NVIDIA GeForce GTX 660 vs NVIDIA Quadro P4000 Comparison

The NVIDIA Quadro P4000 and the NVIDIA GeForce GTX 660 are separated by nearly five years of GPU evolution, and the benchmark data reflects a generational chasm. The Quadro P4000 is the unequivocal performance winner, but the GTX 660 holds a distinct position as a legacy product with a much lower launch MSRP. The data shows a clear hierarchy, with the Quadro P4000 delivering more than triple the compute performance in shared workloads.

Head-to-Head Benchmarks

The two cards share only two benchmark tests in the database, and the Quadro P4000 wins both by enormous margins. In Geekbench OpenCL, the Quadro P4000 scores 36,212 against the GTX 660’s 11,347, a 219.1% advantage. That is a decisive, multi-generational leap in raw compute throughput. In Geekbench Vulkan, the gap widens further: the Quadro P4000 posts 41,786 while the GTX 660 manages 11,415, resulting in a 266.1% delta. These are not marginal improvements; they indicate a completely different performance class.

The average benchmark scores reinforce this hierarchy. The Quadro P4000 averages 9,665 across all recorded tests, while the GTX 660 averages 9,022. While the gap in average scores is a more modest 7.1%, the head-to-head results show that in shared workloads, the Quadro P4000 is overwhelmingly faster. The GTX 660’s average is buoyed by its fewer, less demanding tests, whereas the Quadro P4000’s score is dragged down by its poor DirectX 9 and DirectX 12 results (181 and 40, respectively). In the tests where both cards are present, there is no contest.

The Quadro P4000’s wins are not just in raw numbers. Its Geekbench Vulkan score of 41,786 is 266.1% higher than the GTX 660’s, which means that for any modern API workload, the Pascal card delivers roughly 3.7 times the performance. Even in OpenCL, a compute workload where the GTX 660’s Kepler architecture was once competitive, the Quadro P4000’s 219.1% lead shows that the older card is simply outclassed. The data does not show a single test where the GTX 660 closes the gap to within a competitive range.

Architecture Differences

The architectural chasm between these two cards is stark. The Quadro P4000 is built on the Pascal architecture using the GP104 chip, fabricated on a 16 nm process at TSMC. The GTX 660 uses the Kepler architecture with the GK106 chip, on a 28 nm process. This process shrink alone accounts for significant efficiency and density gains. The Quadro P4000 packs 7,200 million transistors into a 314 mm² die, yielding a transistor density of 22.9M per mm². The GTX 660 has just 2,540 million transistors on a 221 mm² die, with a density of 11.5M per mm². The Pascal chip has nearly three times the transistor count and double the density.

Core configuration differences are equally pronounced. The Quadro P4000 has 1,792 shading units, 112 texture mapping units, and 64 ROPs. The GTX 660 has 960 shading units, 80 TMUs, and only 24 ROPs. This means the Quadro P4000 has 86.7% more shading units, 40% more TMUs, and 166.7% more ROPs. The pixel rate tells the story: the Quadro P4000 achieves 94.72 GPixel/s versus the GTX 660’s 20.64 GPixel/s. Texture rate is similarly lopsided at 165.8 GTexel/s versus 82.56 GTexel/s.

Memory architecture is another major differentiator. The Quadro P4000 has 8 GB of GDDR5 on a 256-bit bus, with 243.3 GB/s of bandwidth. The GTX 660 has 2 GB of GDDR5 on a 192-bit bus, with 144.2 GB/s of bandwidth. The Quadro P4000 offers quadruple the capacity and 68.7% more bandwidth. Clock speeds also favor the newer card: 1202 MHz base and 1480 MHz boost for the Quadro P4000, versus 980 MHz base and 1032 MHz boost for the GTX 660. Memory clocks are 1901 MHz (7.6 Gbps effective) versus 1502 MHz (6 Gbps effective). The FP32 compute output is 5.304 TFLOPS for the Quadro P4000, more than 2.6 times the GTX 660’s 1.981 TFLOPS. The Quadro P4000 also supports FP16 at 82.88 GFLOPS (1:64 ratio); the GTX 660 has no FP16 capability listed.

Power and physical specs differ too. The Quadro P4000 has a 105 W TDP and is single-slot, while the GTX 660 has a 140 W TDP and is dual-slot. Both use a single 6-pin power connector and suggest a 300 W PSU. Both are PCIe 3.0 x16. Display outputs differ: the Quadro P4000 offers 4x DisplayPort 1.4a, while the GTX 660 offers 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2. API support shows the Quadro P4000 with DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The GTX 660 has DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175.

FAQ

Q: Which card has higher raw compute performance in shared benchmarks?

A: The Quadro P4000. In Geekbench OpenCL, it scores 36,212 versus 11,347 for the GTX 660, a 219.1% lead. In Geekbench Vulkan, it scores 41,786 versus 11,415, a 266.1% lead.

Q: What is the memory capacity difference?

A: The Quadro P4000 has 8 GB of GDDR5 on a 256-bit bus. The GTX 660 has 2 GB of GDDR5 on a 192-bit bus. The Quadro P4000 offers four times the capacity and 68.7% more bandwidth (243.3 GB/s vs 144.2 GB/s).

Q: How do the architectures differ in terms of process node?

A: The Quadro P4000 uses the Pascal architecture on a 16 nm TSMC process. The GTX 660 uses the Kepler architecture on a 28 nm TSMC process. The Pascal chip has 7,200 million transistors versus 2,540 million for Kepler.

Q: What is the TDP difference?

A: The Quadro P4000 has a 105 W TDP and is single-slot. The GTX 660 has a 140 W TDP and is dual-slot. Both use a single 6-pin power connector and recommend a 300 W PSU.

