NVIDIA GeForce GTX 660 Ti vs NVIDIA Tesla K40c Comparison

NVIDIA
GEFORCE

NVIDIA 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
VS
NVIDIA
GEFORCE

Tesla K40c

CORE STATE GK180
VRAM 12 GB
CLOCK SPEED 876 MHz
TDP 245 W
BUS WIDTH 384 bit
ARCHITECTURE Kepler
nm
PROCESS 28 nm
LAUNCH DATE 2013

PERFORMANCE BENCHMARKS

geekbench_opencl
15,113
17,468
geekbench_vulkan
15,012
N/A

Analysis: NVIDIA GeForce GTX 660 Ti vs NVIDIA Tesla K40c

Where Each One Wins

The benchmark data splits cleanly between these two Kepler-era NVIDIA cards, but only one of them has a recorded win in the head-to-head comparison. The NVIDIA Tesla K40c takes the single available benchmark victory, the Geekbench OpenCL test, with a score of 17,468 against the GTX 660 Ti's 15,113. That is a 15.6% advantage for the Tesla card, a substantial margin that places it clearly ahead in raw compute throughput as measured by this workload.

The Tesla K40c's win is not a narrow one. A 15.6% delta in OpenCL performance means the K40c completes the same compute tasks in measurably less time, or handles larger datasets within the same time budget. This aligns with its positioning as a compute-oriented accelerator. The GTX 660 Ti, by contrast, has no recorded wins in the head-to-head data. Its only additional benchmark, Geekbench Vulkan with a score of 15,012, is not part of the head-to-head comparison set, but it does show that the GTX 660 Ti can handle modern API workloads. However, that Vulkan score is still below the Tesla K40c's OpenCL result.

Looking at the average benchmark scores tells a similar story on aggregate performance across multiple tests. The Tesla K40c has an average benchmark score of 17,468 from a single recorded test. The GTX 660 Ti averages 15,063 across its two tests (15,113 OpenCL and 15,012 Vulkan). Even when averaging in the Vulkan result, the GTX 660 Ti remains roughly 15.9% behind the Tesla K40c's average. The Tesla card's single score is higher than either of the GTX 660 Ti's individual results, which indicates that the K40c holds a consistent performance advantage regardless of which benchmark you consult.

The use-case split is therefore one-sided in terms of recorded wins, but the underlying hardware differences explain why that is the case. The Tesla K40c is built around the GK180 chip with 2,880 shading units, 240 texture mapping units, and 48 raster operations units. The GTX 660 Ti uses the GK104 chip with 1,344 shading units, 112 texture mapping units, and 24 raster operations units. More shading units generally translate directly into higher floating-point throughput for parallel compute workloads like OpenCL, which is exactly where the Tesla card wins. The GTX 660 Ti's higher clock speeds, 915 MHz base and 980 MHz boost compared to the Tesla's 745 MHz base and 876 MHz boost, help in lighter-threaded tasks, but the Tesla's massive compute resource pool overpowers the clock speed advantage in OpenCL testing.

For gaming-oriented tasks, the GTX 660 Ti has the practical advantage in that it has display outputs, specifically 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2, while the Tesla K40c has no display outputs at all. That means the GTX 660 Ti is the only one of the two that can actually drive a monitor for real-time rendering. The Tesla K40c is a compute accelerator with no video output capability. So while the recorded benchmark wins all belong to the Tesla, the GTX 660 Ti wins in the unbenchmarked but functionally important area of display connectivity. The data does not include gaming benchmarks, so any assessment of gaming performance must rest on the architectural and output differences rather than measured frame rates.

The Verdict

The recorded data points to a clear compute performance hierarchy. The NVIDIA Tesla K40c outperforms the NVIDIA GeForce GTX 660 Ti in the only shared benchmark, Geekbench OpenCL, by 15.6%. The Tesla K40c also sits at the 61st percentile among all GPUs in the database, while the GTX 660 Ti sits at the 57th percentile. That four-point percentile gap further confirms that the K40c occupies a higher tier in the overall performance distribution.

Who should pick the Tesla K40c? Anyone whose work is dominated by parallel compute workloads, OpenCL-heavy rendering, scientific simulation, machine learning inference, or any task that scales with raw FP32 throughput. The K40c delivers 5.046 TFLOPS of FP32 performance, 12 GB of memory on a 384-bit bus with 288.4 GB/s bandwidth, 240 texture mapping units, 48 ROPS, and a 52.56 GPixel/s pixel rate. Those are numbers that matter in compute-dense environments where the GTX 660 Ti's 2.634 TFLOPS, 2 GB memory on a 192-bit bus, 144.2 GB/s bandwidth, 112 TMUs, 24 ROPS, and 27.44 GPixel/s pixel rate would bottleneck quickly on larger datasets. The K40c's 15.6% lead in OpenCL directly reflects this hardware advantage.

