NVIDIA GeForce GTX 480 vs NVIDIA GeForce GTX 660 Ti Comparison

# NVIDIA GeForce GTX 660 Ti vs NVIDIA GeForce GTX 480

The GeForce GTX 660 Ti and the GeForce GTX 480 represent two distinct approaches from NVIDIA across consecutive generations. The GTX 660 Ti belongs to the GeForce 600 series, built on the Kepler architecture with the GK104 chip, while the GTX 480 is part of the GeForce 400 series, using the Fermi architecture and GF100 chip. The database shows a clear generational shift in process technology, memory configuration, and raw compute throughput. The GTX 660 Ti carries a 28 nm process node from TSMC, whereas the GTX 480 uses a 40 nm node. The transistor counts differ substantially, with the GTX 660 Ti integrating 3,540 million transistors on a 294 mm² die, compared to the GTX 480's 3,100 million transistors on a 529 mm² die. Transistor density tells the story: 12.0 million transistors per mm² for the Kepler card versus 5.9 million per mm² for the Fermi card. The GTX 660 Ti is the denser, more modern design.

FAQ

Q: What is the average benchmark score of the NVIDIA GeForce GTX 660 Ti?

A: The average benchmark score, based on Geekbench OpenCL and Vulkan results recorded in the database, is 15,063 for the GTX 660 Ti and 13,300 for the GTX 480.

Q: How does the GTX 660 Ti compare with the GTX 480 in raw compute?

A: The GTX 660 Ti leads in the recorded head-to-head OpenCL test with a score of 15,113 against 13,300, a 13.6% advantage. The GTX 480 has no wins in the database head-to-head comparisons.

Q: Which card has the higher memory bandwidth?

A: The GTX 660 Ti. It uses 2 GB of GDDR5 memory on a 192 bit bus, delivering 144.2 GB/s. The GTX 480 uses 1536 MB of GDDR5 on a 384 bit bus for 177.4 GB/s. The GTX 480 has a higher peak bandwidth, but the GTX 660 Ti shows better overall efficiency in the recorded results.

Q: Are both cards DirectX 12 compatible?

A: Yes. The GTX 660 Ti supports DirectX 12 (11_0) and OpenGL 4.6, plus Vulkan 1.2.175. The GTX 480 supports DirectX 12 (11_0), OpenGL 4.6, and no Vulkan entry in the database.

Architecture Differences

The architectural split between these two is significant. The GTX 660 Ti is Kepler, GF100's successor on the 600 series. The GTX 660 Ti has 1,344 shading units, 112 texture mapping units, and 24 ROPs. The GTX 480 has 480 shading units, 60 texture mapping units, and 48 ROPs. The GTX 660 Ti has a base clock of 915 MHz and a boost of 980 MHz. The GTX 480 is recorded with a memory clock of 924 MHz. Pixel throughput numbers show the Kepler card's lead: 27.44 GPixel/s versus 21.03 GPixel/s. Texel throughput is 109.8 GTexel/s against 42.06 GTexel/s.

The Kepler card also has a higher FP32 compute rating at 2.634 TFLOPS versus 1,345 GFLOPS for the Fermi card. The GTX 660 Ti uses a 192-bit memory interface, while the GTX 480 uses a 384-bit interface. Both cards are dual-slot designs. The GTX 660 Ti uses two 6-pin power connectors with a 450 W suggested PSU; the GTX 480 uses one 6-pin plus one 8-pin connector and a 600 W suggested PSU. The interface is PCIe 3.0 x16 for the GTX 660 Ti and PCIe 2.0 x16 for the GTX 480.

Head-to-Head Benchmarks

The only recorded head-to-head test in the database is a Geekbench OpenCL run. The GTX 660 Ti scores 15,113; the GTX 480 scores 13,300. The delta is 13.6% in favor of the GTX 660 Ti. This is the only direct comparison between the two, and the Kepler card wins it. The database also places the GTX 660 Ti at the 57th percentile of all GPUs, while the GTX 480 sits at the 53rd percentile.