Q: Which card has better API support?

A: The Quadro P4000 supports DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4. The GTX 660 supports DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175.

Q: What are the average benchmark scores for each card?

A: The Quadro P4000 has an average benchmark score of 9,665 across all tests. The GTX 660 has an average score of 9,022. The Quadro P4000’s average is higher, but the head-to-head delta is much larger in individual shared tests.

The Verdict

The data is unambiguous: the Quadro P4000 is the superior performer in every shared metric. It wins 2 out of 2 head-to-head benchmarks, with deltas of 219.1% and 266.1%. Its average benchmark score of 9,665 exceeds the GTX 660’s 9,022, placing it in the 47th percentile of all GPUs versus the GTX 660’s 45th percentile. The Quadro P4000’s nearest rivals are the AMD Radeon Pro WX 2100 (0.1% faster) and the NVIDIA GeForce GTX 960M (0.2% faster), while the GTX 660’s nearest rivals include the NVIDIA TITAN V CEO Edition (0.2% faster) and the GeForce GTX 560 (0.4% faster). This shows that the Quadro P4000 competes in a slightly higher performance tier, despite both cards being near the 45-47th percentile overall.

For users choosing between these two, the Quadro P4000 is the only rational pick for modern workloads. Its 8 GB memory, 5.304 TFLOPS FP32, and newer API support (Vulkan 1.4 vs 1.2.175) make it suitable for current applications. The GTX 660, with its 2 GB memory and 1.981 TFLOPS, is a legacy product that will struggle with memory-intensive tasks. The Quadro P4000’s launch MSRP is 815 USD, while the GTX 660 launched at 229 USD, but the performance delta justifies the price difference for professional use. The GTX 660’s only advantages are its lower launch MSRP and its Geekbench Metal score of 4,305, a test the Quadro P4000 does not have recorded.

Specification Differences

• Process Node: Quadro P4000 is 16 nm; GTX 660 is 28 nm.
• Transistors: Quadro P4000 has 7,200 million; GTX 660 has 2,540 million.
• Die Size: Quadro P4000 is 314 mm²; GTX 660 is 221 mm².
• Transistor Density: Quadro P4000 is 22.9M / mm²; GTX 660 is 11.5M / mm².
• Base Clock: Quadro P4000 is 1202 MHz; GTX 660 is 980 MHz.
• Boost Clock: Quadro P4000 is 1480 MHz; GTX 660 is 1032 MHz.
• Memory Clock: Quadro P4000 is 1901 MHz (7.6 Gbps); GTX 660 is 1502 MHz (6 Gbps).
• Memory Size: Quadro P4000 is 8 GB; GTX 660 is 2 GB.
• Memory Bus: Quadro P4000 is 256 bit; GTX 660 is 192 bit.
• Memory Bandwidth: Quadro P4000 is 243.3 GB/s; GTX 660 is 144.2 GB/s.
• Shading Units: Quadro P4000 has 1792; GTX 660 has 960.
• TMUs: Quadro P4000 has 112; GTX 660 has 80.
• ROPs: Quadro P4000 has 64; GTX 660 has 24.
• Pixel Rate: Quadro P4000 is 94.72 GPixel/s; GTX 660 is 20.64 GPixel/s.
• Texture Rate: Quadro P4000 is 165.8 GTexel/s; GTX 660 is 82.56 GTexel/s.
• FP32: Quadro P4000 is 5.304 TFLOPS; GTX 660 is 1.981 TFLOPS.
• FP16: Quadro P4000 is 82.88 GFLOPS; GTX 660 is not listed.
• TDP: Quadro P4000 is 105 W; GTX 660 is 140 W.
• Slot Width: Quadro P4000 is single-slot; GTX 660 is dual-slot.
• Display Outputs: Quadro P4000 is 4x DisplayPort 1.4a; GTX 660 is 2x DVI, 1x HDMI 1.4a, 1x DisplayPort 1.2.
• DirectX Support: Quadro P4000 is 12 (12_1); GTX 660 is 12 (11_0).
• Vulkan Support: Quadro P4000 is 1.4; GTX 660 is 1.2.175.
• Launch MSRP: Quadro P4000 is 815 USD; GTX 660 is 229 USD.

Where Each One Wins

The Quadro P4000 wins in every category where both cards have data. It dominates in Geekbench OpenCL and Vulkan, with leads of 219.1% and 266.1% respectively. It also wins on compute capability with 5.304 TFLOPS FP32 versus 1.981 TFLOPS. The memory advantage is decisive: 8 GB with 243.3 GB/s bandwidth versus 2 GB with 144.2 GB/s. The Quadro P4000 wins on pixel rate (94.72 GPixel/s vs 20.64 GPixel/s) and texture rate (165.8 GTexel/s vs 82.56 GTexel/s). It wins on efficiency with a 105 W TDP versus 140 W, and on form factor as a single-slot card versus dual-slot. It wins on modern API support with Vulkan 1.4 versus 1.2.175 and DirectX 12_1 versus 11_0. It wins on display outputs with 4x DisplayPort 1.4a versus the older 2x DVI/HDMI/DisplayPort 1.2 combination.

The GTX 660 has no head-to-head benchmark wins. Its only recorded advantage is a Geekbench Metal score of 4,305, a test that the Quadro P4000 does not have a recorded result for. It also has a lower launch MSRP of 229 USD versus 815 USD, which is its primary appeal for legacy or budget-constrained builds. For any workload involving compute, modern APIs, high-resolution textures, or multi-display setups, the Quadro P4000 is the clear choice. The GTX 660 remains relevant only for users who need a basic display adapter with no performance expectations beyond its 1.981 TFLOPS and 2 GB memory, and who prioritize the lower launch MSRP over all other factors.