NVIDIA GeForce GTX 660 vs NVIDIA Quadro P5000 Comparison

FAQ

Q: How do the average benchmark scores of the GTX 660 and Quadro P5000 compare?

A: The GTX 660 has an average benchmark score of 9022, while the Quadro P5000 averages 8039. This puts the GTX 660 at the 45th percentile of all GPUs, versus the 41st percentile for the Quadro P5000, a 0.8% difference in the GTX 660's favor.

Q: Which card wins in Geekbench OpenCL performance?

A: The Quadro P5000 dominates OpenCL compute workloads with a score of 52509, which is 41162 points higher than the GTX 660's 11347. This represents a 78.4% lead for the Quadro P5000 in that specific test.

Q: Which card wins in Geekbench Vulkan performance?

A: The GTX 660 wins the Vulkan test with a score of 11415, compared to the Quadro P5000's 6342. The GTX 660 leads by 80% in this API-specific workload.

Q: What are the memory capacities and types of each card?

A: The GTX 660 uses 2 GB of GDDR5 memory on a 192-bit bus, while the Quadro P5000 uses 16 GB of GDDR5X memory on a 256-bit bus. Memory bandwidth is 144.2 GB/s for the GTX 660 and 288.5 GB/s for the Quadro P5000.

Q: What is the process node difference between the two architectures?

A: The GTX 660 is built on TSMC's 28 nm process with 2,540 million transistors on a 221 mm² die. The Quadro P5000 uses TSMC's 16 nm process, packing 7,200 million transistors into a 314 mm² die.

Q: What are the DirectX feature level differences?

A: The GTX 660 supports DirectX 12 (11_0), while the Quadro P5000 supports DirectX 12 (12_1). This gives the Quadro P5000 a higher feature level for modern graphics APIs.

Architecture Differences

The NVIDIA GeForce GTX 660 and NVIDIA Quadro P5000 represent two distinct GPU architectures from NVIDIA: Kepler and Pascal, respectively. The GTX 660 uses the GK106 chip, while the Quadro P5000 uses the GP104 chip. Both are manufactured by TSMC, but on different process nodes: the GTX 660 uses 28 nm, while the Quadro P5000 uses 16 nm. This process shrink allows the Quadro P5000 to pack significantly more transistors: 7,200 million versus 2,540 million, on a larger die (314 mm² vs 221 mm²). The transistor density also reflects this, with the Quadro P5000 at 22.9M / mm² versus 11.5M / mm² for the GTX 660.

The core configurations differ substantially. The GTX 660 has 960 shading units, 80 texture mapping units (TMUs), and 24 raster operation units (ROPs). The Quadro P5000 scales this up to 2,560 shading units, 160 TMUs, and 64 ROPs. Clock speeds are also higher on the Quadro P5000: a base clock of 1607 MHz and boost of 1733 MHz, compared to 980 MHz base and 1032 MHz boost on the GTX 660.

Memory architecture is a major differentiator. The GTX 660 uses 2 GB of GDDR5 with a 192-bit bus, delivering 144.2 GB/s bandwidth. The Quadro P5000 uses 16 GB of GDDR5X on a 256-bit bus, doubling bandwidth to 288.5 GB/s. Memory clocks are also different: the GTX 660 runs at 1502 MHz (6 Gbps effective), while the Quadro P5000 runs at 1127 MHz (9 Gbps effective). The higher effective speed of the GDDR5X memory compensates for the lower clock frequency.

Compute capabilities diverge sharply. The GTX 660 achieves 1.981 TFLOPS FP32, while the Quadro P5000 reaches 8.873 TFLOPS, a 4.5x difference. The Quadro P5000 also supports FP16 at 138.6 GFLOPS (1:64), while the GTX 660 has no FP16 capability listed. Pixel rate is 20.64 GPixel/s for the GTX 660 versus 110.9 GPixel/s for the Quadro P5000, and texture rates are 82.56 GTexel/s versus 277.3 GTexel/s.

API support shows generational advances. Both support OpenGL 4.6, but the Quadro P5000 supports Vulkan 1.4 and DirectX 12 (12_1), while the GTX 660 supports Vulkan 1.2.175 and DirectX 12 (11_0). Display outputs also differ: the GTX 660 has 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2, while the Quadro P5000 has 1x DVI and 4x DisplayPort 1.4a.

Head-to-Head Benchmarks

The recorded head-to-head data shows two tests with opposite winners, resulting in one win for each card. The most dramatic difference is in Geekbench OpenCL, where the Quadro P5000 scores 52509 versus the GTX 660's 11347. This is a 78.4% advantage for the Quadro P5000, reflecting its much larger compute core count and higher memory bandwidth. The GTX 660's score of 11347 places it far behind in raw compute throughput, which aligns with its older architecture and lower transistor count.

In Geekbench Vulkan, the results flip. The GTX 660 scores 11415, while the Quadro P5000 scores 6342. The GTX 660 leads by 80% in this test. This outcome is notable because the Quadro P5000 has a newer Vulkan implementation (1.4 versus 1.2.175). The GTX 660's higher Vulkan score suggests that its driver optimization or architecture handles this specific workload more efficiently, despite the Quadro P5000's superior raw specifications.

