NVIDIA GeForce GTX 670M

The NVIDIA GeForce GTX 670M is a mobile graphics processor from the GeForce 600M generation, built on the Fermi 2.0 architecture with a 40 nm TSMC process. It packs 336 shading units, 56 texture mapping units, and 24 ROPs, with a 192-bit GDDR5 memory interface. In the Geekbench OpenCL benchmark, it scores 6513 points, placing it at the 37th percentile of all GPUs in the database—a modest position that signals entry-level performance for its era. The GPU draws 75 W, uses an MXM module form factor, and is now end-of-life, having been released on March 21, 2012, between the GeForce 500M and 700M generations.

Benchmark Performance

The GTX 670M’s Geekbench OpenCL score of 6513 is the sole benchmark result in the data pack, and it anchors every comparative analysis. Against its nearest rivals, the performance envelope is extremely tight. The GPU trails the Intel UHD Graphics P750 by 0.4% (Intel scores 6538), sits ahead of the AMD Radeon Vega 10 Mobile by 0.6% (AMD scores 6476), falls behind the AMD Radeon HD 7730M by 0.7% (AMD scores 6560), and leads the NVIDIA Quadro M5000M by 0.8% (NVIDIA scores 6463). These deltas are all under one percentage point, meaning the GTX 670M is functionally tied with each of these competitors in raw OpenCL compute throughput.

The 37th percentile ranking reinforces this picture. The GPU outperforms only 37% of all GPUs in the benchmark database, which suggests that most desktop and many mobile parts deliver higher compute performance. In terms of theoretical throughput, the GTX 670M offers 803.7 GFLOPS of FP32 compute, a texture fill rate of 33.49 GTexel/s, and a pixel rate of 8.372 GPixel/s. These figures align with its modest benchmark standing—they are competitive with the integrated and entry-level discrete parts listed as rivals, but far from the top of the mobile GPU stack. The data shows that the GTX 670M is a baseline performer, suitable for light to moderate workloads, but not for demanding compute or high-end gaming.

Who Should Consider It

Given its 37th percentile ranking and the near-tie with integrated graphics like the Intel UHD Graphics P750, the GTX 670M is best suited for users who need a discrete mobile GPU with modest compute capability. The 1536 MB GDDR5 frame buffer and 72.00 GB/s memory bandwidth are enough for older or less demanding titles at reduced detail settings, though the data pack does not include specific game benchmarks. At lower resolutions and with conservative graphics settings, the GTX 670M can handle casual gaming, media playback, and light productivity tasks. Its 75 W TDP and MXM module design make it a candidate for laptops or portable workstations that require a dedicated GPU without excessive power draw. However, the 37th percentile standing means it will struggle with modern high-end games or compute-intensive applications. Users who prioritize battery life over raw performance may find the GTX 670M adequate, but those seeking higher frame rates or future-proofing should look to parts with better percentile scores.

How It Compares

vs. Intel UHD Graphics P750: The GTX 670M scores 6513 against the Intel’s 6538, a 0.4% deficit. This places the two effectively on par in OpenCL compute, despite the Intel part being an integrated solution. The GTX 670M’s discrete nature and dedicated memory may offer advantages in real-world graphics workloads, but the benchmark data shows no measurable compute lead.

vs. AMD Radeon Vega 10 Mobile: The GTX 670M is 0.6% ahead of the AMD Radeon Vega 10 Mobile, which scores 6476. This is the smallest margin among the four rivals, indicating that the two GPUs are interchangeable in compute performance. The Vega 10 Mobile is also a mobile part, so the comparison is direct—neither holds a significant edge.

vs. AMD Radeon HD 7730M: The GTX 670M trails the AMD Radeon HD 7730M by 0.7%, with the AMD part scoring 6560. This is the largest negative delta in the rival set, but at 0.7% it remains negligible. Both are discrete mobile GPUs from the same era, and the benchmark suggests they are equivalent in OpenCL performance.

vs. NVIDIA Quadro M5000M: The GTX 670M leads the NVIDIA Quadro M5000M by 0.8%, scoring 6513 against 6463. The Quadro M5000M is a professional-grade mobile GPU, so the GTX 670M’s slight edge in this compute test is notable, though the margin is too small to imply a meaningful performance difference in professional applications.

