NVIDIA GeForce GTX 670M vs NVIDIA Quadro K4000 Comparison

Head-to-Head Benchmarks

The database records a single direct comparison between these two mobile and workstation class GPUs, and the result is a narrow but clear victory for the NVIDIA Quadro K4000. In the Geekbench OpenCL test, the Quadro K4000 scores 6816, while the GeForce GTX 670M trails at 6513. That is a 4.4% gap, which is modest in absolute terms but consistent with the architectural leap between the two products.

The GeForce GTX 670M does not win any head-to-head benchmark in the recorded data. Its single OpenCL result of 6513 places it in a tight cluster of competing mobile parts. The nearest rival data shows the GTX 670M is just 0.3% ahead of the GeForce GT 555M (6493), 0.5% ahead of the Quadro M5000M (6481), and 0.6% ahead of the AMD Radeon Vega 10 Mobile (6476). Against the Intel UHD Graphics P750 (6554), the GTX 670M is 0.6% behind. These deltas are so small that they fall within typical run-to-run variance, meaning the GTX 670M effectively sits in a four-way statistical tie at the 6500-point level.

The Quadro K4000 tells a different story when viewed through its average benchmark score. Its average across all recorded tests is 5982, which is dragged down by a weak Geekbench Metal score of 4166. However, its OpenCL score of 6816 and its Vulkan score of 6964 both exceed the GTX 670M's OpenCL figure. The K4000's nearest rivals cluster around the 5980-5990 range: the Quadro K4000M scores 5986 (0.1% ahead), the AMD FirePro W4100 scores 5987 (0.1% ahead), the AMD Radeon HD 8750M scores 5970 (0.2% behind), and the NVIDIA RTX PRO 6000 Blackwell Server scores 5996 (0.2% ahead). This indicates the K4000's average is competitive with a broader set of workstation and mobile parts, even though its peak OpenCL result is notably higher.

In percentile terms, the GTX 670M ranks higher globally at the 38th percentile across all GPUs, while the K4000 sits at the 34th percentile. This counterintuitive result stems from the K4000's Metal score pulling its average down. The OpenCL head-to-head remains the more relevant comparison for compute workloads, and there the K4000 leads by 4.4%. For a user focused solely on OpenCL performance, the K4000 is the clear pick; for a user considering overall benchmark averages across multiple APIs, the GTX 670M appears stronger despite losing the direct comparison.

Architecture Differences

The two GPUs represent two distinct NVIDIA architectures and process nodes. The GeForce GTX 670M uses the GF114 chip built on Fermi 2.0 architecture, fabricated on a 40 nm process at TSMC. The Quadro K4000 uses the GK106 chip built on Kepler architecture, fabricated on a 28 nm process, also at TSMC. This process shrink allows the K4000 to pack 2,540 million transistors into a 221 mm² die, while the GTX 670M contains 1,950 million transistors on a larger 332 mm² die. Transistor density reflects this: the K4000 achieves 11.5M transistors per mm², nearly double the GTX 670M's 5.9M per mm².

The compute resources differ substantially. The Quadro K4000 carries 768 shading units, 64 texture mapping units, and 24 ROPs. The GeForce GTX 670M has 336 shading units, 56 TMUs, and 24 ROPs. The K4000's shading unit count is more than double, and its TMU count is 14% higher. These differences translate directly into throughput figures. The K4000 delivers 1,244.2 GFLOPS of FP32 compute, compared to 803.7 GFLOPS for the GTX 670M, a 54.8% advantage. Pixel fill rate favors the K4000 at 12.96 GPixel/s versus 8.372 GPixel/s, and texture fill rate favors it at 51.84 GTexel/s versus 33.49 GTexel/s.

Memory configurations also diverge. The GTX 670M has 1536 MB of GDDR5 on a 192-bit bus, with memory clocked at 750 MHz (3 Gbps effective) and bandwidth of 72.00 GB/s. The Quadro K4000 has 3 GB of GDDR5 on the same 192-bit bus, but memory runs at 1404 MHz (5.6 Gbps effective) and delivers 134.8 GB/s, which is 87% higher bandwidth. Both GPUs support DirectX 12 (11_0) and OpenGL 4.6. The K4000 additionally supports Vulkan 1.2.175, while the GTX 670M has no recorded Vulkan support.

Physical and power characteristics differ as well. The GTX 670M is an MXM module with a 75 W TDP, no power connectors, and portable-device-dependent display outputs. The K4000 is a single-slot card measuring 241 mm by 111 mm, with an 80 W TDP, a single 6-pin power connector, a suggested 250 W PSU, and PCIe 2.0 x16 interface. Its display outputs include 1x DVI and 2x DisplayPort 1.2. The GTX 670M uses the MXM-B (3.0) bus interface and has no external power requirement.

FAQ

Q: Which GPU has higher OpenCL performance?

