NVIDIA Quadro P6000

The NVIDIA Quadro P6000 is a professional workstation graphics card built on the Pascal architecture, utilizing the GP102 chip manufactured on a 16 nm process at TSMC. With a transistor count of 11,800 million on a 471 mm² die, this dual-slot card was released in late 2016 and is now end-of-life, succeeded by the Quadro Volta generation. The benchmark data positions it at the 92nd percentile of all GPUs, with an average benchmark score of 67,320 across Geekbench OpenCL and Vulkan tests, placing it in a tightly contested performance tier where rival cards are separated by less than two percent.

How It Compares

The closest rival is the AMD Radeon Pro Vega 56, which posts an average score of 67,097. The Quadro P6000 leads by a marginal 0.3%, a difference that falls well within normal run-to-run variance for synthetic benchmarks. This effectively makes the two cards performance equals in aggregate workloads, though the NVIDIA card achieves this with a distinctly different memory configuration and architecture.

Against the NVIDIA Tesla T4, the P6000 holds a 0.9% advantage, with the Tesla T4 scoring 66,733. This is a notable result because the Tesla T4 is a datacenter-oriented card with a much lower power envelope, yet the P6000’s raw compute throughput only barely edges it out. The data suggests that in professional applications, the P6000’s higher shading unit count compensates for architectural differences.

The GeForce RTX 4090, a consumer flagship from a much later generation, scores 66,473, putting it 1.3% behind the Quadro P6000. This is a striking outcome, as the RTX 4090 benefits from several generations of architectural improvements, yet the P6000’s driver-optimized workstation performance and 24 GB frame buffer keep it competitive in these specific benchmark tests.

The Tesla P40, another Pascal-based professional card, scores 66,127, trailing the P6000 by 1.8%. Both cards share the GP102 chip, but the P6000’s higher clock speeds and memory bandwidth give it a clear, if modest, edge. The delta is small enough that workload-specific factors could easily reverse the order, but the aggregate data consistently favors the P6000.

Power and Cooling

The Quadro P6000 carries a thermal design power (TDP) of 250 W, which is substantial for a professional card of its era. The cooling solution is a dual-slot design, indicating that it requires two expansion slots in the chassis and moves heat through a robust heatsink and fan assembly. Power delivery is handled by a single 8-pin PCIe power connector, which is a relatively modest requirement for a card with this performance level, simplifying installation in workstations with standard power supplies.

NVIDIA recommends a 600 W power supply unit for systems using the P6000. This recommendation accounts for the card’s 250 W TDP alongside the rest of the system’s components, including the CPU, storage, and peripherals. The single 8-pin connector means no adapter cables are necessary for most modern power supplies, though older units may need to verify they have an available 8-pin PCIe lead. The bus interface is PCIe 3.0 x16, which is fully compatible with virtually all motherboards from the card’s production period and later, provided the physical slot is available.

Benchmark Performance

The Geekbench OpenCL score for the Quadro P6000 is 63,852, while its Vulkan score is 70,788, yielding the average of 67,320 used for comparisons. The gap between OpenCL and Vulkan performance suggests that the card’s compute capabilities are well-optimized for modern graphics APIs, with Vulkan delivering roughly 10.9% higher scores than OpenCL in this test suite. This indicates strong driver support for low-level API access, which benefits professional visualization workloads.

Comparing the average score to the nearest rivals, the P6000 leads the Radeon Pro Vega 56 by 0.3%, a difference of only 223 points. This is a negligible margin that could be attributed to driver versions or test conditions, but it does demonstrate that the P6000 holds its own against a competitor with a newer architecture. The advantage over the Tesla T4 is 0.9%, or 587 points, which is more consistent but still within a narrow band. Against the RTX 4090, the P6000’s 1.3% lead translates to 847 points, a surprising result given the generational gap, but one that reflects the P6000’s workstation-tuned drivers and the specific nature of these benchmark workloads.

The largest delta is against the Tesla P40, where the P6000 leads by 1.8%, or 1,193 points. This is the clearest separation among the rivals, suggesting that the P6000’s higher boost clock of 1645 MHz versus the P40’s configuration provides a measurable advantage in compute-heavy tasks. The fp32 throughput of 12.63 TFLOPS is a key contributor here, as both cards share the same architecture but the P6000 extracts more performance per watt from the GP102 die.

FAQ

Q: What is the launch MSRP of the NVIDIA Quadro P6000?

A: The launch MSRP was 5,999 USD.

Q: How much memory does the Quadro P6000 have and what type is it?

A: It has 24 GB of GDDR5X memory on a 384-bit bus, providing 432.8 GB/s of bandwidth.

Q: Does the Quadro P6000 support hardware ray tracing?

A: No, the card has no RT cores listed in the specifications, indicating it lacks dedicated ray tracing hardware.

Q: What API versions does the Quadro P6000 support?

A: It supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4.

Q: What is the recommended power supply wattage for a system with this card?

A: NVIDIA recommends a 600 W power supply for systems using the Quadro P6000.

Q: How does the Quadro P6000 compare to the GeForce RTX 4090 in average benchmark score?

A: The Quadro P6000 has an average score of 67,320, which is 1.3% higher than the RTX 4090’s 66,473.

Ray Tracing and Feature Set

The Quadro P6000 does not include dedicated ray tracing cores or tensor cores, as these are absent from the specification list. This places it in the pre-RTX era of NVIDIA professional graphics, where ray tracing was handled through compute shaders rather than specialized hardware. The card does support DirectX 12 with feature level 12_1, which includes some advanced rendering features but not the DXR ray tracing API that requires hardware acceleration. OpenGL 4.6 and Vulkan 1.4 are supported, giving users access to modern graphics pipelines for visualization and compute tasks.

The lack of tensor cores means no dedicated AI acceleration for tasks like denoising or deep learning inference. However, the card’s 3840 shading units and 12.63 TFLOPS of fp32 performance provide substantial general-purpose compute capability. The fp16 throughput is notably limited at 197.4 GFLOPS, operating at a 1:64 ratio relative to fp32, which means half-precision workloads will not see performance benefits. This is a significant limitation for workloads that rely on fp16, such as certain machine learning training scenarios, but it does not affect traditional graphics rendering or fp32-based scientific computing.

Who Should Consider It

The Quadro P6000 is best suited for professionals who need large amounts of memory and reliable compute performance at high resolutions. With 24 GB of GDDR5X memory and 432.8 GB/s of bandwidth, it can handle textures and datasets that exceed the capacity of most consumer cards, making it suitable for 4K and 8K content creation, complex CAD models, and scientific visualization. The pixel rate of 157.9 GPixel/s and texture rate of 394.8 GTexel/s indicate strong fill-rate capabilities for multi-display setups, and the four DisplayPort 1.4a outputs plus one DVI port support high-resolution multi-monitor configurations.

For users working in OpenCL-heavy applications, the P6000’s score of 63,852 demonstrates solid compute performance, while the Vulkan score of 70,788 suggests even better results in modern graphics APIs. The card’s 0.3% lead over the Radeon Pro Vega 56 means it is a viable alternative for AMD-centric workflows, while the 1.3% advantage over the RTX 4090 indicates that the P6000 remains relevant for specific professional tasks despite its age. However, given its end-of-life status, this card is primarily for legacy systems or specialized applications where 24 GB of memory is a hard requirement and newer cards with similar memory capacity are unavailable or cost-prohibitive. Users with workloads that rely on fp16 or hardware ray tracing will need to look to newer generations, as the P6000’s feature set is firmly rooted in the Pascal era.