NVIDIA GeForce GTX 1080 Ti vs NVIDIA GeForce GTX 660 Ti Comparison

NVIDIA
GEFORCE

NVIDIA GeForce GTX 1080 Ti

CORE STATE GP102
VRAM 11 GB
CLOCK SPEED 1582 MHz
TDP 250 W
BUS WIDTH 352 bit
ARCHITECTURE Pascal
nm
PROCESS 16 nm
LAUNCH DATE 2017
VS
NVIDIA
GEFORCE

GeForce GTX 660 Ti

CORE STATE GK104
VRAM 2 GB
CLOCK SPEED 980 MHz
TDP 150 W
BUS WIDTH 192 bit
ARCHITECTURE Kepler
nm
PROCESS 28 nm
LAUNCH DATE 2012

PERFORMANCE BENCHMARKS

3dmark_3dmark_steel_nomad_dx12
2,231
N/A
geekbench_metal
30,624
N/A
geekbench_opencl
67,929
15,113
geekbench_vulkan
40,511
15,012
passmark_directx_10
118
N/A
passmark_directx_11
151
N/A
passmark_directx_12
66
N/A
passmark_directx_9
231
N/A
passmark_g2d
939
N/A
passmark_g3d
18,600
N/A
passmark_gpu_compute
9,632
N/A

Analysis: NVIDIA GeForce GTX 1080 Ti vs NVIDIA GeForce GTX 660 Ti

The GeForce GTX 1080 Ti and GeForce GTX 660 Ti represent two distinct eras of NVIDIA's engineering, separated by a five-year gap in design philosophy. The data shows the 1080 Ti as a high-end Pascal flagship, while the 660 Ti was a mid-range Kepler part. Their benchmark scores reveal a generational chasm, with the newer card dominating in every measurable compute task. This analysis compares their raw performance, architectural foundations, and specifications to clarify where each card stands historically.

FAQ

Q: How large is the performance gap in the Geekbench OpenCL test?

A: The GTX 1080 Ti scores 67,929, which is 349.5% higher than the GTX 660 Ti's 15,113. This represents a massive lead for the Pascal card in general-purpose compute workloads.

Q: Does the older GTX 660 Ti win any of the head-to-head benchmark comparisons?

A: No. The head-to-head data records 2 wins for the GTX 1080 Ti and 0 wins for the GTX 660 Ti, covering the Geekbench OpenCL and Vulkan tests.

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

A: The GTX 1080 Ti has an average benchmark score of 15,548, while the GTX 660 Ti averages 15,063. The newer card is about 3.2% ahead in this aggregate metric.

Q: How do the cards compare in terms of memory bandwidth?

A: The GTX 1080 Ti provides 484.4 GB/s of bandwidth from 11 GB of GDDR5X on a 352-bit interface. The GTX 660 Ti offers 144.2 GB/s from 2 GB of GDDR5 on a 192-bit bus, a difference of over 3x in bandwidth.

Q: What is the difference in their DirectX API support?

A: The GTX 1080 Ti supports DirectX 12 (12_1), whereas the GTX 660 Ti supports DirectX 12 (11_0). This indicates a higher feature level for the newer architecture.

Q: Which card has a higher transistor density?

A: The GTX 1080 Ti, built on a 16 nm TSMC process, packs 25.1M transistors per mm². The GTX 660 Ti, on a 28 nm process, has a density of 12.0M per mm².

The Verdict

The choice between these two cards is dictated by the data, not nostalgia. The GTX 1080 Ti is the unequivocal performance winner, with a 169.9% lead in Vulkan and a 349.5% lead in OpenCL. Its average benchmark score of 15,548 also places it in a higher tier, with its nearest rival being the AMD Radeon Pro W5500 at 15,679 (a -0.8% delta). The GTX 660 Ti, with an average score of 15,063, sits near the AMD Radeon Pro 560X at 15,082 (-0.1% delta), showing it is competitive with much later mid-range parts but not with the 1080 Ti.

