NVIDIA GeForce GTX 660 vs NVIDIA Tesla C2075 Comparison

The NVIDIA Tesla C2075 and the NVIDIA GeForce GTX 660 come from different eras and target completely different workloads. The Tesla C2075 is a compute-oriented card built on the Fermi architecture, while the GTX 660 is a consumer gaming card using the Kepler architecture. Benchmark data in the database shows the GTX 660 holds a single head-to-head win in the only shared test, but the two cards diverge sharply in memory capacity, power draw, and feature support. This analysis breaks down where each card wins, how their architectures differ, and which user should pick which based strictly on recorded measurements.

Where Each One Wins

The GTX 660 wins the only direct benchmark comparison available in the database. In the Geekbench OpenCL test, the GTX 660 scores 11,347 against the Tesla C2075’s 10,400. That is a performance advantage of 8.3% for the GTX 660 in a compute workload. The database records one win for the GTX 660 and zero wins for the Tesla C2075 in head-to-head testing. If the question is purely about raw OpenCL compute speed, the GTX 660 is the faster card.

However, the Tesla C2075 wins in categories not covered by that single test. It carries 6 GB of GDDR5 memory, three times the GTX 660’s 2 GB. For workloads that need large datasets resident on the GPU, such as scientific simulations or large-matrix processing, that capacity advantage matters. The Tesla also uses a 384-bit memory bus versus the GTX 660’s 192-bit bus, and its memory bandwidth of 150.3 GB/s edges out the GTX 660’s 144.2 GB/s. So while the GTX 660 wins on raw compute score, the Tesla C2075 wins on memory capacity and slightly higher bandwidth.

The GTX 660 also wins on pixel throughput and texture throughput. It records a pixel rate of 20.64 GPixel/s versus the Tesla’s 16.07 GPixel/s, and a texture rate of 82.56 GTexel/s versus the Tesla’s 32.14 GTexel/s. For graphics rendering workloads, the GTX 660 is clearly superior. The Tesla C2075 has 48 ROPs against the GTX 660’s 24, but the GTX 660’s higher clock speeds compensate in practice. The GTX 660 also has 960 shading units versus the Tesla’s 448, giving it more parallel compute lanes despite its lower transistor count.

The Tesla C2075 wins on power connector requirements in a different sense: it is a more power-hungry card. Its TDP is 247 W, and it requires a 6-pin and an 8-pin power connector. The GTX 660 has a 140 W TDP and needs only a single 6-pin connector. The suggested power supply for the Tesla is 550 W, while the GTX 660 suggests 300 W. For system integration, the GTX 660 is far easier to accommodate.

Architecture Differences

The Tesla C2075 uses the GF110 chip on the Fermi 2.0 architecture, manufactured on a 40 nm process at TSMC. The GTX 660 uses the GK106 chip on the Kepler architecture, also from TSMC but on a 28 nm process. The process shrink from 40 nm to 28 nm is a significant generational leap. The Tesla packs 3,000 million transistors into a 520 mm² die, giving a transistor density of 5.8 million per square millimeter. The GTX 660 has 2,540 million transistors on a 221 mm² die, yielding a density of 11.5 million per square millimeter. The Kepler architecture achieves nearly double the density per area, which explains how the GTX 660 delivers higher compute throughput with fewer transistors.

Clock speeds differ dramatically. The Tesla C2075 has no base or boost clock listed in the database, but its memory clock is 783 MHz with 3.1 Gbps effective. The GTX 660 has a base clock of 980 MHz and a boost clock of 1,032 MHz, with a memory clock of 1,502 MHz and 6 Gbps effective. The GTX 660’s memory runs at nearly double the effective speed.

Shading unit counts also diverge. The Tesla has 448 shading units, 56 texture mapping units, and 48 ROPs. The GTX 660 has 960 shading units, 80 TMUs, and 24 ROPs. The GTX 660 has more than double the shading units and more TMUs, but half the ROPs. This reflects different design priorities: Kepler emphasizes compute throughput, while Fermi focuses on memory bandwidth and rasterization.

API support differs as well. Both cards support DirectX 12 (11_0) and OpenGL 4.6, but the GTX 660 adds Vulkan 1.2.175 support, while the Tesla C2075 has no Vulkan listing. The GTX 660 also supports PCIe 3.0 x16, while the Tesla uses PCIe 2.0 x16. The GTX 660’s display outputs include 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2, while the Tesla has only a single DVI output. The Tesla is clearly not designed for display use, despite its compute capabilities.

Head-to-Head Benchmarks

The only recorded head-to-head benchmark is the Geekbench OpenCL test. The Tesla C2075 scores 10,400, and the GTX 660 scores 11,347. The database marks the GTX 660 as the winner with a delta of -8.3%, meaning the Tesla trails by that margin. This is a notable gap for a compute-oriented card, especially considering the Tesla’s larger memory pool and higher bandwidth. The GTX 660’s advantage likely stems from its higher clock speeds, more shading units, and more efficient Kepler architecture.

In the broader database, the Tesla C2075 sits at the 48th percentile of all GPUs, with an average benchmark score of 10,400. Its nearest rivals include the AMD Radeon RX 6500M with an average score of 10,362 (the Tesla leads by 0.4%), the AMD Radeon RX 550X at 10,481 (the Tesla trails by 0.8%), the NVIDIA GeForce GTX 950A at 10,273 (the Tesla leads by 1.2%), and the AMD Radeon R9 M275X at 10,582 (the Tesla trails by 1.7%). This places the Tesla in a narrow performance band where small score differences separate cards.

