NVIDIA GeForce GTX 670 vs NVIDIA P106-090 Comparison

The Verdict

The data presents a clear, if unusual, hierarchy. The NVIDIA P106-090 outperforms the NVIDIA GeForce GTX 670 in every head-to-head benchmark recorded, winning both tests. Its average benchmark score of 13470 places it 5.5% above the GTX 670's 12773. The P106-090 also holds a higher percentile rank among all GPUs, sitting at the 54th percentile versus the GTX 670's 52nd. For tasks that leverage OpenCL or Vulkan compute, the P106-090 is the definitive choice, offering a 38.9% lead in Geekbench OpenCL and a 19.6% lead in Geekbench Vulkan.

However, the GTX 670 is not without its own strengths, though these are not reflected in compute benchmarks. It is a full-fledged graphics card with display outputs, whereas the P106-090 has no display outputs at all. The GTX 670 also features a wider 256-bit memory bus and a higher number of shading units (1344 vs. 768), which historically suggests a strong rasterization performance profile. Yet, the benchmark data does not include a head-to-head gaming test, so any such advantage remains speculative based on specifications alone. The P106-090, a mining-specific part, is the superior compute performer; the GTX 670 is the only one of the two that can function as a standard desktop graphics solution.

Architecture Differences

The two cards represent distinct architectural generations and design philosophies. The P106-090 is built on the Pascal architecture, utilizing the GP106 chip fabricated on a 16 nm process at TSMC. In contrast, the GTX 670 uses the Kepler architecture with the GK104 chip, also from TSMC but on a larger 28 nm process. This process difference is stark: the P106-090 packs 4,400 million transistors into a 200 mm² die, achieving a transistor density of 22.0M per mm². The GTX 670, with 3,540 million transistors spread across a 294 mm² die, has a significantly lower density of 12.0M per mm². This explains the P106-090's much lower 75 W TDP compared to the GTX 670's 170 W.

The core configurations also differ substantially. The P106-090 has 768 shading units, 48 texture mapping units (TMUs), and 48 render output units (ROPs). The GTX 670 has more shading units (1344) and TMUs (112) but fewer ROPs (32). Consequently, the GTX 670's texture rate is higher at 109.8 GTexel/s versus the P106-090's 73.49 GTexel/s. However, the P106-090's pixel rate is higher at 73.49 GPixel/s versus the GTX 670's 27.44 GPixel/s, a result of its higher clock speeds. The P106-090's base clock is 1354 MHz with a boost of 1531 MHz, while the GTX 670 runs at a base of 915 MHz and a boost of 980 MHz.

Memory subsystems differ as well. The P106-090 has 3 GB of GDDR5 on a 192-bit bus, yielding a bandwidth of 192.2 GB/s. The GTX 670 has 2 GB of GDDR5 on a wider 256-bit bus, yielding a nearly identical bandwidth of 192.3 GB/s. The P106-090 supports DirectX 12 (12_1) and Vulkan 1.4, while the GTX 670 supports DirectX 12 (11_0) and Vulkan 1.2.175. The P106-090's interface is limited to PCIe 1.0 x1, whereas the GTX 670 has a full PCIe 3.0 x16 connection. Finally, the P106-090 has no display outputs, while the GTX 670 offers 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2.

FAQ

Q: Which GPU has a higher average benchmark score?

A: The NVIDIA P106-090 has a higher average benchmark score of 13470, while the NVIDIA GeForce GTX 670 scores 12773. This represents a 5.5% advantage for the P106-090.

Q: How does the P106-090 perform in Geekbench OpenCL compared to the GTX 670?

A: The P106-090 scores 21304 in Geekbench OpenCL, which is 38.9% higher than the GTX 670's score of 15341. This is the largest performance delta between the two cards in any test.

Q: Is the GTX 670 a viable option for a system requiring display outputs?

A: Yes. The GTX 670 has display outputs including 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2. The P106-090, in contrast, has no display outputs, making it unsuitable as a primary display adapter.

Q: What are the power requirements for these two cards?

A: The P106-090 has a TDP of 75 W and requires a single 6-pin power connector with a suggested PSU of 250 W. The GTX 670 has a much higher TDP of 170 W, needs two 6-pin power connectors, and a suggested PSU of 450 W.

