NVIDIA GeForce GTX 1650 vs NVIDIA GeForce GTX 670M Comparison

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA GeForce GTX 1650 has an average benchmark score of 7472, while the NVIDIA GeForce GTX 670M has an average benchmark score of 6513. The GTX 1650 sits in the 40th percentile of all GPUs, while the GTX 670M sits in the 38th percentile.

Q: What is the transistor count and process node for each GPU?

A: The GTX 1650 uses 4,700 million transistors on a 12 nm process, while the GTX 670M uses 1,950 million transistors on a 40 nm process. Both are fabricated by TSMC, but the GTX 1650 achieves a transistor density of 23.5M per mm² versus 5.9M per mm² for the GTX 670M.

Q: How do the two GPUs compare in memory bandwidth?

A: The GTX 1650 has a memory bandwidth of 128.1 GB/s with a 128 bit bus and 4 GB of GDDR5 memory. The GTX 670M has a bandwidth of 72.00 GB/s with a 192 bit bus and 1536 MB of GDDR5 memory.

Q: What are the API support differences between the two cards?

A: The GTX 1650 supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The GTX 670M supports DirectX 12 (11_0) and OpenGL 4.6, but has no Vulkan support listed in the database.

Q: Which GPU has a higher pixel rate?

A: The GTX 1650 has a pixel rate of 53.28 GPixel/s, compared to the GTX 670M's 8.372 GPixel/s. This represents a substantial advantage in fill-rate capabilities.

Q: What is the release timeline for each card?

A: The GTX 670M was released on 2012-03-21, while the GTX 1650 was released on 2019-04-22. The GTX 1650 has a launch MSRP of 149 USD, while the GTX 670M has no launch MSRP recorded in the database.

Architecture Differences

The two GPUs represent entirely different architectural generations from NVIDIA. The GTX 1650 is built on the Turing architecture using the TU117 chip, while the GTX 670M is based on the older Fermi 2.0 architecture with the GF114 chip. This generational gap spans the GeForce 16 series versus the GeForce 600M series.

The manufacturing process shows a dramatic difference. The GTX 1650 is fabricated on a 12 nm process at TSMC, whereas the GTX 670M uses a 40 nm process, also at TSMC. The newer process allows the GTX 1650 to pack 4,700 million transistors into a 200 mm² die, giving a density of 23.5M per mm². The GTX 670M, by contrast, contains 1,950 million transistors on a much larger 332 mm² die, yielding a density of only 5.9M per mm². The die size difference is notable: the older Fermi chip is 132 mm² larger despite having fewer than half the transistors.

Compute resources differ substantially. The GTX 1650 has 896 shading units, 56 texture mapping units, and 32 ROPs. The GTX 670M has 336 shading units, 56 TMUs, and 24 ROPs. Both cards share the same TMU count, but the GTX 1650 has roughly 2.7 times the shading units and 1.3 times the ROPs. Neither card features ray tracing cores or tensor cores.

Clock behavior also differs. The GTX 1650 has a base clock of 1485 MHz and a boost clock of 1665 MHz, while the GTX 670M has no base or boost clock recorded in the database. The memory clocks show the GTX 1650 running at 2001 MHz with 8 Gbps effective, while the GTX 670M runs at 750 MHz with 3 Gbps effective.

The GTX 1650 achieves a texture rate of 93.24 GTexel/s and a pixel rate of 53.28 GPixel/s. The GTX 670M manages 33.49 GTexel/s and 8.372 GPixel/s respectively. In floating-point performance, the GTX 1650 delivers 2.984 TFLOPS in FP32 with 5.967 TFLOPS in FP16 (at a 2:1 ratio), while the GTX 670M reaches 803.7 GFLOPS in FP32 with no FP16 capability listed.

Power consumption is identical at 75 W TDP for both cards. The GTX 1650 uses no power connectors and has a suggested PSU of 250 W, while the GTX 670M also requires no power connectors but has no suggested PSU recorded. The GTX 1650 is a dual-slot card with dimensions of 229 mm length, 111 mm height, 35 mm width, while the GTX 670M is an MXM module with no dimensions listed.