Who should pick the GTX 660 Ti instead? Anyone needing a functional GPU that can connect to displays. The GTX 660 Ti has 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2 outputs, the Tesla K40c has none. For desktop use, casual gaming, or any workload that requires visual output, the GTX 660 Ti is the only viable option between these two. Its lower power draw, 150 W versus the Tesla's 245 W, and its lower suggested PSU, 450 W versus 550 W, also make it easier to slot into an existing system. The GTX 660 Ti is also shorter, 241 mm versus 267 mm, which helps with case compatibility. Its Vulkan score of 15,012 shows it can handle modern graphics APIs, even if that score is below the Tesla's OpenCL result.

There is no scenario in the data where the GTX 660 Ti wins a compute benchmark against the K40c. The verdict is that the Tesla K40c is the compute champion and the GTX 660 Ti is the display-capable alternative. The choice hinges entirely on whether the workload requires a monitor output or pure compute throughput.

Head-to-Head Benchmarks

The head-to-head benchmark suite contains one test: Geekbench OpenCL. The NVIDIA Tesla K40c scores 17,468, the NVIDIA GeForce GTX 660 Ti scores 15,113, giving the Tesla a 15.6% advantage. This is the only shared benchmark in the database, so all comparative analysis rests on this single measurement.

The 15.6% delta is significant when contextualized against the nearest rivals of each card. The Tesla K40c's nearest rivals in the database are all within 1% of its score: the AMD Radeon Pro 460 at 17,509 (-0.2%), the AMD Radeon Pro 560 at 17,551 (-0.5%), the AMD Radeon 780M at 17,588 (-0.7%), and the NVIDIA GeForce RTX 4060 at 17,639 (-1%). The Tesla card is essentially tied with these modern GPUs. Meanwhile, the GTX 660 Ti's nearest rivals are the AMD Radeon Pro 560X at 15,082 (-0.1%), the AMD Radeon RX 7600 at 15,171 (-0.7%), the NVIDIA GeForce RTX 3050 OEM at 15,199 (-0.9%), and the AMD Radeon 680M at 15,270 (-0.1%). The GTX 660 Ti is within 1.4% of these cards as well.

The gap between the Tesla K40c and the GTX 660 Ti is far larger than the gaps each card has to its own nearest rivals. The Tesla K40c sits in a cluster of 17.5k to 17.6k scores, while the GTX 660 Ti sits in a cluster of 15.0k to 15.3k scores. The 15.6% gap between the two reviewed cards is roughly 15 times larger than the 1% gap that separates the Tesla K40c from its nearest rival, the RTX 4060. In other words, the performance distance between the Tesla K40c and the GTX 660 Ti is substantially wider than the distance between the Tesla K40c and a much newer RTX 4060. That is a striking observation from the data: the Tesla K40c matches a modern RTX 4060 in OpenCL within 1%, while leaving the GTX 660 Ti 15.6% behind.

The GTX 660 Ti's Vulkan score of 15,012 is not part of the head-to-head set because the Tesla K40c has no recorded Vulkan benchmark. This means the Vulkan result cannot be directly compared across both cards. However, it is informative that the GTX 660 Ti's Vulkan score is nearly identical to its OpenCL score (15,012 vs 15,113), suggesting consistent performance across API types for that card. The Tesla K40c's single OpenCL score of 17,468 remains above both GTX 660 Ti results.

The one-benchmark head-to-head limits the scope of the comparison, but the single result is decisive. The Tesla K40c wins the only shared test, and the margin is large enough to be meaningful across workloads that stress similar compute resources.

FAQ

Q: Which GPU has the higher Geekbench OpenCL score?

A: The NVIDIA Tesla K40c scores 17,468, while the NVIDIA GeForce GTX 660 Ti scores 15,113. The Tesla K40c wins by 15.6%.

Q: How do these cards compare to their nearest rivals?

A: The Tesla K40c is within 1% of the AMD Radeon Pro 460, AMD Radeon Pro 560, AMD Radeon 780M, and NVIDIA GeForce RTX 4060. The GTX 660 Ti is within 1.4% of the AMD Radeon Pro 560X, AMD Radeon RX 7600, NVIDIA GeForce RTX 3050 OEM, and AMD Radeon 680M.

Q: Does the GTX 660 Ti support Vulkan?

A: Yes, the GTX 660 Ti has a recorded Geekbench Vulkan score of 15,012. The Tesla K40c has no recorded Vulkan benchmark.

Q: Which card has more memory?

A: The Tesla K40c has 12 GB of GDDR5 memory on a 384-bit bus with 288.4 GB/s bandwidth. The GTX 660 Ti has 2 GB of GDDR5 memory on a 192-bit bus with 144.2 GB/s bandwidth.

Q: Which card has display outputs?

A: The GTX 660 Ti has 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2 outputs. The Tesla K40c has no display outputs.

Q: What are the power requirements for each card?

A: The Tesla K40c has a TDP of 245 W and a suggested PSU of 550 W with 1x 6-pin + 1x 8-pin power connectors. The GTX 660 Ti has a TDP of 150 W and a suggested PSU of 450 W with 2x 6-pin power connectors.