The GTX 660 sits at the 45th percentile, with an average benchmark score of 9,022. That average includes its OpenCL score of 11,347, a Metal score of 4,305, and a Vulkan score of 11,415. The Vulkan score is its highest recorded result. Its nearest rivals include the NVIDIA TITAN V CEO Edition at 9,037 (the GTX 660 trails by 0.2%), the NVIDIA GeForce GTX 560 at 9,058 (the GTX 660 trails by 0.4%), the AMD Radeon 550X at 8,918 (the GTX 660 leads by 1.2%), and the AMD Radeon Pro WX 5100 at 8,863 (the GTX 660 leads by 1.8%). The GTX 660’s average is dragged down by its weak Metal score, but its OpenCL and Vulkan results are strong.

The data shows a clear split: the GTX 660 outperforms the Tesla in OpenCL compute, and it also has a Vulkan score that the Tesla cannot match. The Tesla has no Vulkan score recorded, so any Vulkan workload is automatically a win for the GTX 660. The Tesla’s only compute metric is OpenCL, and it loses there.

The Verdict

The GTX 660 is the better choice for users who need raw compute performance in OpenCL or Vulkan workloads, or who need a card for graphics rendering. Its higher shading unit count, faster clocks, and modern architecture deliver a 8.3% OpenCL win and add Vulkan support. It also draws 107 W less power, requires a smaller power supply, and fits into more systems with a single 6-pin connector. For gaming or general GPU compute, the GTX 660 is the clear winner.

The Tesla C2075 is the better choice only for users who need more than 2 GB of memory. Its 6 GB capacity is triple the GTX 660’s, and its 384-bit bus provides 150.3 GB/s of bandwidth, slightly higher than the GTX 660’s 144.2 GB/s. That memory advantage is useful for datasets that exceed 2 GB, such as large scientific models or high-resolution textures in professional applications. However, the Tesla sacrifices performance, efficiency, and modern API support to get that memory. Its 247 W TDP and dual-slot power connectors make it harder to integrate.

The database records the GTX 660 as the winner in every shared benchmark and gives it the higher peak compute scores. The Tesla C2075’s niche is memory capacity, not speed. If the workload fits within 2 GB, the GTX 660 is faster and more efficient. If the workload needs more than 2 GB, the Tesla is the only option between these two, but users should expect lower compute throughput.

FAQ

Q: Which card has the higher OpenCL benchmark score?

A: The GTX 660 scores 11,347 in Geekbench OpenCL, while the Tesla C2075 scores 10,400. The GTX 660 leads by 8.3%.

Q: Does the Tesla C2075 support Vulkan?

A: No. The database lists no Vulkan support for the Tesla C2075. The GTX 660 supports Vulkan 1.2.175.

Q: How much memory does each card have?

A: The Tesla C2075 has 6 GB of GDDR5 memory. The GTX 660 has 2 GB of GDDR5 memory.

Q: Which card draws less power?

A: The GTX 660 has a 140 W TDP and requires a 300 W power supply. The Tesla C2075 has a 247 W TDP and requires a 550 W power supply.

Q: What is the pixel rate difference between the two?

A: The GTX 660 has a pixel rate of 20.64 GPixel/s, while the Tesla C2075 has a pixel rate of 16.07 GPixel/s. The GTX 660 is faster.

Q: How do the cards compare in the database percentile rankings?

A: The Tesla C2075 sits at the 48th percentile of all GPUs, while the GTX 660 sits at the 45th percentile. The Tesla’s average score is 10,400, and the GTX 660’s average score is 9,022, due to the GTX 660’s low Metal score.

Specification Differences

The two cards differ in nearly every core specification. The Tesla C2075 uses the GF110 chip on Fermi 2.0, while the GTX 660 uses the GK106 chip on Kepler. The process node is 40 nm for the Tesla and 28 nm for the GTX 660. The Tesla has 3,000 million transistors on a 520 mm² die, while the GTX 660 has 2,540 million transistors on a 221 mm² die. Transistor density is 5.8M per mm² for the Tesla and 11.5M per mm² for the GTX 660.

Memory configuration differs significantly. The Tesla has 6 GB at 783 MHz with 3.1 Gbps effective, a 384-bit bus, and 150.3 GB/s bandwidth. The GTX 660 has 2 GB at 1,502 MHz with 6 Gbps effective, a 192-bit bus, and 144.2 GB/s bandwidth. The Tesla’s memory clock is lower, but its wider bus gives it slightly higher bandwidth.

Compute resources differ as well. The Tesla has 448 shading units, 56 TMUs, and 48 ROPs. The GTX 660 has 960 shading units, 80 TMUs, and 24 ROPs. Pixel rate is 16.07 GPixel/s for the Tesla and 20.64 GPixel/s for the GTX 660. Texture rate is 32.14 GTexel/s for the Tesla and 82.56 GTexel/s for the GTX 660. FP32 performance is 1,027.7 GFLOPS for the Tesla and 1.981 TFLOPS for the GTX 660.

Power and connectivity also differ. The Tesla has a 247 W TDP, dual-slot width, and requires 1x 6-pin plus 1x 8-pin power connectors. The GTX 660 has a 140 W TDP, dual-slot width, and requires only 1x 6-pin. Suggested PSU is 550 W for the Tesla and 300 W for the GTX 660. The Tesla uses PCIe 2.0 x16, while the GTX 660 uses PCIe 3.0 x16. Display outputs are 1x DVI for the Tesla and 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2 for the GTX 660.

API support shows the GTX 660 ahead. Both support DirectX 12 (11_0) and OpenGL 4.6, but only the GTX 660 has Vulkan 1.2.175. The Tesla has no Vulkan listing. Physical dimensions are close: the Tesla is 248 mm (9.8 inches) long, and the GTX 660 is 241 mm (9.5 inches) long. Both are end-of-life products, with the Tesla released in 2011 and the GTX 660 in 2012. The GTX 660 had a launch MSRP of 229 USD.