Head-to-Head Benchmarks

The database contains one direct head-to-head benchmark between these two GPUs: Geekbench OpenCL. In this test, the GTX 1650 scores 29629, while the GTX 670M scores 6513. The delta is 354.9% in favor of the GTX 1650, meaning the newer card outperforms the older one by more than four and a half times in raw OpenCL compute throughput.

This single head-to-head result is decisive. The GTX 1650 wins the only shared benchmark in the database, giving it a 1-0 advantage in direct comparison wins. The magnitude of the victory is substantial enough that the performance gap is not close: the GTX 670M would need to more than quadruple its score to match the GTX 1650 in this workload.

Looking at the broader benchmark landscape, the GTX 1650 has a much richer set of recorded results. It has scores for 3DMark Steel Nomad DX12 (305), Passmark DirectX 10 (39), DirectX 11 (58), DirectX 12 (35), DirectX 9 (124), G2D (561), G3D (7880), and GPU Compute (3048), plus Geekbench Vulkan (33042). The GTX 670M only has the one Geekbench OpenCL result in the database.

The average benchmark scores reinforce the head-to-head result. The GTX 1650 averages 7472 across all recorded tests, while the GTX 670M averages 6513. The percentiles also show separation: the GTX 1650 sits at the 40th percentile of all GPUs, and the GTX 670M sits at the 38th percentile. While both are in the lower-middle range of the overall GPU distribution, the GTX 1650 is positioned higher.

The nearest rival data for each card provides context for their respective tiers. The GTX 1650 is closest to the AMD Radeon HD 8850M (delta 0.3%), Intel UHD Graphics 750 (delta 0.4%), AMD Radeon R7 350 (delta 0.6%), and trails the Intel Arc A310 by 1%. The GTX 670M is closest to the NVIDIA GeForce GT 555M (delta 0.3%), NVIDIA Quadro M5000M (delta 0.5%), AMD Radeon Vega 10 Mobile (delta 0.6%), and trails the Intel UHD Graphics P750 by 0.6%. This places each card in a tight cluster of similarly performing GPUs, but the clusters are separated by roughly 15% in average score.

Specification Differences

The two GPUs differ across nearly every major specification category. The process node moves from 40 nm to 12 nm, a three-generation jump in manufacturing technology. Transistor count increases from 1,950 million to 4,700 million, and die size actually shrinks from 332 mm² to 200 mm². Transistor density improves from 5.9M per mm² to 23.5M per mm².

Memory configuration changes significantly. The GTX 670M uses 1536 MB of GDDR5 on a 192 bit bus with 72.00 GB/s bandwidth. The GTX 1650 uses 4 GB of GDDR5 on a 128 bit bus with 128.1 GB/s bandwidth. Although the bus width narrows from 192 bit to 128 bit, the higher memory clock (2001 MHz versus 750 MHz) more than compensates, delivering 78% more bandwidth.

Compute unit counts differ as follows: shading units increase from 336 to 896, TMUs stay identical at 56, and ROPs increase from 24 to 32. The clock speeds show the GTX 1650 with a base of 1485 MHz and boost of 1665 MHz, while the GTX 670M has no recorded base or boost clocks.

Raw throughput metrics all favor the GTX 1650. Pixel rate rises from 8.372 GPixel/s to 53.28 GPixel/s, a 6.4x improvement. Texture rate rises from 33.49 GTexel/s to 93.24 GTexel/s, a 2.8x improvement. FP32 compute rises from 803.7 GFLOPS to 2.984 TFLOPS, a 3.7x improvement. The GTX 1650 adds FP16 support at 5.967 TFLOPS, while the GTX 670M has no FP16 capability.

