Suggested PSU | Not listed | 250 W |

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

| Display Outputs | Portable Device Dependent | 2x DVI, 1x mini-DisplayPort 1.2 |

| Vulkan API | 1.4 | 1.2.175 |

| Length | Not listed | 202 mm (8 inches) |

| Height | Not listed | 111 mm (4.4 inches) |

| Release Date | 2016-03-24 | 2013-02-28 |

| Predecessor | GeForce 800M | Quadro Fermi |

| Successor | GeForce 10 Mobile | Quadro Maxwell |

| Launch MSRP | Not listed | 599 USD |

FAQ

Q: Which GPU has higher memory bandwidth?

A: The Quadro K2000D has 64.00 GB/s bandwidth, which is over 4.4 times the 14.40 GB/s of the GeForce 920MX.

Q: What is the difference in shading units?

A: The K2000D has 384 shading units, while the 920MX has 256, a 50% advantage for the K2000D.

Q: Which card won the Geekbench OpenCL test?

A: The GeForce 920MX scored 4068 against the K2000D’s 3919, a delta of 3.8% in favor of the 920MX.

Q: How do their average benchmark scores compare?

A: The K2000D averages 3919 across its recorded tests, while the 920MX averages 3528. The K2000D sits at the 23rd percentile of all GPUs, versus the 920MX’s 21st.

Q: Do both cards support the same DirectX version?

A: Yes, both support DirectX 12 (11_0) and OpenGL 4.6. Vulkan support differs: the 920MX supports 1.4, while the K2000D supports 1.2.175.

Q: What is the TDP difference?

A: The 920MX is rated at 16 W, while the K2000D is rated at 51 W. The K2000D also recommends a 250 W power supply, while the 920MX has no such recommendation.

Where Each One Wins

The GeForce 920MX wins in the Geekbench OpenCL benchmark, taking the only direct head-to-head test by 3.8%. It also has a higher transistor density (13.2M / mm² versus 10.8M / mm²) and a more modern architecture (Maxwell versus Kepler). Its 16 W TDP makes it far more power-efficient, and its PCIe 3.0 x8 interface is newer than the K2000D’s PCIe 2.0 x16. For users constrained by power budgets or needing a mobile MXM module, the 920MX is the logical choice. Its Vulkan 1.4 support also offers broader compatibility for modern compute APIs.

The Quadro K2000D wins on nearly every raw specification that matters for sustained professional workloads. It has 384 shading units versus 256, 32 TMUs versus 24, and 16 ROPs versus 8. Its FP32 compute of 732.7 GFLOPS is 44% higher than the 920MX’s 508.4 GFLOPS. Memory bandwidth of 64.00 GB/s versus 14.40 GB/s is a massive advantage for large datasets. It also has dedicated display outputs (2x DVI, 1x mini-DisplayPort) and a fixed single-slot form factor with defined dimensions, making it a predictable desktop component. Its average benchmark score of 3919 exceeds the 920MX’s 3528, and its 23rd percentile ranking is higher than the 920MX’s 21st.

For use-case selection: choose the 920MX for low-power mobile applications, light OpenCL compute, or scenarios where the host device provides display output. Choose the K2000D for workstation environments where memory bandwidth, shader throughput, and multi-display output are critical, and where the 51 W TDP and 250 W PSU recommendation are acceptable. The K2000D’s 599 USD launch MSRP also positions it as a professional-grade product, though pricing should not be the deciding factor based on the data alone. The benchmark data shows a near-tie in OpenCL, but the K2000D’s superior specifications and higher average score make it the more capable all-round performer for demanding tasks.