NVIDIA GeForce GTX 670M vs NVIDIA Quadro K4000M Comparison

The Verdict

The recorded benchmark data favors the NVIDIA GeForce GTX 670M for raw compute workloads. In the sole head-to-head measurement, Geekbench OpenCL, the GTX 670M scores 6513 against the Quadro K4000M’s 5986, a decisive 8.8% advantage. The GTX 670M also places higher in the database’s overall percentile ranking, sitting at the 38th percentile versus the K4000M’s 34th. For users whose primary concern is absolute OpenCL throughput, the GTX 670M is the clear choice from the data.

However, the Quadro K4000M is not without merit. It carries significantly more memory, 4 GB versus 1536 MB, and a wider memory bus, 256 bit versus 192 bit. This configuration yields a higher memory bandwidth of 89.60 GB/s compared to the GTX 670M’s 72.00 GB/s. For workloads that are memory-capacity sensitive, such as large dataset manipulation or certain professional visualization tasks, the K4000M’s memory advantage is a tangible asset. The data suggests a split decision: the GTX 670M for compute performance, the K4000M for memory capacity and bandwidth.

Users seeking a professional-grade mobile GPU with Kepler architecture and Vulkan support should lean toward the Quadro K4000M. The GTX 670M, while faster in the recorded benchmark, offers no Vulkan support in the database, whereas the K4000M lists Vulkan 1.2.175. This feature gap could matter for modern applications that rely on Vulkan. Both cards are end-of-life products, so the decision is about matching the right legacy hardware to the specific workload.

Architecture Differences

The two GPUs come from different NVIDIA architectures and process nodes. The GeForce GTX 670M uses the GF114 chip built on the Fermi 2.0 architecture, fabricated on a 40 nm process at TSMC. The Quadro K4000M employs the GK104 chip with the Kepler architecture, also from TSMC but on a smaller 28 nm process. This process difference is substantial: the K4000M packs 3,540 million transistors into a 294 mm² die, yielding a transistor density of 12.0M per mm². The GTX 670M has 1,950 million transistors on a larger 332 mm² die, giving a density of 5.9M per mm². The Kepler chip is therefore more than twice as dense.

Core counts also differ sharply. The GTX 670M has 336 shading units, 56 texture mapping units, and 24 ROPs. The K4000M more than doubles the shading units to 960, with 80 TMUs and 32 ROPs. These higher counts translate into higher theoretical pixel and texture rates: the K4000M reaches 12.02 GPixel/s and 48.08 GTexel/s, while the GTX 670M manages 8.372 GPixel/s and 33.49 GTexel/s. Despite this, the GTX 670M still wins the OpenCL benchmark, suggesting that raw throughput in those specific rates does not directly determine compute performance.

Clock speeds differ as well. The GTX 670M has no listed base or boost clock, but its memory runs at 750 MHz, yielding 3 Gbps effective. The K4000M has a base and boost clock of 601 MHz, with memory at 700 MHz or 2.8 Gbps effective. The lower memory clock on the K4000M is offset by its wider 256-bit bus, producing the higher overall bandwidth. Both cards use the MXM-B (3.0) bus interface and are MXM modules, with power connectors listed as none. The K4000M has a higher TDP of 100 W versus the GTX 670M’s 75 W.

Feature support also diverges. Both cards list DirectX 12 (11_0) and OpenGL 4.6. The K4000M adds Vulkan 1.2.175 support, while the GTX 670M has no Vulkan entry. This is a meaningful architectural difference for software that targets Vulkan. Both are end-of-life, with the GTX 670M released on 2012-03-21 and the K4000M on 2012-05-31. The GTX 670M’s predecessor is the GeForce 500M and its successor is the GeForce 700M; the K4000M’s predecessor is Quadro Fermi-M and its successor is Quadro Maxwell-M.

Where Each One Wins

Based strictly on the database, the GTX 670M wins the only head-to-head benchmark. It achieves an 8.8% higher Geekbench OpenCL score, which places it ahead in compute-oriented tasks that stress OpenCL execution. The GTX 670M also has a higher average benchmark score, 6513 versus 5986, and a higher percentile rank, 38 versus 34. Users running OpenCL-based workloads, such as certain physics simulations or data processing pipelines, should expect the GTX 670M to deliver faster results.

The K4000M wins in memory-focused scenarios. Its 4 GB frame buffer is more than double the GTX 670M’s 1536 MB, and its 89.60 GB/s bandwidth exceeds the GTX 670M’s 72.00 GB/s. For applications that load large textures, hold extensive geometry data, or require substantial VRAM for rendering, the K4000M has a clear capacity advantage. The K4000M also offers Vulkan support, which the GTX 670M lacks, making it the only option of the two for Vulkan-based rendering or compute.

The K4000M’s higher theoretical pixel and texture rates, 12.02 GPixel/s and 48.08 GTexel/s, suggest it may excel in fill-rate-bound scenarios, even though the OpenCL score is lower. Conversely, the GTX 670M’s higher FP32 throughput, 803.7 GFLOPS versus 1,153.9 GFLOPS for the K4000M, is actually lower on paper, yet it wins the compute benchmark. This indicates that the GTX 670M’s Fermi architecture extracts better real-world compute efficiency from its resources in this specific test. The verdict: GTX 670M for compute wins, K4000M for memory capacity and Vulkan compatibility.

