GPU Comparison

The benchmark data positions the NVIDIA GeForce 920MX as the clear performance leader over the NVIDIA GeForce GT 635M, despite both cards occupying a similar low tier in the overall GPU landscape. The 920MX wins the only direct head-to-head comparison, though its victory is narrower than its architectural advantages might suggest. Both cards are end-of-life products aimed at the entry-level mobile segment, but they represent different generations of NVIDIA's design philosophy, with the 920MX delivering superior compute and throughput metrics across the board.

Head-to-Head Benchmarks

The sole direct benchmark comparison available is the Geekbench OpenCL test, and the results show the NVIDIA GeForce 920MX outperforming the NVIDIA GeForce GT 635M with a score of 4068 against 3740. This represents an 8.1% advantage for the 920MX, a meaningful gap in compute performance but not a generational leap. The deltaPct of -8.1% in the head-to-head data confirms the 920MX as the winner, though the margin is modest enough that real-world differences in gaming or general GPU-accelerated tasks would be noticeable but not transformative.

Context from the nearest rivals puts this win into perspective. The GT 635M's score of 3740 places it just 0.6% ahead of the NVIDIA Quadro 3000M (3718) and the NVIDIA GeForce GT 740M (3717), while sitting 1.2% above the NVIDIA GeForce 825M (3694). On the other side, the 920MX's OpenCL result of 4068 is 1.8% above the NVIDIA GeForce GT 545 (3594) and 1.9% above the NVIDIA RTX 5000 Mobile Ada Generation (3596), though it trails the Intel UHD Graphics 710 (3792) by 1.4% in the GT 635M's rival list. The 920MX also has a Vulkan score of 2987, a test the GT 635M does not have data for, indicating broader API support that could matter for newer applications.

Average benchmark scores complicate the picture slightly. The GT 635M has an average score of 3740, matching its single OpenCL result, while the 920MX averages 3528 across its two tests, dragged down by its lower Vulkan score of 2987. This means the 920MX's OpenCL superiority is counterbalanced by weaker Vulkan performance, though the OpenCL test remains the more directly comparable metric between the two cards. The percentile rankings are close: the GT 635M sits at the 22nd percentile of all GPUs, while the 920MX sits at the 21st, indicating both are firmly entry-level parts with similar overall standing.

Where Each One Wins

The NVIDIA GeForce 920MX wins decisively in raw compute throughput. Its FP32 performance of 508.4 GFLOPS is over 2.7 times the GT 635M's 182.4 GFLOPS, a massive advantage for any workload that relies on shader or compute units. Texture rate follows a similar pattern, with the 920MX delivering 23.83 GTexel/s versus 7.600 GTexel/s for the GT 635M, a 3.1x improvement. Pixel rate also favors the 920MX at 7.944 GPixel/s compared to 1.900 GPixel/s, a 4.2x gap that directly impacts fill-rate-bound scenarios like high-resolution rendering or heavy post-processing effects.

The 920MX also wins on efficiency and connectivity. Its TDP of 16 W is less than half the GT 635M's 35 W, meaning the newer card delivers substantially higher performance while drawing significantly less power, a critical factor for thin-and-light laptops. The 920MX also uses a PCIe 3.0 x8 bus interface compared to the GT 635M's PCIe 2.0 x16, offering newer protocol support even with fewer lanes. Vulkan 1.4 API support is exclusive to the 920MX, while the GT 635M has no Vulkan listing, giving the newer card access to modern graphics APIs that can improve performance in recent titles.

The NVIDIA GeForce GT 635M retains one notable advantage: memory bandwidth. Its 128-bit bus width delivers 28.80 GB/s of bandwidth, exactly double the 920MX's 64-bit bus and 14.40 GB/s. In memory-intensive workloads where bandwidth is the bottleneck, the GT 635M could potentially close the gap or even edge ahead, though its much lower compute and texture rates would likely still hold it back overall. The GT 635M also has a larger process node at 40 nm versus 28 nm, which is not a performance advantage but indicates older manufacturing technology.

Architecture Differences

The two cards come from entirely different NVIDIA architectures. The GT 635M is built on the Fermi architecture with the GF108 chip, while the 920MX uses the Maxwell architecture with the GM108S chip. This generational shift is immediately apparent in the specs: the 920MX packs 256 shading units, 24 texture mapping units, and 8 ROPs, versus 96 shading units, 16 TMUs, and 4 ROPs on the GT 635M. The 920MX has nearly three times the shader count and double the ROPs, which explains its commanding lead in pixel and texture rates despite its narrower memory bus.

Process technology differs significantly. The GT 635M is fabricated on a 40 nm process by TSMC, while the 920MX uses a 28 nm process, also from TSMC. This node shrink allows the 920MX to pack 1,020 million transistors into a die size of just 77 mm², compared to the GT 635M's 585 million transistors on a 116 mm² die. The transistor density tells the story: the 920MX achieves 13.2 million transistors per square millimeter versus only 5.0 million on the GT 635M, a 2.6x density improvement that enables both higher performance and lower power consumption.

