NVIDIA GeForce GT 635M vs NVIDIA GeForce GT 740 Comparison

# Head-to-Head Benchmarks

The only direct benchmark comparison available between the NVIDIA GeForce GT 635M and the NVIDIA GeForce GT 740 is the Geekbench OpenCL test. In this single head-to-head matchup, the GT 740 posts a score of 3847 against the GT 635M's 3740. That translates to a 2.8% advantage for the GT 740, a relatively narrow margin that suggests the two GPUs occupy a similar performance tier in compute workloads. The delta is small enough that real-world differences in OpenCL applications could easily be obscured by driver optimizations, system memory configuration, or thermal conditions.

Looking at the broader context, the GT 635M's 3740 score places it at the 22nd percentile of all GPUs in the database. Its nearest rivals include the NVIDIA Quadro 3000M at 3718 (0.6% slower), the NVIDIA GeForce GT 740M at 3717 (0.6% slower), and the NVIDIA GeForce 825M at 3694 (1.2% slower). The only competitor ahead of it in that immediate cluster is the Intel UHD Graphics 710 at 3792, which beats it by 1.4%. The GT 635M is thus bracketed by a tight group of mobile and entry-level desktop parts, all within roughly 1.5% of each other.

The GT 740, despite winning the head-to-head, sits at the 21st percentile overall — one point below the GT 635M. Its average benchmark score across all tests is 3431, which is lower than its OpenCL result because it also runs Metal and Vulkan workloads. Its nearest rivals tell a similar story: the NVIDIA Quadro 2000M at 3434 is 0.1% faster, the NVIDIA GeForce 920MX at 3528 is 2.7% faster, while the Intel HD Graphics P4600 at 3389 trails by 1.2% and the Intel HD Graphics 530 at 3332 trails by 3%. The GT 740's position in the rankings is therefore essentially equivalent to the GT 635M's, despite the slight OpenCL win.

What is striking is how close these two parts are in compute performance despite their architectural differences. A 2.8% delta in OpenCL is well within the noise floor for many workloads. The GT 635M's single benchmark score of 3740 is its only data point, while the GT 740 has three scores: OpenCL at 3847, Vulkan at 4083, and Metal at 2363. The Vulkan score is notably higher than OpenCL, suggesting the GT 740's Kepler architecture handles Vulkan more efficiently than OpenCL. The Metal score, however, is dramatically lower — a 38.6% drop from the OpenCL result — which likely reflects driver maturity or API-specific limitations rather than raw hardware capability.

Since the GT 635M has no Vulkan or Metal scores in the database, a direct comparison across those APIs is impossible. The available data shows only the OpenCL matchup, and in that test the GT 740 wins by a modest 2.8%. For users prioritizing compute performance in OpenCL, the GT 740 has a slight edge. For those curious about Vulkan performance, the GT 740's 4083 score indicates it is about 6.1% faster in Vulkan than in OpenCL, which may be relevant for newer applications that leverage that API.

# FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA GeForce GT 635M has an average benchmark score of 3740, while the NVIDIA GeForce GT 740 has an average of 3431. The GT 635M is therefore about 9% higher on average, though this is skewed by the GT 740's low Metal score.

Q: How does the GT 740 perform in Vulkan compared to OpenCL?

A: The GT 740 scores 4083 in Geekbench Vulkan and 3847 in Geekbench OpenCL. That makes the Vulkan result approximately 6.1% higher than OpenCL, indicating better relative performance in Vulkan workloads.

Q: What is the GT 635M's closest competitor in the database?

A: The NVIDIA Quadro 3000M is the closest, with an average score of 3718 versus the GT 635M's 3740 — a delta of only 0.6%. The NVIDIA GeForce GT 740M is nearly identical at 3717, also 0.6% behind.

Q: Does the GT 740 support Vulkan while the GT 635M does not?

A: Yes. The GT 740 lists Vulkan API support at version 1.2.175, while the GT 635M has no Vulkan API information listed in the database.

Q: Which GPU has a higher pixel fill rate?