FAQ

Q: Which GPU has the higher benchmark score in the database?

A: The NVIDIA GeForce GTX 670M, with a Geekbench OpenCL score of 6513, is 8.8% higher than the Quadro K4000M’s 5986.

Q: Does the Quadro K4000M have more memory?

A: Yes, the K4000M has 4 GB of GDDR5 memory, while the GTX 670M has 1536 MB. The K4000M also has a wider 256-bit bus and higher bandwidth of 89.60 GB/s.

Q: Which GPU supports Vulkan?

A: Only the Quadro K4000M lists Vulkan support, specifically version 1.2.175. The GTX 670M has no Vulkan entry in the database.

Q: What are the TDP differences?

A: The GTX 670M has a TDP of 75 W, while the Quadro K4000M has a TDP of 100 W.

Q: Are both GPUs still in production?

A: No, both are end-of-life products. The GTX 670M was released on 2012-03-21, and the K4000M on 2012-05-31.

Q: How do the shading unit counts compare?

A: The Quadro K4000M has 960 shading units, significantly more than the GTX 670M’s 336. Despite this, the GTX 670M wins the recorded compute benchmark.

Head-to-Head Benchmarks

The database contains a single head-to-head comparison: Geekbench OpenCL. In this test, the GTX 670M scores 6513 and the K4000M scores 5986. The delta percentage is 8.8%, and the winner is the GTX 670M. This is a substantial margin in compute performance, indicating that the GTX 670M’s Fermi architecture is more efficient in this workload despite having fewer shading units and a smaller memory bus.

For context, the GTX 670M’s nearest rivals in the database include the GeForce GT 555M at 6493 (0.3% slower), the Quadro M5000M at 6481 (0.5% slower), and the AMD Radeon Vega 10 Mobile at 6476 (0.6% slower). The Intel UHD Graphics P750 scores 6554, which is 0.6% higher than the GTX 670M. This places the GTX 670M in a tight cluster of similarly performing GPUs, with the K4000M trailing by a clear margin.

The K4000M’s nearest rivals tell a similar story. The AMD FirePro W4100 scores 5987, essentially tied with the K4000M at 0% delta. The NVIDIA Quadro K4000 scores 5982, just 0.1% lower. The NVIDIA RTX PRO 6000 Blackwell Server scores 5996, which is 0.2% higher, and the GeForce GTX 770M scores 6000, also 0.2% higher. The K4000M sits at the lower end of this group, while the GTX 670M sits at the upper end of its group. This reinforces the head-to-head result: the GTX 670M is the faster compute part.

Looking at other metrics, the K4000M’s theoretical FP32 throughput is 1,153.9 GFLOPS, which is 43.6% higher than the GTX 670M’s 803.7 GFLOPS. Yet the OpenCL benchmark shows the opposite result. This discrepancy highlights that theoretical peak throughput does not always translate to benchmark performance. The K4000M also has higher pixel and texture rates, but again, these do not produce a higher OpenCL score. The recorded data clearly favors the GTX 670M in this single compute test.

Specification Differences

| Specification | NVIDIA GeForce GTX 670M | NVIDIA Quadro K4000M |

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

| Architecture | Fermi 2.0 | Kepler |

| Chip | GF114 | GK104 |

| Process Node | 40 nm | 28 nm |

| Transistors | 1,950 million | 3,540 million |

| Die Size | 332 mm² | 294 mm² |

| Transistor Density | 5.9M / mm² | 12.0M / mm² |

| Memory Size | 1536 MB | 4 GB |

| Memory Type | GDDR5 | GDDR5 |

| Memory Bus Width | 192 bit | 256 bit |

| Memory Bandwidth | 72.00 GB/s | 89.60 GB/s |

| Shading Units | 336 | 960 |

| TMUs | 56 | 80 |

| ROPs | 24 | 32 |

| Pixel Rate | 8.372 GPixel/s | 12.02 GPixel/s |

| Texture Rate | 33.49 GTexel/s | 48.08 GTexel/s |

| FP32 | 803.7 GFLOPS | 1,153.9 GFLOPS |

| TDP | 75 W | 100 W |

| Vulkan API | None | 1.2.175 |

| Release Date | 2012-03-21 | 2012-05-31 |

| Predecessor | GeForce 500M | Quadro Fermi-M |

| Successor | GeForce 700M | Quadro Maxwell-M |

The two GPUs share the same manufacturer, bus interface (MXM-B 3.0), form factor (MXM Module), power connector setup (none), and display outputs (portable device dependent). Both support DirectX 12 (11_0) and OpenGL 4.6. The GTX 670M’s memory clock is 750 MHz (3 Gbps effective), while the K4000M’s is 700 MHz (2.8 Gbps effective). The K4000M has a base and boost clock of 601 MHz, whereas the GTX 670M has no listed base or boost clock. Neither has a launch MSRP in the database.