For any user prioritizing modern compute performance, even in legacy APIs, the GTX 1080 Ti is the only logical choice. The data shows its FP32 throughput of 11.34 TFLOPS versus 2.634 TFLOPS for the older card, a 4.3x difference that explains the benchmark results. The GTX 660 Ti, however, holds a niche for those interested in the Kepler architecture's efficiency, operating at a 150 W TDP compared to the 1080 Ti's 250 W. Its smaller 294 mm² die and 3,540 million transistors also make it a historically significant piece, but not a competitive one. The verdict is clear: the GTX 1080 Ti is a high-performance part, while the GTX 660 Ti is a legacy product.

Head-to-Head Benchmarks

The benchmark data provides a stark illustration of generational progress. In the Geekbench OpenCL test, the GTX 1080 Ti scores 67,929 against the GTX 660 Ti's 15,113. This 349.5% delta is the largest in the comparison and reflects the massive increase in shading units (3,584 vs 1,344) and texture mapping units (224 vs 112). The Pascal card's 11.34 TFLOPS FP32 performance is over four times the Kepler card's 2.634 TFLOPS, directly translating to this compute-heavy workload advantage.

The Vulkan test shows a similarly decisive result. The GTX 1080 Ti scores 40,511, while the GTX 660 Ti manages 15,012. This 169.9% delta, while smaller than the OpenCL gap, still demonstrates a clear victory for the newer architecture. The 1080 Ti's support for Vulkan 1.4 and higher pixel rate (139.2 GPixel/s vs 27.44 GPixel/s) likely contributes to its superior performance in this modern graphics API. The older card's Vulkan 1.2.175 support is functional but cannot overcome the raw hardware disparity.

These two tests are the only head-to-head comparisons available, but they are sufficient to establish the hierarchy. The GTX 1080 Ti wins both, with winsA at 2 and winsB at 0. The average benchmark scores corroborate this, with the 1080 Ti at 15,548 and the 660 Ti at 15,063. The closest rival to the 1080 Ti is the AMD Radeon R9 M380, which is only 0.2% behind, indicating the 1080 Ti's performance is well-established in its class. For the 660 Ti, its nearest competitor is the AMD Radeon Pro 560X at -0.1%, showing it still trades blows with much newer mid-range hardware despite its age.

Specification Differences

The specification sheets highlight two cards from different performance tiers. The GTX 1080 Ti has 3,584 shading units, 224 TMUs, and 88 ROPs, compared to the GTX 660 Ti's 1,344 shading units, 112 TMUs, and 24 ROPs. This is a fundamental difference in processing capacity, with the newer card offering more than double the TMUs and over triple the ROPs. The pixel rate of 139.2 GPixel/s for the 1080 Ti versus 27.44 GPixel/s for the 660 Ti reflects this ROP advantage.

Memory configuration is another major divider. The GTX 1080 Ti uses 11 GB of GDDR5X on a 352-bit bus, delivering 484.4 GB/s. The GTX 660 Ti uses 2 GB of GDDR5 on a 192-bit bus, providing 144.2 GB/s. This 3.4x bandwidth difference is critical for high-resolution textures and modern game assets. The memory clock also differs, with the 1080 Ti running at 1376 MHz (11 Gbps effective) versus the 660 Ti's 1502 MHz (6 Gbps effective).

Power and physical requirements also diverge. The GTX 1080 Ti has a 250 W TDP and requires a 600 W suggested PSU, using 1x 6-pin + 1x 8-pin connectors. The GTX 660 Ti is more modest, with a 150 W TDP, a 450 W suggested PSU, and 2x 6-pin connectors. The newer card is also longer at 267 mm versus 241 mm, reflecting its more complex cooler and larger PCB. Both are dual-slot cards, but the 1080 Ti's dimensions are more demanding for case compatibility.

Architecture Differences

The architectural divide is rooted in the manufacturing process and core design. The GTX 1080 Ti uses the GP102 chip on a 16 nm TSMC process, packing 11,800 million transistors into a 471 mm² die. The GTX 660 Ti uses the GK104 chip on a 28 nm TSMC process, with 3,540 million transistors on a 294 mm² die. This process shrink from 28 nm to 16 nm allows for a transistor density of 25.1M per mm² for Pascal versus 12.0M per mm² for Kepler, enabling the massive core count increase.

