NVIDIA GeForce GTX 1060 5 GB vs NVIDIA P104-100 Comparison

# Head-to-Head Benchmarks

The data presents a clear and somewhat lopsided picture when these two Pascal-era NVIDIA cards are placed side by side. Across the two shared benchmark tests, the NVIDIA P104-100 wins decisively, leaving the GeForce GTX 1060 5 GB without a single head-to-head victory. The most striking result comes from the Geekbench OpenCL test, where the P104-100 scores 52,368 against the GTX 1060 5 GB's 33,397. This represents a 36.2% advantage for the P104-100 — a substantial margin that suggests a fundamentally different level of raw compute throughput. In real-world terms, this gap is not a marginal edge but a categorical one, indicating that the P104-100 is operating in a different performance tier for OpenCL workloads.

The Vulkan benchmark tells a similar story, though with a slightly narrower margin. Here, the P104-100 posts 45,165 while the GTX 1060 5 GB manages 33,991. The delta shrinks to 24.7% in favor of the P104-100, yet the outcome remains unambiguous. While Vulkan is often viewed as a lower-overhead API that can equalize hardware differences, the data here shows the P104-100 still holds a commanding lead. It is importantly the GTX 1060 5 GB's Vulkan score (33,991) is remarkably close to its OpenCL result (33,397), differing by less than 2%. The P104-100, by contrast, shows a larger spread between its OpenCL and Vulkan scores, suggesting its architecture responds differently to these two APIs.

Looking at the broader context, the GTX 1060 5 GB's average benchmark score stands at 33,694, which places it in the 78th percentile of all GPUs. Its nearest rivals in the database include the NVIDIA RTX A5000 (33,622, just 0.2% lower) and the AMD Radeon RX 7700S (33,849, 0.5% higher). This tight clustering indicates that the GTX 1060 5 GB sits at a well-populated performance plateau. The P104-100, with an average score of 32,982, holds the 77th percentile, rubbing shoulders with the NVIDIA T600 Mobile (32,849, 0.4% lower) and the NVIDIA T550 Mobile (33,161, 0.5% higher). Despite winning both head-to-head tests, the P104-100's average score is actually slightly lower than the GTX 1060 5 GB's — a fascinating paradox that stems from the P104-100's third benchmark result (3DMark Steel Nomad DX12 at 1,413), which drags down its overall average. This demonstrates that while the P104-100 excels in compute-oriented tests, its performance is not uniformly superior across all benchmark types.

# Where Each One Wins

The performance split between these two cards is stark: the P104-100 wins both shared benchmarks, while the GTX 1060 5 GB wins none. Yet the average scores tell a more nuanced story. The GTX 1060 5 GB's higher average (33,694 versus 32,982) suggests it may hold advantages in workloads not captured by the head-to-head tests. With only two directly comparable benchmarks, the database paints an incomplete picture, but the available data points to a clear division of labor.

For compute-heavy, general-purpose workloads — as measured by Geekbench OpenCL and Vulkan — the P104-100 is the clear victor. Its 36.2% lead in OpenCL and 24.7% lead in Vulkan indicate that tasks relying on raw shader throughput, such as data-parallel processing or certain rendering pipelines, will favor the P104-100 significantly. This aligns with its specification profile: 1,920 shading units versus 1,280, and 6.655 TFLOPS FP32 performance versus 4.375 TFLOPS. The P104-100 appears purpose-built for computational grunt work, a conclusion reinforced by its "Mining GPUs" generation label.

The GTX 1060 5 GB, however, is not without its niches. Its higher average benchmark score suggests it may be better rounded, or at least more consistent, across diverse test suites. The 3DMark Steel Nomad DX12 result for the P104-100 (1,413) pulls its average down substantially, indicating a potential weakness in gaming-oriented DX12 workloads. The GTX 1060 5 GB has no equivalent 3DMark result in the database, so a direct comparison is impossible. Still, the GTX 1060 5 GB's GeForce 10-series lineage and consumer-oriented feature set — including display outputs — position it as the more versatile card for end users. If a workload involves both compute and traditional graphics output, the GTX 1060 5 GB is the only one of the two that can drive a monitor, given the P104-100 has no display outputs at all.

# Architecture Differences

Both cards are built on NVIDIA's Pascal architecture, but they use different physical chips. The GTX 1060 5 GB is based on the GP106 die, while the P104-100 employs the larger GP104 die. This is a fundamental distinction: the GP104 is a bigger, more complex chip, with 7,200 million transistors packed into a 314 mm² die. The GP106, by contrast, houses 4,400 million transistors on a 200 mm² die. The transistor density tells an interesting story — the GP104 achieves 22.9 million transistors per mm², while the GP106 manages 22.0 million per mm². This marginal difference, less than 5%, is surprising given the significant size gap, and suggests that the GP104 is not simply a scaled-up version but a chip with a more efficient layout or different balance of components.

Both are manufactured by TSMC on the same 16 nm process node, so the architectural differences are purely design-driven rather than process-driven. The GP104's larger transistor budget translates directly into more execution resources: 1,920 shading units, 120 texture mapping units, and 64 raster operations pipelines. The GP106, while still capable, offers 1,280 shading units, 80 TMUs, and 40 ROPs. This 50% advantage in shading units and TMUs, along with a 60% advantage in ROPs, explains the P104-100's dominant benchmark performance.

