GPU Comparison

NVIDIA’s GeForce 920MX and GeForce GT 745M are both end-of-life mobile parts, but they represent two very different design philosophies. The 920MX is a low-power Maxwell chip built for efficiency, while the GT 745M is a higher-power Kepler chip with a wider memory interface. The benchmark data shows a split decision: the 920MX wins in OpenCL compute, while the GT 745M dominates in Vulkan graphics workloads. Their average benchmark scores also tell a story, the GT 745M holds a 12.0% overall lead (3953 vs 3528), placing it in the 23rd percentile of all GPUs versus the 920MX’s 21st percentile.

Where Each One Wins

The NVIDIA GeForce 920MX is the pick for compute-oriented tasks that leverage OpenCL. In the Geekbench OpenCL test, it scores 4068, which is 13.6% higher than the GT 745M’s 3580. This is a significant margin for a chip with fewer shading units and a narrower memory bus. The 920MX’s advantage here likely stems from its Maxwell architecture’s higher clock efficiency, its base clock of 965 MHz and boost of 993 MHz are substantially above the GT 745M’s unspecified clocks, and its FP32 throughput of 508.4 GFLOPS exceeds the GT 745M’s 421.6 GFLOPS. For users running OpenCL-accelerated applications like video encoding filters or physics simulations, the 920MX is the stronger choice despite its lower overall benchmark standing.

The NVIDIA GeForce GT 745M wins decisively in graphics-oriented workloads, specifically Vulkan. Its Geekbench Vulkan score of 5502 crushes the 920MX’s 2987, a 45.7% difference in favor of the GT 745M. This is a massive gap that reflects the GT 745M’s hardware advantages: 384 shading units (vs 256), 32 texture mapping units (vs 24), 16 ROPs (vs 8), and a 128-bit memory bus (vs 64-bit). The GT 745M also uses faster GDDR5 memory at 4 Gbps effective, yielding 64.00 GB/s of bandwidth versus the 920MX’s DDR3 at 14.40 GB/s. For gaming or any Vulkan-based rendering, the GT 745M is in a different class entirely. Its average benchmark score of 3953 also puts it 2.7% ahead of the Quadro 2000M (3434) among its nearest rivals, whereas the 920MX sits 1.8% behind the GT 545 (3594).

Architecture Differences

The two GPUs come from different architectural generations, which explains most of their behavioral divergence. The 920MX is built on the Maxwell architecture using the GM108S chip, fabricated on a 28 nm process at TSMC. It packs 1,020 million transistors into a 77 mm² die, achieving a transistor density of 13.2M / mm². In contrast, the GT 745M uses the older Kepler architecture with the GK107 chip, also on a 28 nm process but with 1,270 million transistors across a larger 118 mm² die, resulting in a lower density of 10.8M / mm².

Memory configuration is a key differentiator. The 920MX uses 2 GB of DDR3 on a 64-bit bus, delivering 14.40 GB/s of bandwidth, a modest figure by any standard. The GT 745M also has 2 GB of memory, but it is GDDR5 on a 128-bit bus, quadrupling bandwidth to 64.00 GB/s. This 4.4x bandwidth advantage is the single biggest architectural reason why the GT 745M wins so heavily in Vulkan, where texture streaming and frame buffer access are memory-bandwidth-bound.

The compute resources also differ significantly. The 920MX has 256 shading units, 24 TMUs, and 8 ROPs, with a pixel rate of 7.944 GPixel/s and a texture rate of 23.83 GTexel/s. The GT 745M has 384 shading units, 32 TMUs, and 16 ROPs, but its pixel rate is lower at 4.392 GPixel/s and texture rate at 17.57 GTexel/s. This is counterintuitive: the GT 745M has more raw units but lower fill rates, likely due to lower clock speeds. However, the GT 745M’s FP32 throughput of 421.6 GFLOPS is lower than the 920MX’s 508.4 GFLOPS, confirming that Maxwell’s per-clock efficiency is superior.

Power and interface differences round out the picture. The 920MX has a TDP of 16 W and uses an MXM Module slot width with no power connectors, while the GT 745M draws 45 W and is listed as an IGP (integrated graphics processor) with a PCIe 3.0 x16 interface. The 920MX uses PCIe 3.0 x8. Both support DirectX 12 (11_0) and OpenGL 4.6, but the 920MX supports Vulkan 1.4 while the GT 745M is limited to Vulkan 1.2.175. The 920MX also has a newer release date (2016-03-24) compared to the GT 745M (2013-03-31), and its predecessor is the GeForce 800M, whereas the GT 745M’s predecessor is the GeForce 600M.

Head-to-Head Benchmarks

The head-to-head data reveals a clean split: each GPU wins one benchmark, and the margins are starkly asymmetric.