Q: Which GPU supports a newer version of Vulkan?

A: The NVIDIA P106-090 supports Vulkan 1.4, while the NVIDIA GeForce GTX 670 supports the older Vulkan 1.2.175 version.

Q: What is the memory configuration difference?

A: The P106-090 has 3 GB of GDDR5 memory on a 192-bit bus. The GTX 670 has 2 GB of GDDR5 memory on a wider 256-bit bus. Despite the bus width difference, their memory bandwidths are nearly identical at 192.2 GB/s and 192.3 GB/s, respectively.

Specification Differences

| Field | NVIDIA P106-090 | NVIDIA GeForce GTX 670 |

|:--- |:--- |:--- |

| Chip | GP106 | GK104 |

| Architecture | Pascal | Kepler |

| Generation | Mining GPUs | GeForce 600 |

| Process Node | 16 nm | 28 nm |

| Transistors | 4,400 million | 3,540 million |

| Die Size | 200 mm² | 294 mm² |

| Transistor Density | 22.0M / mm² | 12.0M / mm² |

| Base Clock | 1354 MHz | 915 MHz |

| Boost Clock | 1531 MHz | 980 MHz |

| Memory Clock | 2002 MHz (8 Gbps effective) | 1502 MHz (6 Gbps effective) |

| Memory Size | 3 GB | 2 GB |

| Memory Bus Width | 192 bit | 256 bit |

| Shading Units | 768 | 1344 |

| TMUs | 48 | 112 |

| ROPs | 48 | 32 |

| Pixel Rate | 73.49 GPixel/s | 27.44 GPixel/s |

| Texture Rate | 73.49 GTexel/s | 109.8 GTexel/s |

| FP32 | 2.352 TFLOPS | 2.634 TFLOPS |

| FP16 | 36.74 GFLOPS (1:64) | None |

| TDP | 75 W | 170 W |

| Power Connectors | 1x 6-pin | 2x 6-pin |

| Suggested PSU | 250 W | 450 W |

| Bus Interface | PCIe 1.0 x1 | PCIe 3.0 x16 |

| Display Outputs | No outputs | 2x DVI, 1x HDMI 1.4a, 1x DisplayPort 1.2 |

| DirectX Support | 12 (12_1) | 12 (11_0) |

| Vulkan Support | 1.4 | 1.2.175 |

| Release Date | 2017-07-30 | 2012-05-09 |

| Predecessor | None | GeForce 500 |

| Successor | None | GeForce 700 |

| Launch MSRP | None | 399 USD |

Head-to-Head Benchmarks

The head-to-head data is limited to two compute-centric tests, but the results are decisive. The NVIDIA P106-090 wins both, securing a 2-0 victory in the benchmark suite.

The most significant win for the P106-090 comes in Geekbench OpenCL. Here, it scores 21304 against the GTX 670's 15341, a substantial 38.9% lead. This large delta is a strong indicator of the architectural efficiency of the Pascal design, particularly its 16 nm process and higher clock speeds. The P106-090's boost clock of 1531 MHz is significantly higher than the GTX 670's 980 MHz, which directly contributes to its compute throughput. While the GTX 670 has more shading units, the P106-090's per-clock efficiency appears to be far superior, as evidenced by this massive score gap.

The second test, Geekbench Vulkan, shows a narrower but still comfortable margin for the P106-090. It scores 18596, which is 19.6% ahead of the GTX 670's 15553. The P106-090's support for Vulkan 1.4, a newer API revision than the GTX 670's 1.2.175, may play a role in this performance advantage. Even in a test that may favor the newer driver and API path, the P106-090 maintains a clear edge.

It is worth remembering the GTX 670's higher FP32 throughput (2.634 TFLOPS vs. 2.352 TFLOPS) and superior texture rate (109.8 GTexel/s vs. 73.49 GTexel/s) do not translate into a win in these tests. This suggests that the benchmark workload is not purely dependent on raw pixel or texture processing. The P106-090's higher pixel rate (73.49 GPixel/s vs. 27.44 GPixel/s) and its architectural advancements likely play a more critical role in these specific compute workloads. The data shows a clear, consistent pattern: the newer Pascal-based mining card is the superior compute processor, despite its lower TDP and fewer shading units.