NVIDIA GeForce GT 635M vs NVIDIA GeForce GT 645M Comparison

Head-to-Head Benchmarks

The only direct comparison available in the database is the Geekbench OpenCL test, and the result is a clear win for the older Fermi-based part. The NVIDIA GeForce GT 635M scores 3740, while the NVIDIA GeForce GT 645M scores 2680. That is a 28.3% deficit for the GT 645M, meaning the GT 635M is roughly 39.6% faster in this single workload. This is an unusual outcome because the GT 645M carries a newer architecture and a much larger shader count, but the recorded data shows the GT 635M ahead by a wide margin.

Looking at broader averages, the GT 645M has an average benchmark score of 4411 across three tests (Geekbench Metal, OpenCL, and Vulkan), while the GT 635M has only one recorded test (OpenCL) with an average of 3740. However, the head-to-head result is the only apples-to-apples comparison, and it favors the GT 635M. The GT 645M does have additional benchmark results that are not directly comparable to the GT 635M, so the overall picture is limited to that single OpenCL run.

The GT 645M’s other scores are worth noting for context. Its Geekbench Metal score is 5679, and its Vulkan score is 4875, both of which are higher than its OpenCL score. But since the GT 635M has no Metal or Vulkan results in the database, those numbers cannot be used for a direct comparison. The only valid head-to-head is the OpenCL test, where the GT 635M wins decisively.

The percentile ranking reinforces this split. The GT 645M sits at the 26th percentile of all GPUs, while the GT 635M sits at the 22nd percentile. That means the GT 645M is actually placed slightly higher in the overall distribution, despite losing the OpenCL head-to-head. This suggests that the GT 645M’s Metal and Vulkan scores are pulling its average up, but for OpenCL specifically, the GT 635M is the stronger part.

Nearest rivals for the GT 645M include the NVIDIA GeForce 930M (average score 4388, 0.5% lower), the Intel Iris Pro Graphics 5200 (4360, 1.2% lower), and the NVIDIA GeForce RTX 4070 GDDR6 (4335, 1.8% lower), with the AMD Radeon R7 M260 (4499) coming in 1.9% higher. These deltas are all tiny, meaning the GT 645M’s average score of 4411 places it in a very tight cluster. For the GT 635M, its nearest rivals are the NVIDIA Quadro 3000M (3718, 0.6% lower), the NVIDIA GeForce GT 740M (3717, 0.6% lower), and the NVIDIA GeForce 825M (3694, 1.2% lower), with the Intel UHD Graphics 710 (3792) sitting 1.4% higher. Again, the differences are small, but the GT 635M’s average is slightly below that cluster’s midpoint.

In terms of raw compute, the GT 645M has a much higher theoretical peak. Its FP32 throughput is 599.0 GFLOPS versus 182.4 GFLOPS for the GT 635M, a factor of about 3.3. Pixel rate is 6.240 GPixel/s versus 1.900 GPixel/s, and texture rate is 24.96 GTexel/s versus 7.600 GTexel/s. These theoretical numbers suggest the GT 645M should dominate, but the OpenCL benchmark does not reflect that. The recorded data shows the opposite, so the practical outcome favors the GT 635M in that specific test.

One possible explanation is driver maturity or workload characteristics, but the database does not include such details. What is clear is that on the single shared benchmark, the GT 635M wins by a substantial margin, and that is the only cross-comparison available.

The Verdict

If you are choosing strictly on the recorded benchmark data, the NVIDIA GeForce GT 635M is the safer pick for OpenCL-based workloads. It wins the head-to-head by 28.3% and holds a higher score in that specific test. The GT 645M does have a better average score when you include its Metal and Vulkan results, but those tests are not available for the GT 635M, so they do not factor into a direct comparison.

The GT 645M is the better choice if you expect to use applications that leverage Metal or Vulkan, since its scores there are much higher than its OpenCL result. But for a pure OpenCL scenario, the GT 635M is ahead. The data does not support a claim that the GT 645M is faster overall; it only shows that it has a higher average across a different set of tests.

From a positioning standpoint, the GT 645M’s percentile rank is 26, versus 22 for the GT 635M. That means the GT 645M is placed higher in the global database, but the margin is small and the only direct benchmark contradicts that ranking. A practical builder should not assume the newer architecture automatically wins.