Geekbench OpenCL, 920MX wins by 13.6%: The 920MX scores 4068 against the GT 745M’s 3580. This is a comfortable win for the Maxwell part, driven by its higher FP32 throughput (508.4 GFLOPS vs 421.6 GFLOPS) and faster pixel rate (7.944 GPixel/s vs 4.392 GPixel/s). The 920MX’s OpenCL performance places it 2.7% above the Quadro 2000M (3434) among its nearest rivals, showing it competes well with older workstation parts. The GT 745M, by contrast, sits 0.1% below the AMD Radeon R5 M420 (3956) in its own rival list, indicating it is slightly behind in compute.

Geekbench Vulkan, GT 745M wins by 45.7%: This is a landslide. The GT 745M scores 5502 while the 920MX manages only 2987. The 45.7% delta is the largest in either direction, and it reflects the GT 745M’s superior memory subsystem and higher shading unit count. The 920MX’s Vulkan score is so low that it falls 45.7% behind its rival, a gap that no amount of architecture efficiency can bridge. For context, the GT 745M’s Vulkan score is nearly double the 920MX’s OpenCL score, suggesting that the Kepler part is far better optimized for modern graphics APIs.

The net result is a 1-1 tie in head-to-head wins, but the magnitude of the Vulkan victory gives the GT 745M a higher average benchmark score (3953 vs 3528). This means that in any mixed workload, or any pure gaming scenario, the GT 745M is the superior part. The 920MX’s OpenCL win is notable but narrow in context, and it only matters for compute-specific applications.

FAQ

Q: Which GPU is faster overall?

A: The GT 745M has a higher average benchmark score of 3953 compared to the 920MX’s 3528, a 12.0% advantage. It also ranks in the 23rd percentile of all GPUs versus the 920MX’s 21st.

Q: Why does the 920MX win in OpenCL despite being slower overall?

A: The 920MX has higher FP32 throughput (508.4 GFLOPS vs 421.6 GFLOPS) and a faster pixel rate (7.944 GPixel/s vs 4.392 GPixel/s). Its Maxwell architecture delivers better per-clock compute efficiency, which translates to a 13.6% OpenCL win.

Q: Is the GT 745M better for gaming?

A: Yes, based on Vulkan performance. The GT 745M scores 5502 in Geekbench Vulkan, which is 45.7% higher than the 920MX’s 2987. Its larger memory bandwidth (64.00 GB/s vs 14.40 GB/s) and more shading units (384 vs 256) make it far more capable for graphics rendering.

Q: What are the major memory differences?

A: The 920MX uses 2 GB of DDR3 on a 64-bit bus with 14.40 GB/s bandwidth. The GT 745M uses 2 GB of GDDR5 on a 128-bit bus with 64.00 GB/s bandwidth, a 4.4x bandwidth advantage.

Q: How do their power requirements compare?

A: The 920MX has a TDP of 16 W, while the GT 745M draws 45 W. The 920MX is an MXM Module with no power connectors, whereas the GT 745M is an IGP with a PCIe 3.0 x16 interface.

Q: Which GPU has better API support?

A: Both support DirectX 12 (11_0) and OpenGL 4.6. The 920MX supports Vulkan 1.4, while the GT 745M is limited to Vulkan 1.2.175. The 920MX also has a newer release date (2016-03-24 vs 2013-03-31).

Specification Differences

| Specification | NVIDIA GeForce 920MX | NVIDIA GeForce GT 745M |

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

| Architecture | Maxwell | Kepler |

| Chip | GM108S | GK107 |

| Generation | GeForce 900M | GeForce 700M |

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

| Die Size | 77 mm² | 118 mm² |

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

| Memory Type | DDR3 | GDDR5 |

| Memory Bus Width | 64 bit | 128 bit |

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

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

| Shading Units | 256 | 384 |

| TMUs | 24 | 32 |

| ROPs | 8 | 16 |

| Pixel Rate | 7.944 GPixel/s | 4.392 GPixel/s |

| Texture Rate | 23.83 GTexel/s | 17.57 GTexel/s |

| FP32 | 508.4 GFLOPS | 421.6 GFLOPS |

| TDP | 16 W | 45 W |

| Slot Width | MXM Module | IGP |

| Power Connectors | None | Not specified |

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

| Vulkan Version | 1.4 | 1.2.175 |

| Release Date | 2016-03-24 | 2013-03-31 |

| Predecessor | GeForce 800M | GeForce 600M |

| Successor | GeForce 10 Mobile | GeForce 800M |

The specification sheet reads like a classic efficiency-versus-brute-force comparison. The 920MX has 20% fewer transistors, a 35% smaller die, and a 64% lower TDP, yet it posts higher fill rates and FP32 performance. The GT 745M counters with a 4.4x bandwidth advantage, 50% more shading units, and double the ROPs. Neither part is modern by current standards, but the data shows the 920MX is the better compute chip, while the GT 745M is the better graphics chip. For a laptop buyer choosing between these two, the decision hinges entirely on whether the workload is OpenCL compute or Vulkan gaming, there is no middle ground.