Memory architecture differs significantly as well. The GTX 1060 5 GB uses 5 GB of GDDR5 on a 160-bit bus, while the P104-100 uses 4 GB of GDDR5X on a 256-bit bus. The P104-100's memory configuration yields 320.3 GB/s of bandwidth — exactly double the GTX 1060 5 GB's 160.2 GB/s. The GDDR5X memory runs at 10 Gbps effective, versus 8 Gbps for the GTX 1060 5 GB's GDDR5. Interestingly, the P104-100 has less total memory (4 GB versus 5 GB), but its wider bus and faster memory technology more than compensate in bandwidth terms. This is a classic trade-off: more capacity versus more speed. For workloads that are bandwidth-bound, the P104-100 holds a clear edge; for those that require large working sets, the GTX 1060 5 GB's extra gigabyte could prove valuable.

Feature-wise, both cards support DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. Neither has ray tracing or tensor cores, as these are pre-RTX architectures. The GTX 1060 5 GB includes display outputs (1x DVI, 1x HDMI 2.0, 3x DisplayPort 1.4a), while the P104-100 has none — a clear signal of its intended use case as a compute or mining card rather than a consumer graphics solution.

# Specification Differences

The P104-100 outperforms the GTX 1060 5 GB in nearly every raw specification category, with one notable exception. Clock speeds favor the P104-100: its base clock runs at 1,607 MHz versus 1,506 MHz, and its boost clock reaches 1,733 MHz versus 1,709 MHz. While the boost clocks are close, the base clock gap of over 100 MHz gives the P104-100 a head start in sustained workloads.

Memory differences are substantial. The GTX 1060 5 GB offers 5 GB of GDDR5 on a 160-bit bus, while the P104-100 provides 4 GB of GDDR5X on a 256-bit bus. Memory bandwidth heavily favors the P104-100: 320.3 GB/s versus 160.2 GB/s. The P104-100 also runs faster memory at 10 Gbps effective, compared to 8 Gbps for the GTX 1060 5 GB.

Compute resources are where the P104-100 truly flexes. It has 1,920 shading units versus 1,280, 120 TMUs versus 80, and 64 ROPs versus 40. This translates to 6.655 TFLOPS FP32 performance versus 4.375 TFLOPS, a 52% advantage. Pixel and texture rates follow suit: 110.9 GPixel/s versus 68.36 GPixel/s, and 208.0 GTexel/s versus 136.7 GTexel/s, respectively.

Power and connectivity present a mixed picture. The GTX 1060 5 GB has a listed TDP of 120 W and requires a 300 W power supply, while the P104-100 has no TDP listed but only suggests a 200 W PSU. The GTX 1060 5 GB uses a single 6-pin power connector; the P104-100 needs a single 8-pin. Physically, the P104-100 is longer at 267 mm (10.5 inches) versus 250 mm (9.8 inches) for the GTX 1060 5 GB. Both are dual-slot cards. The bus interface differs dramatically: the GTX 1060 5 GB uses PCIe 3.0 x16, while the P104-100 is limited to PCIe 1.0 x4 — a potential bottleneck for data transfer despite its compute prowess.

# FAQ

Q: Which card has higher raw compute performance?

A: The P104-100 is far ahead, with 6.655 TFLOPS FP32 versus 4.375 TFLOPS for the GTX 1060 5 GB, and 1,920 shading units versus 1,280.

Q: Why does the P104-100 have a lower average benchmark score despite winning both head-to-head tests?

A: The P104-100's average (32,982) is dragged down by its 3DMark Steel Nomad DX12 score of 1,413, a test the GTX 1060 5 GB does not have a result for. Its two Geekbench wins (52,368 OpenCL and 45,165 Vulkan) are not enough to offset that low 3DMark score.

Q: Which card has more memory bandwidth?

A: The P104-100, with 320.3 GB/s on a 256-bit GDDR5X interface, exactly double the GTX 1060 5 GB's 160.2 GB/s from a 160-bit GDDR5 bus.

Q: Can the P104-100 connect to a display?

A: No. The P104-100 has no display outputs, while the GTX 1060 5 GB offers 1x DVI, 1x HDMI 2.0, and 3x DisplayPort 1.4a.

Q: Do these cards support modern graphics APIs?

A: Yes, both support DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. Neither supports ray tracing or tensor cores.

Q: How do these cards compare to their nearest rivals in the database?

A: The GTX 1060 5 GB sits within 0.9% of the AMD Radeon RX 480's average score (33,997), while the P104-100 is within 0.7% of the AMD Radeon Pro 570 (33,207). Both hold similar percentile rankings at 78 and 77, respectively.

# The Verdict

The data supports a clear but nuanced conclusion. For compute-intensive tasks that leverage OpenCL or Vulkan, the P104-100 is the superior choice. Its 36.2% lead in OpenCL and 24.7% lead in Vulkan, backed by 52% more shading units and double the memory bandwidth, make it a formidable compute engine. Users running GPU-accelerated workloads that fit within 4 GB of memory will find the P104-100 significantly faster.

However, the GTX 1060 5 GB is the more balanced and practical option. Its higher average benchmark score (33,694 versus 32,982) and higher percentile ranking (78 versus 77) suggest better all-around consistency. It offers 5 GB of memory — 25% more capacity than the P104-100 — which could be crucial for datasets that exceed 4 GB. Most importantly, it has display outputs, making it a functional graphics card for everyday use. The GTX 1060 5 GB also uses a standard PCIe 3.0 x16 interface, while the P104-100 is hobbled by PCIe 1.0 x4, which could severely limit data transfer in real-world systems.

The P104-100's lack of display outputs and its mining-oriented lineage make it a poor choice for general-purpose computing. It is a specialized tool for a specific job. The GTX 1060 5 GB, despite losing every head-to-head benchmark, is the card that most users should choose — it can render graphics, connect to monitors, and still deliver respectable compute performance. The P104-100 is for those who need maximum compute throughput and have no need for video output. The data is clear: if you want a graphics card, pick the GTX 1060 5 GB. If you want a compute accelerator, the P104-100 is the winner.