NVIDIA GeForce GTX 660 Ti vs NVIDIA P106-090 Comparison

Head-to-Head Benchmarks

The recorded data shows a clear split between these two NVIDIA offerings, with the P106-090 taking both head-to-head benchmark victories. In Geekbench OpenCL, the P106-090 scores 21,304 points against the GTX 660 Ti's 15,113 points, a 29.1% advantage. This is a substantial margin, placing the mining-oriented card well ahead in compute workloads that leverage OpenCL's parallel processing capabilities.

The gap narrows somewhat in Geekbench Vulkan, where the P106-090 posts 18,596 points versus the GTX 660 Ti's 15,012 points. The 19.3% delta still favors the Pascal-based card, but the older Kepler architecture holds up better in this API, suggesting that Vulkan's lower overhead benefits the GTX 660 Ti's larger shading unit count relative to its overall compute throughput.

Looking at the broader benchmark picture, the two cards occupy similar percentile territory despite their different benchmark profiles. The GTX 660 Ti sits at the 57th percentile of all GPUs, while the P106-090 lands at the 54th percentile. Their average benchmark scores tell a more nuanced story: the GTX 660 Ti averages 15,063 points, while the P106-090 averages 13,470 points. This discrepancy arises because the P106-090's benchmark suite includes the 3DMark Steel Nomad DX12 test, where it scores only 509 points, dragging down its average considerably.

The nearest rival data confirms that both cards compete in the same performance tier. The GTX 660 Ti's closest competitors include the AMD Radeon Pro 560X (15,082 average, 0.1% ahead), the AMD Radeon RX 7600 (15,171 average, 0.7% ahead), and the NVIDIA GeForce RTX 3050 OEM (15,199 average, 0.9% ahead). The P106-090's nearest rivals include the NVIDIA GeForce GTX 570 (13,515 average, 0.3% ahead), the AMD Radeon Pro 555 (13,407 average, 0.5% behind), and the AMD Radeon RX 9070 XT (13,543 average, 0.5% ahead). These figures place the two cards within striking distance of mid-range GPUs from multiple generations, though neither breaks into the upper echelons of performance.

It is worth remembering the GTX 660 Ti wins zero head-to-head matchups in the database, while the P106-090 wins both. However, the average benchmark scores suggest that the GTX 660 Ti is the more consistent performer across a wider range of tests, even if it loses the specific comparisons included in the head-to-head table. The P106-090's raw compute advantage in OpenCL and Vulkan does not translate into a higher average score, indicating that its performance is more workload-dependent than the GTX 660 Ti's more evenly distributed showing.

FAQ

Q: Which card wins in Geekbench OpenCL performance?

A: The NVIDIA P106-090 wins decisively with 21,304 points, beating the GTX 660 Ti's 15,113 points by 29.1%.

Q: How do the average benchmark scores compare between the two cards?

A: The GTX 660 Ti averages 15,063 points, which is higher than the P106-090's 13,470 points, despite the P106-090 winning both head-to-head tests.

Q: What is the performance gap in Vulkan workloads specifically?

A: The P106-090 leads in Geekbench Vulkan with 18,596 points, a 19.3% advantage over the GTX 660 Ti's 15,012 points.

Q: Which GPU ranks higher in the overall percentile standings?

A: The GTX 660 Ti ranks at the 57th percentile of all GPUs, three points above the P106-090's 54th percentile.

Q: How does the 3DMark Steel Nomad DX12 result affect the comparison?

A: The P106-090 scores 509 points in this test, and this low result significantly lowers its average benchmark score relative to its strong Geekbench showings.

Q: Are the two cards closely matched against their nearest competitors?

A: Yes, both sit within 1.4% of their nearest rivals. The GTX 660 Ti trails the AMD Radeon Pro 560X by just 0.1%, while the P106-090 trails the GeForce GTX 570 by 0.3%.

Architecture Differences

The architectural divide between these GPUs spans two distinct NVIDIA generations. The GTX 660 Ti uses the GK104 chip built on the Kepler architecture, fabricated on a 28 nm process at TSMC. The P106-090, by contrast, employs the GP106 chip based on the newer Pascal architecture, also produced by TSMC but on a 16 nm process node.

The transistor counts reveal an interesting inversion. The P106-090 packs 4,400 million transistors into a 200 mm² die, yielding a density of 22.0 million transistors per square millimeter. The GTX 660 Ti contains 3,540 million transistors on a larger 294 mm² die, resulting in a density of 12.0 million per square millimeter. The Pascal chip's smaller, denser design reflects the process node advantage.

The compute configuration differs substantially. The GTX 660 Ti fields 1,344 shading units, 112 texture mapping units, and 24 raster output units. The P106-090 uses only 768 shading units and 48 TMUs but compensates with 48 ROPs, double the GTX 660 Ti's count. This ROP advantage explains why the P106-090's pixel rate reaches 73.49 GPixel/s, more than 2.6 times the GTX 660 Ti's 27.44 GPixel/s. The texture rate, however, favors the GTX 660 Ti at 109.8 GTexel/s against the P106-090's 73.49 GTexel/s.