A: The GT 740 has a pixel rate of 7.944 GPixel/s, compared to the GT 635M's 1.900 GPixel/s. That is a 4.2x advantage for the GT 740 in pixel throughput.

Q: What is the memory bandwidth difference between the two?

A: The GT 740 delivers 80.19 GB/s of bandwidth, while the GT 635M provides 28.80 GB/s. The GT 740 offers roughly 2.8 times the memory bandwidth of the GT 635M.

# The Verdict

The data presents a nuanced picture. In raw compute benchmarks, the GT 740 wins the only direct comparison — OpenCL at 3847 versus 3740, a 2.8% margin. However, the GT 635M holds a higher average benchmark score (3740 versus 3431) due to the GT 740's weak Metal showing. Users who rely on OpenCL should prefer the GT 740, as it demonstrates a measurable, though small, advantage. Users on macOS or Metal-based applications would see significantly worse performance on the GT 740, given its 2363 Metal score — a 36.9% deficit compared to its own OpenCL result.

The GT 740 is the more capable part in most respects. It has quadruple the shading units (384 versus 96), double the texture mapping units (32 versus 16), quadruple the ROPs (16 versus 4), and more than four times the FP32 throughput (762.6 GFLOPS versus 182.4 GFLOPS). Its memory subsystem is vastly superior: GDDR5 at 80.19 GB/s versus DDR3 at 28.80 GB/s, with the same 128-bit bus width. These specification advantages suggest the GT 740 should substantially outperform the GT 635M in graphics-intensive tasks, even if the OpenCL benchmark shows only a modest gap.

The GT 635M's strengths lie in its efficiency and integration. It has a 35W TDP versus the GT 740's 64W, and it is designed as an IGP (integrated graphics processor) with portable-device-dependent display outputs. The GT 740, by contrast, is a single-slot desktop card requiring a 6-pin power connector and a 250W suggested PSU. For a laptop or compact system, the GT 635M is the only viable option. For a desktop with adequate power, the GT 740 is clearly the better performer.

Both GPUs are end-of-life, so the choice is likely between used or refurbished parts. The GT 740 was released on 2014-05-28, while the GT 635M came earlier on 2012-03-21. The GT 740 is also from a newer generation (GeForce 700 versus GeForce 600M), which aligns with its architectural advantages.

# Specification Differences

The two GPUs differ across nearly every specification category. The GT 635M uses the GF108 chip built on a 40 nm process, while the GT 740 uses the GK107 chip on a 28 nm process. Transistor counts tell a dramatic story: the GT 635M has 585 million transistors on a 116 mm² die, yielding a density of 5.0M transistors per mm². The GT 740 packs 1,270 million transistors into a 118 mm² die, achieving 10.8M per mm² — more than double the density.

Memory configurations diverge sharply. The GT 635M has 2 GB of DDR3 memory with a 128-bit bus and 28.80 GB/s bandwidth. The GT 740 has 1024 MB of GDDR5 memory, also on a 128-bit bus, but with 80.19 GB/s bandwidth. The GT 740's memory clock is 1253 MHz (5 Gbps effective), compared to the GT 635M's 900 MHz (1800 Mbps effective). The GT 740's smaller capacity but faster type and bandwidth makes it better suited for bandwidth-hungry workloads.

Compute resources are heavily skewed toward the GT 740. It has 384 shading units, 32 TMUs, and 16 ROPs, versus the GT 635M's 96 shading units, 16 TMUs, and 4 ROPs. Pixel rate is 7.944 GPixel/s versus 1.900 GPixel/s; texture rate is 31.78 GTexel/s versus 7.600 GTexel/s; FP32 is 762.6 GFLOPS versus 182.4 GFLOPS. The GT 740 also supports Vulkan 1.2.175, while the GT 635M lists no Vulkan support. Both support DirectX 12 (11_0) and OpenGL 4.6.