API support changes as well. DirectX support moves from 12 (11_0) to 12 (12_1), and Vulkan support is added at version 1.4. OpenGL remains at 4.6 for both. The bus interface changes from MXM-B (3.0) to PCIe 3.0 x16, reflecting the GTX 670M's laptop module form factor versus the GTX 1650's desktop card form factor. The GTX 1650 has display outputs of 1x DVI, 1x HDMI 2.0, and 1x DisplayPort 1.4a, while the GTX 670M's display outputs are portable device dependent.

Both cards share a 75 W TDP and no power connectors. The GTX 1650 is dual-slot with specified dimensions, while the GTX 670M is an MXM module with no dimensions recorded. Both are end-of-life products. The GTX 670M's predecessors and successors are the GeForce 500M and GeForce 700M respectively, while the GTX 1650's are GeForce 10 and GeForce 20.

Where Each One Wins

The GTX 1650 wins in every category where data exists for comparison. In the only direct head-to-head benchmark, Geekbench OpenCL, it leads by 354.9%. In raw specifications, it leads in shading units, ROPs, pixel rate, texture rate, FP32 compute, memory bandwidth, and memory capacity. It also adds Vulkan support, which the GTX 670M lacks entirely.

The GTX 1650 also wins on efficiency metrics. It delivers 2.984 TFLOPS from 4,700 million transistors on a 200 mm² die, while the GTX 670M delivers 803.7 GFLOPS from 1,950 million transistors on a 332 mm² die. The newer card achieves roughly 3.7 times the FP32 throughput while using the same 75 W TDP and a smaller physical die.

The GTX 670M's only advantages are in areas that do not translate to performance. It has a wider memory bus at 192 bit versus 128 bit, and a larger die size at 332 mm² versus 200 mm². Neither of these is a practical benefit in benchmark results. The wider bus is offset by the much lower memory clock, and the larger die is a consequence of the older, less dense manufacturing process.

For use cases, the GTX 1650 is the clear choice for any workload that depends on compute throughput, fill rate, or modern API support. Its Vulkan 1.4 support and DirectX 12 (12_1) feature level make it suitable for contemporary graphics APIs. The GTX 670M, with only DirectX 12 (11_0) and no Vulkan, is limited to older API workloads.

The GTX 670M's form factor as an MXM module means it was designed for portable devices, while the GTX 1650 is a dual-slot desktop card with standard display outputs. For anyone building or upgrading a desktop system, the GTX 1650 is the only one of the two that fits standard PCIe slots. For a legacy laptop with an MXM slot, the GTX 670M would be the only option, though its performance ceiling is far lower.

The Verdict

The data points to a decisive conclusion: the GTX 1650 is the superior GPU in nearly every measurable way. Its 354.9% lead in the sole head-to-head benchmark, its 7.4x advantage in pixel rate, its 2.8x advantage in texture rate, and its 3.7x advantage in FP32 compute all indicate a generational leap rather than an incremental improvement.

The GTX 1650's average benchmark score of 7472 versus the GTX 670M's 6513 places it about 15% higher on average, and its 40th percentile ranking versus the 38th percentile confirms the separation. The nearest rival clusters for each card show that while both sit in competitive tiers, the GTX 1650's tier is measurably faster.

For users choosing between these two, the GTX 1650 is the recommended pick for any desktop application. It offers more memory (4 GB versus 1536 MB), higher bandwidth (128.1 GB/s versus 72.00 GB/s), better API support including Vulkan, and dramatically higher compute throughput, all within the same 75 W TDP envelope. Its launch MSRP was 149 USD.

The GTX 670M should only be considered in the specific scenario of a legacy MXM-based laptop that requires a drop-in replacement. In that narrow context, its 336 shading units, 1536 MB of memory, and 803.7 GFLOPS of compute may suffice for older games or basic tasks. But even there, the performance gap to the GTX 1650 is so large that no workload where both cards are viable options would favor the GTX 670M.

The record is clear: one head-to-head benchmark, one win for the GTX 1650, with a delta of 354.9%. The specification tables tell the same story across every meaningful metric. The GTX 1650 is a modern GPU with modern capabilities, while the GTX 670M is a legacy part from an older architecture that cannot keep pace.