A: The NVIDIA Quadro K4000 scores 6816 in Geekbench OpenCL, which is 4.4% higher than the GeForce GTX 670M's 6513.

Q: Does the GTX 670M win any recorded benchmark?

A: No. The database shows zero wins for the GTX 670M in head-to-head comparisons, with the K4000 taking the only recorded victory.

Q: How does the GTX 670M compare to its nearest rivals?

A: The GTX 670M is 0.3% ahead of the GeForce GT 555M, 0.5% ahead of the Quadro M5000M, and 0.6% ahead of the AMD Radeon Vega 10 Mobile. It is 0.6% behind the Intel UHD Graphics P750.

Q: What is the memory bandwidth difference?

A: The Quadro K4000 offers 134.8 GB/s bandwidth, while the GeForce GTX 670M offers 72.00 GB/s, making the K4000's memory subsystem 87% faster.

Q: Which GPU has a higher global percentile ranking?

A: The GeForce GTX 670M ranks at the 38th percentile across all GPUs, while the Quadro K4000 ranks at the 34th percentile, despite the K4000 winning the OpenCL head-to-head.

Q: What is the launch MSRP of the Quadro K4000?

A: The Quadro K4000 has a launch MSRP of 1,269 USD. The GTX 670M has no recorded launch MSRP.

The Verdict

The data supports a clear split decision based on workload. For OpenCL compute performance, the NVIDIA Quadro K4000 is the stronger part, delivering 4.4% higher scores than the GTX 670M and offering 54.8% more FP32 throughput. Its memory bandwidth advantage of 87% further reinforces its suitability for bandwidth-sensitive tasks. The K4000 also brings Vulkan support, which the GTX 670M lacks, and double the shading units.

However, the GTX 670M holds a higher global percentile ranking (38th versus 34th) and a higher average benchmark score in the database. This suggests that across a broad mix of workloads and API tests, the GTX 670M is more consistent, even though it loses the specific OpenCL comparison. The K4000's average is dragged down by its Metal score of 4166, which is far lower than its OpenCL and Vulkan results.

For a user prioritizing raw compute throughput and memory bandwidth in OpenCL or Vulkan workloads, the Quadro K4000 is the justified choice. For a user looking at overall benchmark averages or working in environments where the Metal API matters, the GeForce GTX 670M appears more balanced. The GTX 670M also offers a lower TDP of 75 W versus 80 W, and it requires no external power connector, making it easier to integrate into portable MXM systems. The K4000 demands a 6-pin power connector and a 250 W PSU suggestion, which suits a desktop workstation slot instead.

The 4.4% OpenCL delta is real but not transformative. The architectural differences, however, are substantial: Kepler's 28 nm process and doubled shading unit count give the K4000 a clear compute ceiling that Fermi 2.0 cannot reach. The GTX 670M's advantage lies in its efficiency profile and its more favorable position in the global percentile distribution. Buyers should weigh the specific API and workload mix rather than rely on a single benchmark result.

Specification Differences

| Field | NVIDIA GeForce GTX 670M | NVIDIA Quadro K4000 |

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

| Chip | GF114 | GK106 |

| Architecture | Fermi 2.0 | Kepler |

| Generation | GeForce 600M | Quadro Kepler (Kx000) |

| Process Node | 40 nm | 28 nm |

| Transistors | 1,950 million | 2,540 million |

| Die Size | 332 mm² | 221 mm² |

| Transistor Density | 5.9M / mm² | 11.5M / mm² |

| Memory Clock | 750 MHz (3 Gbps effective) | 1404 MHz (5.6 Gbps effective) |

| Memory Size | 1536 MB | 3 GB |

| Memory Bandwidth | 72.00 GB/s | 134.8 GB/s |

| Shading Units | 336 | 768 |

| TMUs | 56 | 64 |

| ROPs | 24 | 24 |

| Pixel Rate | 8.372 GPixel/s | 12.96 GPixel/s |

| Texture Rate | 33.49 GTexel/s | 51.84 GTexel/s |

| FP32 | 803.7 GFLOPS | 1,244.2 GFLOPS |

| TDP | 75 W | 80 W |

| Slot Width | MXM Module | Single-slot |

| Power Connectors | None | 1x 6-pin |

| Suggested PSU | None | 250 W |

| Bus Interface | MXM-B (3.0) | PCIe 2.0 x16 |

| Display Outputs | Portable Device Dependent | 1x DVI, 2x DisplayPort 1.2 |

| Vulkan Support | None | 1.2.175 |

| Release Date | 2012-03-21 | 2013-02-28 |

| Predecessor | GeForce 500M | Quadro Fermi |

| Successor | GeForce 700M | Quadro Maxwell |

| Launch MSRP | None | 1,269 USD |

| Average Benchmark Score | 6513 | 5982 |

| Percentile vs All GPUs | 38 | 34 |