The average benchmark scores provide additional context. The GTX 660's average of 9022 is higher than the Quadro P5000's 8039, a difference of about 11%. However, the nearest rivals for each card are quite different. The GTX 660's closest rival is the NVIDIA TITAN V CEO Edition with an average score of 9037, a mere 0.2% difference. Other nearby rivals include the GeForce GTX 560 (9058, 0.4% ahead) and AMD Radeon 550X (8918, 1.2% behind). The Quadro P5000's nearest rival is the GeForce GTX 880M with a score of 8040, a 0% difference. The GeForce GTX 650 Ti (8053) and GTX 650 Ti Boost (8067) are also close, within 0.3% of the Quadro P5000's score.

These averages are influenced by the different benchmark suites each card was tested with. The GTX 660 has three Geekbench results, while the Quadro P5000 has ten results across multiple test types, including 3DMark Steel Nomad DX12 (1330), Passmark DirectX 10 (77), DirectX 11 (102), DirectX 12 (44), DirectX 9 (170), G2D (674), G3D (12634), and GPU Compute (6508). The inclusion of lower-scoring DirectX tests in the Quadro P5000's suite pulls its average down, while the GTX 660's limited test set skews its average higher.

Specification Differences

The following specifications differ between the two cards:

• Chip: GK106 (GTX 660) vs GP104 (Quadro P5000)
• Architecture: Kepler vs Pascal
• Generation: GeForce 600 vs Quadro Pascal (Px000)
• Process Node: 28 nm vs 16 nm
• Transistors: 2,540 million vs 7,200 million
• Die Size: 221 mm² vs 314 mm²
• Transistor Density: 11.5M / mm² vs 22.9M / mm²
• Base Clock: 980 MHz vs 1607 MHz
• Boost Clock: 1032 MHz vs 1733 MHz
• Memory Clock: 1502 MHz (6 Gbps effective) vs 1127 MHz (9 Gbps effective)
• Memory Size: 2 GB vs 16 GB
• Memory Type: GDDR5 vs GDDR5X
• Memory Bus Width: 192 bit vs 256 bit
• Memory Bandwidth: 144.2 GB/s vs 288.5 GB/s
• Shading Units: 960 vs 2560
• TMUs: 80 vs 160
• ROPs: 24 vs 64
• Pixel Rate: 20.64 GPixel/s vs 110.9 GPixel/s
• Texture Rate: 82.56 GTexel/s vs 277.3 GTexel/s
• FP32 Performance: 1.981 TFLOPS vs 8.873 TFLOPS
• FP16 Performance: Not listed vs 138.6 GFLOPS (1:64)
• TDP: 140 W vs 180 W
• Power Connectors: 1x 6-pin vs 1x 8-pin
• Suggested PSU: 300 W vs 450 W
• Display Outputs: 2x DVI, 1x HDMI 1.4a, 1x DisplayPort 1.2 vs 1x DVI, 4x DisplayPort 1.4a
• DirectX Support: 12 (11_0) vs 12 (12_1)
• Vulkan Support: 1.2.175 vs 1.4
• Dimensions: 241 mm (9.5 inches) length vs 267 mm (10.5 inches) length, 111 mm (4.4 inches) height
• Release Date: 2012-09-05 vs 2016-09-30
• Predecessor: GeForce 500 vs Quadro Maxwell
• Successor: GeForce 700 vs Quadro Volta
• Launch MSRP: 229 USD vs 2,499 USD (stated once)

Where Each One Wins

The GTX 660 wins in the Geekbench Vulkan benchmark, scoring 80% higher than the Quadro P5000. This makes it the better choice for workloads that rely on Vulkan API performance specifically, such as certain cross-platform games or compute tasks that use this API. Its shorter length (241 mm vs 267 mm) and lower power requirements (140 W TDP with a 300 W suggested PSU versus 180 W and 450 W) also make it easier to integrate into compact systems with modest power supplies. The GTX 660's average benchmark score of 9022 is higher than the Quadro P5000's 8039, and its 45th percentile ranking versus 41st indicates better overall standing in the database's GPU hierarchy.

The Quadro P5000 wins decisively in Geekbench OpenCL, with a 78.4% lead over the GTX 660. This makes it the superior choice for OpenCL compute workloads, which are common in professional applications such as video rendering, scientific simulations, and machine learning inference. Its 16 GB of GDDR5X memory is eight times larger than the GTX 660's 2 GB, and its 288.5 GB/s bandwidth is double. The Quadro P5000's 8.873 TFLOPS FP32 performance is 4.5 times higher, and its 2560 shading units provide substantial parallel compute throughput. Its higher DirectX feature level (12_1) and newer Vulkan 1.4 support also make it more future-proof for modern graphics APIs. The Quadro P5000's professional positioning is reflected in its four DisplayPort 1.4a outputs, which support high-resolution multi-monitor configurations, and its taller profile (111 mm height) suggests a more robust cooling solution.

In practical terms, the GTX 660 is the better pick for scenarios where Vulkan performance is the priority, or where system constraints on power and physical space are tight. The Quadro P5000 is the better pick for compute-intensive professional workloads, large memory footprints, and applications that benefit from newer API feature levels. The head-to-head results show a split: one win each, with the Quadro P5000's OpenCL victory being more lopsided (78.4% delta) than the GTX 660's Vulkan win (80% delta). The average benchmark scores favor the GTX 660, but this is skewed by the different test suites each card was subjected to, so the per-test results provide a more reliable comparison of their respective strengths.