Looking at rival cards, the GTX 660 Ti's nearest rivals in the database are tight. The AMD Radeon Pro 560X averages 15,082 (0.1% slower than the GTX 660 Ti). The AMD Radeon RX 7600 averages 15,171 (0.7% slower). The NVIDIA GeForce RTX 3050 OEM averages 15,199 (0.9% slower). The AMD Radeon 680M averages 15,270 (1.4% slower). The GTX 660 Ti leads this group by a narrow but consistent margin.

For the GTX 480, the nearest rivals are similarly close. The AMD Radeon Pro 555 averages 13,321 (0.2% slower). The AMD FirePro M6100 averages 13,354 (0.4% slower). The AMD Radeon RX 5500M averages 13,356 (0.4% slower). The AMD Radeon HD 8950M averages 13,376 (0.6% slower). The GTX 480 leads this cluster by a small margin.

Specification Differences

The key differences in the database are in the memory subsystem and the board interface. The GTX 660 Ti uses 2 GB GDDR5 on a 192-bit bus, while the GTX 480 uses 1536 MB GDDR5 on a 384-bit bus. Bandwidth is 144.2 GB/s versus 177.4 GB/s. The GTX 480 has a higher bandwidth figure. The GTX 660 Ti has more shading units (1344 vs 480), more texture units (112 vs 60), and more ROPs (24 vs 48). The die size is also very different: 294 mm² for the GTX 660 Ti versus 529 mm² for the GTX 480. The 28 nm node makes the smaller die possible.

Clock behavior also differs. The GTX 660 Ti has a 915 MHz base and 980 MHz boost. The GTX 480 memory clock is 924 MHz. The GTX 660 Ti has a pixel rate of 27.44 GPixel/s and texture rate of 109.8 GTexel/s. The GTX 480 is 21.03 GPixel/s and 42.06 GTexel/s. The Kepler card has a significant compute advantage.

The power delivery is different, too. The GTX 660 Ti is a dual-slot card with two 6-pin connectors and a 450 W power supply. The GTX 480 is dual-slot, uses one 6-pin and one 8-pin connector, and recommends 600 W.

The physical dimensions tell the story of the two boards. The GTX 660 Ti is 241 mm (9.5 inches) long. The GTX 480 is 267 mm (10.5 inches) long. The GTX 480 does not have a listed height or width in the database.

Where Each One Wins

Choose the GTX 660 Ti for the OpenCL compute lead. It is 13.6% ahead of the GTX 480 in the only recorded head-to-head and also holds a higher global percentile (57th vs 53rd). The GTX 660 Ti also wins on pixel throughput, texture fill, shading units, and transistor density. It is the newer, more efficient design.

Pick the GTX 480 if the memory subsystem matters more. It has a larger 384-bit bus, higher bandwidth at 177.4 GB/s, and more memory per board at 1536 MB. It is also the only one of the two with a high bandwidth figure, the GTX 660 Ti has a lower 144.2 GB/s. For applications that are sensitive to raw memory bandwidth rather than the OpenCL score, this is the card with more memory bandwidth.

The Verdict

The benchmark data makes a consistent case. The GTX 660 Ti wins the only recorded head-to-head test, and leads in a wide range of compute and pixel-rate metrics. The GTX 480 cannot match the Kepler card in density, shading throughput, or fill rate. It retains a single advantage in memory bandwidth and a slightly higher global percentile ranking, but no head-to-head win.

The NVIDIA GeForce GTX 660 Ti is the pick for compute workloads, for OpenCL and Vulkan performance, and for users who want the Kepler feature set (DX 12, OpenGL 4.6, Vulkan support). The GTX 480 is the fallback if memory bandwidth is the only metric on the table, but the overall data favors the newer 600-series card.