The Pascal architecture in the 1080 Ti is built for higher throughput, with a base clock of 1481 MHz and boost clock of 1582 MHz. Kepler's GK104 runs at 915 MHz base and 980 MHz boost. Despite the lower clocks, the 1080 Ti's 11.34 TFLOPS FP32 output dwarfs the 660 Ti's 2.634 TFLOPS. The FP16 performance also differs, with the 1080 Ti offering 177.2 GFLOPS at a 1:64 ratio, while the 660 Ti has no FP16 data listed. This indicates a focus on compute flexibility in the newer generation.

Display outputs and API support show further evolution. The GTX 1080 Ti provides 1x HDMI 2.0 and 3x DisplayPort 1.4a, while the GTX 660 Ti offers 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2. The newer card's support for DirectX 12 (12_1) and Vulkan 1.4, versus DirectX 12 (11_0) and Vulkan 1.2.175 for the older card, reflects a higher feature set. Neither card has ray tracing or tensor cores, as these are absent from the data for both. The 1080 Ti's release in March 2017 and the 660 Ti's in August 2012 also place them at the start and end of their respective architectural eras.

DETAILED SPECIFICATIONS

SPECIFICATION
GTX 1080 Ti
GTX 660 Ti
Core Specs
Shading Units
3,584
1,344 -62.5%
Shaders
3,584
1,344 -62.5%
TMUs
224
112 -50.0%
ROPs
88
24 -72.7%
SM Count
28
—
Clocks
Base Clock
1481 MHz
915 MHz
Boost Clock
1582 MHz
980 MHz
Memory Clock
1376 MHz 11 Gbps effective
1502 MHz 6 Gbps effective
Memory
Memory Size
11 GB
2 GB
VRAM (MB)
11,264
2,048 -81.8%
Memory Type
GDDR5X
GDDR5
Memory Bus
352 bit
192 bit
Bandwidth
484.4 GB/s
144.2 GB/s
Cache
L1 Cache
48 KB (per SM)
16 KB (per SMX)
L2 Cache
2.75 MB
384 KB
Performance
Pixel Rate
139.2 GPixel/s
27.44 GPixel/s
Texture Rate
354.4 GTexel/s
109.8 GTexel/s
FP32 (TFLOPS)
11.34 TFLOPS
2.634 TFLOPS
FP64 (TFLOPS)
354.4 GFLOPS (1:32)
109.8 GFLOPS (1:24)
FP16 (TFLOPS)
177.2 GFLOPS (1:64)
—
Power
TDP
250 W
150 W
TDP (W)
250
150 -40.0%
Suggested PSU
600 W
450 W
Power Connectors
1x 6-pin + 1x 8-pin
2x 6-pin
Architecture
Architecture
Pascal
Kepler
GPU Name
GP102
GK104
Generation
GeForce 10
GeForce 600
Process Size
16 nm
28 nm
Transistors
11,800 million
3,540 million
Die Size
471 mm²
294 mm²
Foundry
TSMC
TSMC
Density
25.1M / mm²
12.0M / mm²
API Support
DirectX
12 (12_1)
12 (11_0)
OpenGL
4.6
4.6
Vulkan
1.4
1.2.175
OpenCL
3.0
3.0
CUDA
6.1
3.0
Shader Model
6.8
6.5 (5.1)
Physical
Slot Width
Dual-slot
Dual-slot
Length
267 mm 10.5 inches
241 mm 9.5 inches
Height
112 mm 4.4 inches
—
Outputs
1x HDMI 2.03x DisplayPort 1.4a
2x DVI1x HDMI 1.4a1x DisplayPort 1.2
Bus Interface
PCIe 3.0 x16
PCIe 3.0 x16
Other
Launch Price
699 USD
299 USD
Production
End-of-life
End-of-life
Predecessor
GeForce 900
GeForce 500
Successor
GeForce 20
GeForce 700
View GeForce GTX 1080 Ti Details View GeForce GTX 660 Ti Details