NVIDIA GeForce GTX 680 vs NVIDIA P106-090 Comparison

The NVIDIA GeForce GTX 680 and the NVIDIA P106-090 represent two very different approaches to GPU design, separated by five years of architectural evolution. The GTX 680 is a flagship Kepler part from 2012, while the P106-090 is a Pascal-based mining card from 2017 with no display outputs. Benchmark data shows the P106-090 holds a decisive lead in compute workloads, but the GTX 680 counters with a broader feature set and higher raw shading throughput. The following analysis breaks down their performance, specifications, and suitability based solely on the provided data.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA GeForce GTX 680 scores 14150 on average, while the NVIDIA P106-090 scores 13470. This places the GTX 680 0.9% ahead of its nearest rival, the NVIDIA Tesla K10, and the P106-090 0.3% behind its closest competitor, the NVIDIA GeForce GTX 570.

Q: How do the two cards compare in OpenCL performance?

A: The P106-090 wins decisively in Geekbench OpenCL, scoring 21304 versus the GTX 680’s 16906. This represents a 20.6% advantage for the Pascal-based card, which is the largest performance gap between the two in any benchmark test.

Q: Which card has a higher pixel fill rate?

A: The P106-090 achieves a pixel rate of 73.49 GPixel/s, which is more than double the GTX 680’s 33.86 GPixel/s. This is despite the GTX 680 having a higher texture rate of 135.4 GTexel/s compared to the P106-090’s 73.49 GTexel/s.

Q: What is the difference in power consumption between the two cards?

A: The GTX 680 has a TDP of 195 W and requires a 450 W suggested PSU, while the P106-090 has a TDP of 75 W and a 250 W suggested PSU. The P106-090 also uses a single 6-pin power connector, whereas the GTX 680 needs two 6-pin connectors.

Q: Do both cards support the same DirectX version?

A: No. The GTX 680 supports DirectX 12 (11_0), while the P106-090 supports DirectX 12 (12_1). The P106-090 also has a newer Vulkan version of 1.4 compared to the GTX 680’s 1.2.175.

Q: Which card has more shading units and texture mapping units?

A: The GTX 680 has 1536 shading units and 128 TMUs, which is double the P106-090’s 768 shading units and 48 TMUs. However, the P106-090 has 48 ROPs versus the GTX 680’s 32 ROPs.

The Verdict

The data presents a clear split decision. For pure compute performance in OpenCL and Vulkan workloads, the NVIDIA P106-090 is the superior card. It wins both head-to-head benchmark tests, with a 20.6% lead in OpenCL and a 5.1% lead in Vulkan. Its higher clock speeds (1354 MHz base, 1531 MHz boost) and newer Pascal architecture overcome the GTX 680’s larger shading unit count.

However, the GTX 680 is the more versatile product. It features display outputs (2x DVI, 1x HDMI 1.4a, 1x DisplayPort 1.2), while the P106-090 has no outputs at all. The GTX 680 also offers higher FP32 compute at 3.250 TFLOPS versus the P106-090’s 2.352 TFLOPS, and it uses a wider 256-bit memory bus. For gaming or any task requiring a visual output, the GTX 680 is the only viable choice.

The P106-090’s 75 W TDP makes it far more power-efficient than the 195 W GTX 680, and its 3 GB of VRAM exceeds the GTX 680’s 2 GB. Benchmark results indicate that users prioritizing raw compute density and power efficiency should choose the P106-090, while those needing a functional graphics card with display connectivity should pick the GTX 680. The P106-090 wins 2 out of 2 head-to-head benchmarks, but the GTX 680 holds a higher average score and better overall percentile ranking (55th versus 54th).

Head-to-Head Benchmarks

The only shared benchmark tests between the two cards are Geekbench OpenCL and Geekbench Vulkan, and the P106-090 dominates both. In OpenCL, the P106-090 scores 21304 against the GTX 680’s 16906, a delta of -20.6% from the GTX 680’s perspective. This is a substantial victory, driven by the P106-090’s higher boost clock of 1531 MHz versus 1058 MHz and its newer Pascal architecture.