Physical and power characteristics differ as well. The GT 635M is an IGP with no power connectors and a 35W TDP, while the GT 740 is a single-slot card measuring 145 mm (5.7 inches) in length, with a 1x 6-pin power connector and a 64W TDP. The GT 740 has a suggested PSU of 250W. The bus interface is PCIe 2.0 x16 on the GT 635M versus PCIe 3.0 x16 on the GT 740. Display outputs are portable-device-dependent on the GT 635M, while the GT 740 offers 2x DVI and 1x mini-HDMI 1.4a.

# Architecture Differences

The architectural gap is fundamental. The GT 635M is based on NVIDIA's Fermi architecture, which debuted in 2010 and was designed for the GeForce 400 and 500 series. The GT 740 uses Kepler, the architecture that succeeded Fermi and powered the GeForce 600 and 700 series. Kepler introduced significant efficiency improvements and a different shader design philosophy. Fermi relied on a more complex scheduling model, while Kepler simplified the front-end and increased the number of cores per SM, allowing for higher throughput with lower clock speeds.

The process node difference is substantial: 40 nm for the GT 635M versus 28 nm for the GT 740. This shrinks the GT 740's transistors significantly, which explains why it packs more than double the transistor count (1,270 million versus 585 million) on a nearly identical die size (118 mm² versus 116 mm²). The density figure of 10.8M transistors per mm² for the GT 740 versus 5.0M for the GT 635M quantifies this leap.

The shading unit count is the most telling architectural difference. The GT 740's 384 shading units are exactly four times the GT 635M's 96. In Kepler, each streaming multiprocessor (SM) contains 192 CUDA cores, whereas Fermi's SM contained 32 or 48 depending on the variant. This allows the GT 740 to execute far more parallel threads, which shows up in its 762.6 GFLOPS FP32 rate versus 182.4 GFLOPS for the GT 635M. The GT 740's texture rate of 31.78 GTexel/s versus 7.600 GTexel/s and pixel rate of 7.944 GPixel/s versus 1.900 GPixel/s follow the same pattern.

Cache hierarchies and memory controllers are not detailed in the data, but the memory type difference (GDDR5 versus DDR3) and the effective clock speed (5 Gbps versus 1800 Mbps) indicate a more efficient memory subsystem on the GT 740. The GT 635M's higher memory capacity (2 GB versus 1024 MB) is its only advantage in the memory department, likely aimed at laptop use where memory size matters for larger textures.

# Where Each One Wins

The GT 740 wins in virtually every performance metric where both GPUs have data. It is faster in OpenCL (3847 versus 3740), has higher pixel and texture rates, more compute units, and significantly more memory bandwidth. For gaming, 3D rendering, or any graphics-intensive workload on a desktop, the GT 740 is the clear choice. Its Vulkan score of 4083 suggests it handles modern APIs well, and its GDDR5 memory provides the bandwidth needed for high-resolution textures and effects.

The GT 635M wins in efficiency and portability. Its 35W TDP is nearly half the GT 740's 64W, making it suitable for thin-and-light laptops where power draw and heat dissipation are critical. Its IGP form factor and portable-device-dependent display outputs mean it is integrated into a system rather than occupying a slot. The 2 GB memory capacity is double the GT 740's 1024 MB, which can be an advantage in memory-constrained scenarios, though the slower DDR3 type limits its practical benefit.

In compute workloads, the GT 635M's single OpenCL score of 3740 is within 2.8% of the GT 740's 3847, and its average benchmark score of 3740 actually exceeds the GT 740's 3431 average. This means for pure compute tasks that are not memory-bandwidth-bound, the GT 635M is nearly as capable as the GT 740, despite having a quarter of the shading units. This is likely due to the GT 635M's higher effective clock utilization or the nature of the OpenCL workload tested.

For users with a desktop system and a power supply of at least 250W, the GT 740 is the superior choice in almost every scenario. For users with a laptop or a low-power SFF system, the GT 635M is the only practical option, and its compute performance is respectable enough for light workloads. The GT 740's single-slot size and 6-pin connector mean it requires a standard desktop case with available PCIe power. The GT 635M's lack of power connectors and IGP design makes it invisible to the user in terms of installation.