NVIDIA GeForce GT 635M vs NVIDIA GeForce GT 745M Comparison

# Head-to-Head Benchmarks

The data presents a single direct comparison between the NVIDIA GeForce GT 745M and the NVIDIA GeForce GT 635M, and the result is a narrow victory for the older Fermi-based part. In the Geekbench OpenCL test, the GT 635M scores 3740 against the GT 745M's 3580, a delta of -4.3% from the perspective of the GT 745M. This means the GT 635M finishes roughly 4.3% ahead in this specific workload, which is a surprisingly tight margin given the architectural differences between the two.

However, the broader benchmark picture reveals a more complex story. The GT 745M has been tested across three separate Geekbench workloads — Metal, OpenCL, and Vulkan — while the GT 635M has only a single OpenCL result. The GT 745M's Vulkan score of 5502 is its strongest showing, while its Metal result of 2777 is its weakest. Its OpenCL score of 3580 sits between the two. The avgBenchmarkScore for the GT 745M is 3953, which incorporates all three tests, while the GT 635M's avgBenchmarkScore is 3740, based solely on its OpenCL run.

This creates an interesting analytical question: is the GT 635M genuinely faster in compute tasks, or is its single OpenCL result an outlier? The percentile rankings suggest the two are extremely close in overall performance. The GT 745M sits at the 23rd percentile of all GPUs, while the GT 635M is at the 22nd percentile. That one-percentile gap is negligible in real-world terms, but the delta in their average scores — 3953 versus 3740 — represents a 5.7% advantage for the GT 745M. The head-to-head OpenCL result contradicts that average, which suggests the GT 635M may have a specific strength in OpenCL that does not translate to other APIs.

Where Each One Wins

Looking strictly at the benchmark data, the GT 635M wins the only direct head-to-head test available — Geekbench OpenCL — with a 4.3% margin. This is its sole recorded victory, and it is the only data point we have for this GPU. The implications are narrow but real: in OpenCL compute workloads, the GT 635M appears to hold a slight edge over the GT 745M, despite being based on older Fermi architecture and having significantly fewer shading units.

The GT 745M, by contrast, shows its strength in the breadth of its benchmark coverage. Its average score of 3953 is higher than the GT 635M's 3740, and it demonstrates capability across Metal, OpenCL, and Vulkan APIs. The Vulkan score of 5502 is particularly notable — it is roughly 54% higher than its own OpenCL score, indicating that the Kepler architecture scales much better with Vulkan's modern API design. The GT 635M has no Vulkan support listed in its API specifications, which means it cannot even participate in that comparison.

For use-case analysis, the data suggests the GT 745M is the more versatile choice for applications that leverage Vulkan or Metal — both of which are common in modern gaming and graphics workloads. The GT 635M, with its single OpenCL strength, may be adequate for legacy compute tasks but lacks the API breadth to compete in contemporary software environments. The GT 745M's pixel rate of 4.392 GPixel/s versus 1.900 GPixel/s for the GT 635M further reinforces its advantage in graphics output, even if the OpenCL compute race is closer than expected.

Architecture Differences

The architectural gulf between these two GPUs is substantial, yet the benchmark results suggest the gap has narrowed in practice. The GT 745M uses the GK107 chip built on TSMC's 28 nm process, packing 1,270 million transistors into a 118 mm² die. The GT 635M uses the GF108 chip on TSMC's older 40 nm node, with 585 million transistors on a 116 mm² die. The transistor density difference is stark: 10.8 million transistors per mm² for the GT 745M versus 5.0 million for the GT 635M. Kepler is fundamentally a more efficient architecture, yet the Fermi part still manages to win the OpenCL test.

The compute resources tell a similar story of Kepler superiority on paper. The GT 745M has 384 shading units, 32 texture mapping units, and 16 raster output pipelines. The GT 635M has just 96 shading units, 16 TMUs, and 4 ROPs. That is a 4x difference in shading units and a 2x difference in TMUs and ROPs. The GT 745M's FP32 throughput is 421.6 GFLOPS versus 182.4 GFLOPS for the GT 635M — a 2.3x advantage. Yet in OpenCL, the GT 635M wins. This suggests that raw compute specifications do not always translate directly to benchmark performance, possibly due to driver maturity, memory bandwidth bottlenecks, or workload characteristics.

Memory is another major divergence. Both cards have 2 GB of VRAM and a 128-bit bus, but the GT 745M uses GDDR5 at 1000 MHz (4 Gbps effective) delivering 64.00 GB/s of bandwidth. The GT 635M uses DDR3 at 900 MHz (1800 Mbps effective) for just 28.80 GB/s. The GT 745M has more than double the memory bandwidth, which should help in texture-heavy and bandwidth-sensitive workloads. The memory clock difference — 1000 MHz versus 900 MHz — is relatively modest, but the GDDR5 versus DDR3 distinction is decisive for throughput.