Neither part has a launch MSRP in the database, so no pricing information is available. Both are end-of-life products, released in 2012, with the GT 635M coming first in March and the GT 645M following in September. Both are IGP (integrated) solutions with no power connectors, so they are meant for laptops where the display output is portable device dependent.

For an experienced builder or buyer looking at used or legacy laptops, the GT 635M offers a better OpenCL result, but the GT 645M has a stronger feature set in terms of API support and theoretical compute. The verdict depends entirely on which workloads you run. If OpenCL is your primary metric, choose the GT 635M. If you need Vulkan support or higher Metal scores, the GT 645M is the only one with those results in the database.

Where Each One Wins

The GT 635M wins in the only direct benchmark: Geekbench OpenCL. It posts 3740 versus 2680, a 28.3% lead. This is the most concrete finding in the entire comparison. If your software uses OpenCL for acceleration, the GT 635M is the better performer according to the recorded data.

The GT 645M wins in theoretical specifications and in API breadth. It has 384 shading units versus 96, 32 TMUs versus 16, and 16 ROPs versus 4. Its FP32 output is 599.0 GFLOPS versus 182.4 GFLOPS. It also supports Vulkan 1.2.175, while the GT 635M has no Vulkan entry in the database. For applications that rely on Vulkan or Metal, the GT 645M has recorded scores that are significantly higher than its own OpenCL score, but there is no equivalent data for the GT 635M.

The GT 645M also has a higher pixel rate (6.240 GPixel/s versus 1.900 GPixel/s) and texture rate (24.96 GTexel/s versus 7.600 GTexel/s), which could matter for fill-rate-bound tasks, although no benchmark directly confirms that. The GT 635M, by contrast, has a lower TDP at 35 W versus 32 W for the GT 645M, so the GT 645M is slightly more power-efficient on paper.

In terms of memory, both have 2 GB of DDR3 on a 128-bit bus with 28.80 GB/s bandwidth. That is identical, so memory capacity and bandwidth do not differentiate them. The GT 635M uses a 40 nm process while the GT 645M uses 28 nm, but that does not translate into a benchmark win in the head-to-head.

For use-case planning, the GT 645M is better for modern API workloads, especially Vulkan, since it has a recorded Vulkan score of 4875. The GT 635M has no Vulkan support listed, so it may not work with Vulkan-based applications at all. The GT 645M also has a Metal score of 5679, which is its highest recorded result. The GT 635M has no Metal score, so it is unclear if it supports Metal or how it would perform.

The GT 635M is better for OpenCL-specific tasks, such as compute-heavy filters or physics simulations that use that API. Its OpenCL score is 3740, which is higher than the GT 645M’s 2680. That is the only clear-cut win in the database, but it is a decisive one.

FAQ

Q: Which GPU has a higher average benchmark score?

A: The NVIDIA GeForce GT 645M has an average benchmark score of 4411, while the NVIDIA GeForce GT 635M has an average of 3740. However, the GT 645M has three recorded tests (Metal, OpenCL, Vulkan), while the GT 635M has only one (OpenCL), so the averages are not directly comparable.

Q: Is the GT 635M faster than the GT 645M in OpenCL?

A: Yes. In the Geekbench OpenCL test, the GT 635M scores 3740 versus 2680 for the GT 645M, a 28.3% difference in favor of the GT 635M. This is the only head-to-head benchmark recorded in the database.

Q: Does the GT 645M support Vulkan?

A: Yes, the GT 645M lists Vulkan 1.2.175 support and has a Geekbench Vulkan score of 4875. The GT 635M has no Vulkan version listed and no Vulkan benchmark score in the database.

Q: What is the memory configuration of both GPUs?

A: Both have 2 GB of DDR3 memory on a 128-bit bus, with a bandwidth of 28.80 GB/s. There is no difference in memory size, type, bus width, or bandwidth.

Q: Which GPU has more shading units?

A: The GT 645M has 384 shading units, while the GT 635M has 96. The GT 645M also has 32 TMUs and 16 ROPs, versus 16 TMUs and 4 ROPs for the GT 635M.

Q: Are these GPUs still in production?

A: No, both are marked as end-of-life. The GT 635M was released in March 2012, and the GT 645M was released in September 2012. Both have the GeForce 500M series as a predecessor and the GeForce 700M series as a successor.