Memory configuration differs in bus width but not capacity. Both cards ship with 2 GB of DDR3 memory, but the GT 635M uses a 128-bit interface while the 920MX is limited to 64-bit. This is why the GT 635M's bandwidth of 28.80 GB/s doubles the 920MX's 14.40 GB/s, despite identical memory clocks of 900 MHz with 1800 Mbps effective speed. The 920MX compensates with higher core clocks, running at 965 MHz base and 993 MHz boost, while the GT 635M has no listed base or boost clock, only its memory speed.

Feature support also diverges. Both cards support DirectX 12 (11_0) and OpenGL 4.6, but the 920MX adds Vulkan 1.4 support, which the GT 635M lacks entirely. The 920MX also lists a boost clock, a feature absent from the GT 635M's specifications. Bus interface differences are notable: the 920MX uses PCIe 3.0 x8 while the GT 635M relies on PCIe 2.0 x16, with the 920MX's newer standard offering potentially better bandwidth efficiency despite fewer lanes.

FAQ

Q: Which GPU has the higher raw compute performance?

A: The NVIDIA GeForce 920MX, with FP32 performance of 508.4 GFLOPS versus 182.4 GFLOPS for the GT 635M, a 2.7x advantage. This is reflected in its Geekbench OpenCL score of 4068 versus 3740, an 8.1% lead.

Q: Does the GT 635M have any advantage over the 920MX?

A: Yes, in memory bandwidth. The GT 635M's 128-bit bus delivers 28.80 GB/s, exactly double the 920MX's 14.40 GB/s from its 64-bit bus. This could benefit bandwidth-bound workloads, though the 920MX's higher compute rates likely offset this in most scenarios.

Q: How do their power requirements compare?

A: The 920MX is far more power-efficient, with a TDP of 16 W versus 35 W for the GT 635M. This makes the 920MX better suited for portable devices where battery life and thermals are primary concerns.

Q: Which card supports more modern graphics APIs?

A: The 920MX supports Vulkan 1.4 in addition to DirectX 12 (11_0) and OpenGL 4.6. The GT 635M only lists DirectX 12 (11_0) and OpenGL 4.6, with no Vulkan support specified.

Q: What are the core configuration differences?

A: The 920MX has 256 shading units, 24 TMUs, and 8 ROPs, while the GT 635M has 96 shading units, 16 TMUs, and 4 ROPs. The 920MX also features a boost clock of 993 MHz, which the GT 635M lacks.

Q: How do the two cards rank overall among all GPUs?

A: They are nearly identical in percentile ranking, with the GT 635M at the 22nd percentile and the 920MX at the 21st percentile of all GPUs. This indicates both are entry-level parts with similar overall standing, despite the 920MX's per-test superiority.

Specification Differences

| Specification | NVIDIA GeForce GT 635M | NVIDIA GeForce 920MX |

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

| Architecture | Fermi | Maxwell |

| Chip | GF108 | GM108S |

| Process Node | 40 nm | 28 nm |

| Transistors | 585 million | 1,020 million |

| Die Size | 116 mm² | 77 mm² |

| Transistor Density | 5.0M / mm² | 13.2M / mm² |

| Base Clock | N/A | 965 MHz |

| Boost Clock | N/A | 993 MHz |

| Memory Bus Width | 128 bit | 64 bit |

| Memory Bandwidth | 28.80 GB/s | 14.40 GB/s |

| Shading Units | 96 | 256 |

| TMUs | 16 | 24 |

| ROPs | 4 | 8 |

| Pixel Rate | 1.900 GPixel/s | 7.944 GPixel/s |

| Texture Rate | 7.600 GTexel/s | 23.83 GTexel/s |

| FP32 Performance | 182.4 GFLOPS | 508.4 GFLOPS |

| TDP | 35 W | 16 W |

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

| Vulkan Support | N/A | 1.4 |

| Generation | GeForce 600M | GeForce 900M |

| Release Date | 2012-03-21 | 2016-03-24 |

The Verdict

The NVIDIA GeForce 920MX is the better GPU in nearly every measurable category. Its Geekbench OpenCL score of 4068 beats the GT 635M's 3740 by 8.1%, and its FP32 performance of 508.4 GFLOPS is nearly three times higher. The 920MX also offers dramatically better texture and pixel rates, lower power consumption at 16 W versus 35 W, and modern Vulkan API support. For any user choosing between these two for a laptop, the 920MX is the clear recommendation based on performance data alone.

The GT 635M's only significant advantage is memory bandwidth at 28.80 GB/s versus 14.40 GB/s, which could matter in specific bandwidth-intensive applications. However, this benefit is unlikely to overcome the 920MX's massive leads in shader count, texture rate, and pixel throughput. The GT 635M might be preferable in niche scenarios where memory bandwidth is the absolute bottleneck, but such cases are rare given the 920MX's overall compute superiority.

The percentile rankings tell a nuanced final story. Both cards sit at the bottom of the GPU hierarchy, with the GT 635M at the 22nd percentile and the 920MX at the 21st percentile. This means neither card is suitable for demanding modern gaming or professional workloads, and the performance difference between them is small in absolute terms. The 920MX is the better choice for users who want the most capable entry-level option, while the GT 635M only makes sense if memory bandwidth is the sole priority and the older card is significantly more available or compatible with a specific system.