The feature set also differs. The GT 745M supports Vulkan 1.2.175, while the GT 635M lists no Vulkan support. Both support DirectX 12 (11_0) and OpenGL 4.6. The bus interface is PCIe 3.0 x16 on the GT 745M versus PCIe 2.0 x16 on the GT 635M, which affects data transfer speeds between the GPU and system memory. The GT 745M has a TDP of 45 W, while the GT 635M draws 35 W. Both are integrated into laptops (IGP form factor), and the GT 745M's higher power envelope likely reflects its larger compute array and faster memory.

The Verdict

The data presents a counterintuitive conclusion: the older, less powerful GPU wins the only direct benchmark comparison, yet the newer GPU holds a higher average score and broader API support. For users whose primary workload is OpenCL compute, the GT 635M appears to offer a slight edge — a 4.3% advantage in the head-to-head test. This is a narrow margin, but it is the only direct evidence we have.

For everyone else, the GT 745M is the stronger choice based on the aggregate data. Its average benchmark score of 3953 beats the GT 635M's 3740 by 5.7%, and its Vulkan support opens the door to modern graphics workloads that the GT 635M cannot handle. The GT 745M's 2.3x advantage in FP32 throughput, 2.2x advantage in memory bandwidth, and 4x advantage in shading units suggest it should outperform the GT 635M in most scenarios, even if the OpenCL result tells a different story.

The percentile rankings reinforce this: the GT 745M sits at the 23rd percentile versus the GT 635M's 22nd, placing the former marginally higher in the global GPU hierarchy. The nearest rival data also shows the GT 745M trading blows with the NVIDIA GeForce 830M (avg score 3957, delta -0.1%) and AMD Radeon R5 M420 (avg score 3956, delta -0.1%), while the GT 635M competes with the NVIDIA Quadro 3000M (avg score 3718, delta 0.6%) and Intel UHD Graphics 710 (avg score 3792, delta -1.4%). This places the GT 745M in a slightly higher performance tier.

The verdict is nuanced but clear: the GT 745M is the better all-around GPU, but the GT 635M is not the pushover its specifications suggest. If OpenCL compute is the sole criterion, the GT 635M wins. If the user needs Vulkan, higher memory bandwidth, or better overall average performance, the GT 745M is the data-backed choice.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA GeForce GT 745M has an avgBenchmarkScore of 3953, while the GT 635M has an avgBenchmarkScore of 3740.

Q: What is the result of the direct head-to-head OpenCL test?

A: The GT 635M wins the Geekbench OpenCL test with a score of 3740 against the GT 745M's 3580, a delta of -4.3% from the GT 745M's perspective.

Q: Does the GT 635M support Vulkan?

A: No, the GT 635M lists no Vulkan support in its API specifications, while the GT 745M supports Vulkan 1.2.175.

Q: How do the memory bandwidth figures compare?

A: The GT 745M has 64.00 GB/s of bandwidth from GDDR5 memory, while the GT 635M has 28.80 GB/s from DDR3 memory.

Q: What are the transistor counts for each GPU?

A: The GT 745M has 1,270 million transistors, while the GT 635M has 585 million transistors.

Q: Which GPU has a higher pixel rate?

A: The GT 745M has a pixel rate of 4.392 GPixel/s, while the GT 635M has a pixel rate of 1.900 GPixel/s.

Specification Differences

| Specification | NVIDIA GeForce GT 745M | NVIDIA GeForce GT 635M |

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

| Chip | GK107 | GF108 |

| Architecture | Kepler | Fermi |

| Generation | GeForce 700M | GeForce 600M |

| Process Node | 28 nm | 40 nm |

| Transistors | 1,270 million | 585 million |

| Die Size | 118 mm² | 116 mm² |

| Transistor Density | 10.8M / mm² | 5.0M / mm² |

| Memory Type | GDDR5 | DDR3 |

| Memory Clock | 1000 MHz (4 Gbps effective) | 900 MHz (1800 Mbps effective) |

| Memory Bandwidth | 64.00 GB/s | 28.80 GB/s |

| Shading Units | 384 | 96 |

| TMUs | 32 | 16 |

| ROPs | 16 | 4 |

| Pixel Rate | 4.392 GPixel/s | 1.900 GPixel/s |

| Texture Rate | 17.57 GTexel/s | 7.600 GTexel/s |

| FP32 | 421.6 GFLOPS | 182.4 GFLOPS |

| TDP | 45 W | 35 W |

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

| Vulkan Support | 1.2.175 | None |

| Release Date | 2013-03-31 | 2012-03-21 |

| Predecessor | GeForce 600M | GeForce 500M |

| Successor | GeForce 800M | GeForce 700M |

| Vulkan Benchmark Score | 5502 | Not tested |

| Metal Benchmark Score | 2777 | Not tested |

| OpenCL Benchmark Score | 3580 | 3740 |