Architecture Differences

The GT 645M is built on the GK107 chip using the Kepler architecture, while the GT 635M uses the GF108 chip with the older Fermi architecture. This is a fundamental generational shift. Kepler is more efficient per clock and per transistor, but the Fermi part still wins the OpenCL benchmark in this data.

The process node differs as well. The GT 645M uses 28 nm, while the GT 635M uses 40 nm. Both are manufactured by TSMC, but the smaller node allows the GT 645M to pack 1,270 million transistors on a 118 mm² die, giving a transistor density of 10.8M per mm². The GT 635M has 585 million transistors on a 116 mm² die, for a density of 5.0M per mm². That means the GT 645M has more than double the transistor count and double the density.

The GT 645M has 384 shading units, 32 TMUs, and 16 ROPs. The GT 635M has 96 shading units, 16 TMUs, and 4 ROPs. This is a 4x difference in shader count and a 2x difference in TMUs, with a 4x difference in ROPs. The theoretical rates reflect this: the GT 645M reaches 599.0 GFLOPS FP32, 24.96 GTexel/s, and 6.240 GPixel/s, versus 182.4 GFLOPS, 7.600 GTexel/s, and 1.900 GPixel/s for the GT 635M.

Clock speeds are only listed for the GT 645M, with a base of 709 MHz and a boost of 780 MHz. The GT 635M has no base or boost clock in the database. Memory clocks are the same for both: 900 MHz, with 1800 Mbps effective. Because the memory bus and type are identical, bandwidth is also identical at 28.80 GB/s.

API support differs. The GT 645M lists DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. The GT 635M lists DirectX 12 (11_0) and OpenGL 4.6, but no Vulkan version. This means the GT 645M is ahead in API feature coverage, particularly for Vulkan.

Both use a PCIe interface, but the GT 645M uses PCIe 3.0 x16, while the GT 635M uses PCIe 2.0 x16. The GT 645M also has a lower TDP at 32 W versus 35 W for the GT 635M, despite having far more compute resources. Both are IGP solutions with no power connectors and portable device dependent display outputs.

The GT 645M has a transistor density of 10.8M per mm², which is more than twice the GT 635M’s 5.0M per mm². This is a direct result of the 28 nm process versus 40 nm. The die sizes are similar (118 mm² versus 116 mm²), but the GT 645M fits far more transistors on that area.

In terms of production status, both are end-of-life. The GT 635M launched earlier in 2012, and the GT 645M came later that same year. Both share the same predecessor and successor in the GeForce 500M and 700M series, respectively.

The architecture difference explains why the GT 645M has a higher theoretical throughput, but it does not explain the OpenCL result. The database records the GT 635M as the winner in that test, so the newer architecture does not automatically translate to better real-world performance in every API.

Specification Differences

The two GPUs differ in several key specifications. The chip is different: GK107 for the GT 645M versus GF108 for the GT 635M. The architecture is different: Kepler versus Fermi. The process node is different: 28 nm versus 40 nm.

Transistor count is 1,270 million versus 585 million, and die size is 118 mm² versus 116 mm². Transistor density is 10.8M per mm² versus 5.0M per mm². These are all clear numerical differences.

Clock speeds: the GT 645M has a base clock of 709 MHz and a boost of 780 MHz. The GT 635M has no base or boost clock listed. Memory clock is the same for both at 900 MHz with 1800 Mbps effective.

Shading units: 384 versus 96. TMUs: 32 versus 16. ROPs: 16 versus 4. Pixel rate: 6.240 GPixel/s versus 1.900 GPixel/s. Texture rate: 24.96 GTexel/s versus 7.600 GTexel/s. FP32: 599.0 GFLOPS versus 182.4 GFLOPS.

TDP: 32 W versus 35 W. Bus interface: PCIe 3.0 x16 versus PCIe 2.0 x16. Vulkan support: 1.2.175 for the GT 645M, none listed for the GT 635M. DirectX and OpenGL versions are identical: 12 (11_0) and 4.6.

Memory is identical: 2 GB DDR3, 128-bit bus, 28.80 GB/s bandwidth. Both have no power connectors and are IGP with portable device dependent outputs. Both are end-of-life, with the GT 635M released in March 2012 and the GT 645M in September 2012. Neither has a launch MSRP in the database.