FAQ

Q: What is the Geekbench OpenCL score of the GTX 670M?

A: The GTX 670M scores 6513 points in the Geekbench OpenCL benchmark, placing it at the 37th percentile of all GPUs in the database.

Q: What memory configuration does the GTX 670M use?

A: It has 1536 MB of GDDR5 memory on a 192-bit bus, providing 72.00 GB/s of bandwidth. The memory clock is 750 MHz, which translates to 3 Gbps effective.

Q: Does the GTX 670M support DirectX 12?

A: Yes, it supports DirectX 12 (11_0) and OpenGL 4.6. Vulkan support is not listed in the data pack.

Q: What is the thermal design power (TDP) of the GTX 670M?

A: The TDP is 75 W. The GPU uses an MXM module slot and requires no external power connectors.

Q: When was the GTX 670M released?

A: It was released on March 21, 2012, and is now end-of-life. Its predecessor is the GeForce 500M series and its successor is the GeForce 700M series.

Q: How many shading units does the GTX 670M have?

A: It has 336 shading units, along with 56 texture mapping units and 24 ROPs.

Power and Cooling

The GTX 670M has a TDP of 75 W, which is modest for a discrete mobile GPU. It is designed as an MXM module with an MXM-B (3.0) bus interface, and it requires no external power connectors—power is drawn entirely from the MXM slot. The data pack does not specify a suggested PSU, but given the low TDP, a standard laptop power supply is sufficient. The 40 nm process node and 1,950 million transistors on a 332 mm² die contribute to the relatively low power envelope. The GPU’s end-of-life status and 75 W rating make it suitable for systems where thermal and power budgets are constrained. Because it is an MXM module, cooling is typically handled by the host laptop’s cooling solution; the GPU itself does not include a reference cooler in the data pack.

Ray Tracing and Feature Set

The GTX 670M does not include dedicated ray tracing cores or tensor cores—the data pack lists both as null. Its architecture is Fermi 2.0, which predates NVIDIA’s RT and tensor core technologies. The feature set is defined by its API support: DirectX 12 (11_0) and OpenGL 4.6. Vulkan support is not listed, so users should not assume Vulkan compatibility. The GPU includes 336 shading units, 56 TMUs, and 24 ROPs, which provide the raw compute and rasterization capabilities. The absence of ray tracing and tensor cores means the GTX 670M cannot accelerate ray-traced effects or AI-based features like DLSS. For modern games that rely on these technologies, the GTX 670M would be limited to traditional rasterization methods, and its 37th percentile performance further constrains its ability to handle such workloads.

Memory Subsystem

The GTX 670M is equipped with 1536 MB of GDDR5 memory, a 192-bit memory bus, and a bandwidth of 72.00 GB/s. The memory clock is 750 MHz, yielding an effective data rate of 3 Gbps. This configuration is typical for an entry-level mobile GPU of its generation. The 192-bit bus is narrower than higher-end parts, and the 72.00 GB/s bandwidth is modest by modern standards. At high resolutions or with heavy texture filtering, the memory subsystem may become a bottleneck, as the bandwidth is insufficient to feed the GPU’s 336 shading units at maximum throughput. The 1536 MB capacity is also limited for large texture sets or high-resolution assets. However, for the target use case of lighter gaming and productivity at lower settings, the memory subsystem provides adequate capacity and speed. The data shows that the GTX 670M’s memory bandwidth is in line with its nearest rivals, all of which fall within a narrow performance band, so it does not stand out positively or negatively in this regard.