Peak FP32 performance slightly favors the GTX 660 Ti at 2.634 TFLOPS versus the P106-090's 2.352 TFLOPS. The P106-090 does include a nominal FP16 capability of 36.74 GFLOPS at a 1:64 ratio, while the GTX 660 Ti has no listed FP16 performance, confirming its older compute profile.

API support differs as well. The GTX 660 Ti supports DirectX 12 at feature level 11_0, while the P106-090 supports DirectX 12 at the higher 12_1 feature level. Both cards list OpenGL 4.6, but the Vulkan support varies: the GTX 660 Ti lists Vulkan 1.2.175, whereas the P106-090 lists Vulkan 1.4.

The P106-090 belongs to the "Mining GPUs" generation, which is reflected in its most conspicuous architectural omission: it has no display outputs. The GTX 660 Ti, by contrast, includes 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2 connectivity. The mining card's lack of video output is a defining architectural choice, trading display functionality for a lower power envelope.

Specification Differences

The two cards diverge across nearly every specification category. The memory subsystem shows the P106-090 with 3 GB of GDDR5 on a 192-bit bus, delivering 192.2 GB/s of bandwidth. The GTX 660 Ti has 2 GB of GDDR5 on the same 192-bit bus, yielding 144.2 GB/s. The higher memory clock of the P106-090, rated at 2002 MHz or 8 Gbps effective, drives this bandwidth advantage over the GTX 660 Ti's 1502 MHz or 6 Gbps effective memory.

Clock speeds heavily favor the Pascal card. The P106-090 runs at a base clock of 1354 MHz and a boost clock of 1531 MHz. The GTX 660 Ti operates at 915 MHz base and 980 MHz boost. These clocks reflect the efficiency gains of the newer architecture and process node.

Power consumption presents one of the starkest contrasts. The P106-090 draws only 75 W TDP with a single 6-pin power connector and a suggested 250 W power supply. The GTX 660 Ti consumes 150 W TDP, requires two 6-pin connectors, and calls for a 450 W power supply. The mining card's power efficiency is exceptional by comparison, halving the power draw while delivering comparable compute performance in certain workloads.

The bus interface differs dramatically. The GTX 660 Ti uses a full PCIe 3.0 x16 connection, while the P106-090 is limited to PCIe 1.0 x1. This narrow interface, combined with the lack of display outputs, clearly marks the P106-090 as a specialized compute device rather than a general-purpose graphics card. The physical dimensions are similar, with the GTX 660 Ti measuring 241 mm (9.5 inches) and the P106-090 measuring 250 mm (9.8 inches). Both are dual-slot cards.

Release timing places these products five years apart. The GTX 660 Ti launched on 2012-08-15, while the P106-090 arrived on 2017-07-30. The GTX 660 Ti had a launch MSRP of 299 USD; the P106-090 has no recorded launch price. The GTX 660 Ti belongs to the GeForce 600 generation with a predecessor in GeForce 500 and a successor in GeForce 700, while the P106-090 has no recorded predecessor or successor, existing as a standalone mining product.

The Verdict

The data paints a clear picture for different use cases. The NVIDIA P106-090 is the stronger choice for compute-heavy applications where OpenCL and Vulkan performance matters most. Its 29.1% lead in OpenCL and 19.3% lead in Vulkan over the GTX 660 Ti, combined with its 75 W power draw, make it a remarkably efficient compute solution. The 3 GB memory capacity and 192.2 GB/s bandwidth provide a solid foundation for memory-intensive workloads. However, the PCIe 1.0 x1 interface severely limits data transfer rates, and the complete absence of display outputs renders it useless for gaming or any task requiring visual output.

The GeForce GTX 660 Ti, despite its age, offers a more balanced profile. Its higher average benchmark score of 15,063 points versus the P106-090's 13,470 points, and its 57th percentile ranking versus the 54th percentile, indicate broader overall capability. The standard display outputs, full PCIe 3.0 x16 bandwidth, and higher texture rate of 109.8 GTexel/s make it a more versatile piece of hardware. Its 2.634 TFLOPS FP32 performance edges out the P106-090's 2.352 TFLOPS. The tradeoff comes in power consumption: 150 W TDP with a 450 W suggested PSU, double that of the mining card.

For gaming or general desktop use, the GTX 660 Ti is the only viable option between the two, as the P106-090 cannot drive a display at all. For dedicated compute tasks, the P106-090's raw speed and efficiency are compelling, provided the PCIe bottleneck and lack of outputs are acceptable. The GTX 660 Ti's higher percentile ranking and average score suggest it remains the stronger all-around performer, even if the P106-090 wins the specific head-to-head benchmarks recorded. Users needing a display-capable card should choose the GTX 660 Ti; those building a headless compute rig may find the P106-090's speed and low power draw appealing. The data does not support a single universal winner, only a clear division of strengths.