The Vulkan test is closer but still favors the P106-090. It scores 18596 compared to the GTX 680’s 17646, a 5.1% advantage. This narrower margin suggests that the GTX 680’s higher texture rate (135.4 GTexel/s) helps it remain competitive in API-level workloads, even though it falls short overall.

Interestingly, the GTX 680 has a higher average benchmark score of 14150 versus 13470 for the P106-090, despite losing both head-to-head tests. This discrepancy arises because the GTX 680 has three benchmark entries (including Geekbench Metal at 7897), while the P106-090 includes a 3DMark Steel Nomad DX12 score of 509, which drags down its average. The GTX 680’s nearest rival, the Tesla K10, sits just 0.9% below it, while the P106-090’s closest competitor, the GTX 570, is only 0.3% ahead.

Specification Differences

The two cards diverge sharply on nearly every specification. The GTX 680 uses a 28 nm process with 3,540 million transistors on a 294 mm² die, while the P106-090 uses a 16 nm process with 4,400 million transistors on a smaller 200 mm² die. This gives the P106-090 a transistor density of 22.0M / mm² versus the GTX 680’s 12.0M / mm².

Clock speeds favor the P106-090 significantly: 1354 MHz base and 1531 MHz boost versus the GTX 680’s 1006 MHz base and 1058 MHz boost. Memory configurations also differ: the GTX 680 has 2 GB GDDR5 on a 256-bit bus with 192.3 GB/s bandwidth, while the P106-090 has 3 GB GDDR5 on a 192-bit bus with 192.2 GB/s bandwidth. Effective memory speed is 6 Gbps for the GTX 680 and 8 Gbps for the P106-090.

The GTX 680 features 1536 shading units, 128 TMUs, and 32 ROPs, whereas the P106-090 has 768 shading units, 48 TMUs, and 48 ROPs. Pixel rate heavily favors the P106-090 at 73.49 GPixel/s versus 33.86 GPixel/s, but texture rate favors the GTX 680 at 135.4 GTexel/s versus 73.49 GTexel/s. FP32 compute is higher on the GTX 680 at 3.250 TFLOPS compared to 2.352 TFLOPS on the P106-090.

Power and connectivity differ dramatically: the GTX 680 draws 195 W with a 450 W PSU requirement, while the P106-090 draws only 75 W with a 250 W PSU. The GTX 680 uses PCIe 3.0 x16, but the P106-090 is limited to PCIe 1.0 x1. The GTX 680 has full display outputs, while the P106-090 has none. The GTX 680 measures 256 mm in length, and the P106-090 is 250 mm.

Architecture Differences

The GTX 680 is built on the Kepler architecture with the GK104 chip, representing the GeForce 600 generation. It was released on 2012-03-21 and is now end-of-life. Kepler was designed for high shading throughput, which explains the GTX 680’s 1536 shading units and 128 TMUs. Its 28 nm TSMC process was state-of-the-art at launch, and the card’s 3,540 million transistors reflect that era’s design priorities.

The P106-090 uses the Pascal architecture with the GP106 chip, part of the “Mining GPUs” generation. Released on 2017-07-30, it benefits from a more advanced 16 nm TSMC process, packing 4,400 million transistors into a smaller 200 mm² die. Pascal introduced significant clock speed improvements, evidenced by the P106-090’s 1531 MHz boost clock. The architecture also supports DirectX 12 (12_1) and Vulkan 1.4, while Kepler only reaches DirectX 12 (11_0) and Vulkan 1.2.175.

The GTX 680 has a predecessor in the GeForce 500 series and a successor in the GeForce 700 series, whereas the P106-090 has no listed predecessor or successor. The P106-090’s lack of display outputs is a defining architectural choice, as it was designed exclusively for mining workloads. The GTX 680’s 2x 6-pin power connectors and 195 W TDP reflect a desktop graphics card design, while the P106-090’s single 6-pin and 75 W TDP indicate a more constrained, compute-focused product. The P106-090 also has a 1:64 FP16 ratio, yielding 36.74 GFLOPS, a feature absent from the GTX 680’s specifications.