Architecture Differences

Both cards use NVIDIA's Kepler architecture, both are fabricated on TSMC's 28 nm process, and both support DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. Beyond those shared traits, the architecture diverges significantly in the underlying chips. The Tesla K40c uses the GK180 chip, the GTX 660 Ti uses the GK104. The Tesla generation is labeled Tesla Kepler (Kxx series), while the GTX 660 Ti belongs to the GeForce 600 generation. The Tesla K40c uses 7,080 million transistors spread across a 561 mm² die, yielding a transistor density of 12.1 million transistors per square millimeter. The GTX 660 Ti uses 3,540 million transistors on a 294 mm² die, with 12.0 million transistors per square millimeter density. The Tesla K40c packs roughly twice as many transistors on a chip that is roughly 1.9 times larger.

The compute resources differ dramatically. The Tesla K40c has 2,880 shading units, 240 texture mapping units, and 48 raster operations units. The GTX 660 Ti has 1,344 shading units, 112 texture mapping units, and 24 raster operations units. The Tesla K40c has exactly 2.14 times the shading units, 2.14 times the TMUs, and 2 times the ROPS of the GTX 660 Ti. This near-doubling of every major compute block explains the 15.6% OpenCL lead, though the lead is smaller than the raw unit count difference suggests, likely because clock speeds differ. The GTX 660 Ti runs at 915 MHz base and 980 MHz boost, the Tesla K40c at 745 MHz base and 876 MHz boost. The GTX 660 Ti's higher clocks close some of the gap, but not all of it.

Neither card has ray tracing cores or tensor cores. Both are pure raster and compute designs. The Tesla K40c comes from the Tesla Fermi generation's successor line, with Tesla Maxwell as its successor. The GTX 660 Ti succeeds the GeForce 500 generation and is succeeded by GeForce 700. The Tesla K40c's production status is end-of-life, as is the GTX 660 Ti's.

Specification Differences

The two cards differ in nearly every physical and performance specification. The Tesla K40c is the larger card at 267 mm (10.5 inches) in length, the GTX 660 Ti is 241 mm (9.5 inches). Both are dual-slot designs. The Tesla K40c has a base clock of 745 MHz with a boost of 876 MHz, the GTX 660 Ti has a base clock of 915 MHz with a boost of 980 MHz. Memory clock is the same on both: 1502 MHz with 6 Gbps effective. Memory size differs substantially, 12 GB on the Tesla K40c versus 2 GB on the GTX 660 Ti. Memory bus width is 384-bit on the Tesla K40c and 192-bit on the GTX 660 Ti. Memory bandwidth is 288.4 GB/s versus 144.2 GB/s.

Pixel rate is 52.56 GPixel/s on the Tesla K40c versus 27.44 GPixel/s on the GTX 660 Ti. Texture rate is 210.2 GTexel/s versus 109.8 GTexel/s. FP32 compute is 5.046 TFLOPS versus 2.634 TFLOPS. Neither card has FP16 capability listed. Power consumption differs with the Tesla K40c drawing 245 W and the GTX 660 Ti drawing 150 W. The suggested PSU is 550 W for the Tesla K40c and 450 W for the GTX 660 Ti. Power connectors are 1x 6-pin + 1x 8-pin on the Tesla K40c and 2x 6-pin on the GTX 660 Ti.

Display outputs are a major differentiator. The Tesla K40c has no outputs, the GTX 660 Ti has 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2. Both use PCIe 3.0 x16 interfaces. Both support DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. The Tesla K40c has a launch MSRP of 7,699 USD. The GTX 660 Ti has a launch MSRP of 299 USD. The Tesla K40c was released in October 2013, the GTX 660 Ti in August 2012. The Tesla K40c uses the GK180 chip, the GTX 660 Ti uses the GK104. Both are 28 nm TSMC parts, but the Tesla K40c has 7,080 million transistors versus 3,540 million on the GTX 660 Ti, and a 561 mm² die versus 294 mm². The Tesla K40c is a compute accelerator with no display output, the GTX 660 Ti is a consumer graphics card with full display capability.

DETAILED SPECIFICATIONS

SPECIFICATION
GTX 660 Ti
Tesla K40c
Core Specs
Shading Units
1,344
2,880 +114.3%
Shaders
1,344
2,880 +114.3%
TMUs
112
240 +114.3%
ROPs
24
48 +100.0%
Clocks
Base Clock
915 MHz
745 MHz
Boost Clock
980 MHz
876 MHz
Memory Clock
1502 MHz 6 Gbps effective
1502 MHz 6 Gbps effective
Memory
Memory Size
2 GB
12 GB
VRAM (MB)
2,048
12,288 +500.0%
Memory Type
GDDR5
GDDR5
Memory Bus
192 bit
384 bit
Bandwidth
144.2 GB/s
288.4 GB/s
Cache
L1 Cache
16 KB (per SMX)
16 KB (per SMX)
L2 Cache
384 KB
1536 KB
Performance
Pixel Rate
27.44 GPixel/s
52.56 GPixel/s
Texture Rate
109.8 GTexel/s
210.2 GTexel/s
FP32 (TFLOPS)
2.634 TFLOPS
5.046 TFLOPS
FP64 (TFLOPS)
109.8 GFLOPS (1:24)
1.682 TFLOPS (1:3)
Power
TDP
150 W
245 W
TDP (W)
150
245 +63.3%
Suggested PSU
450 W
550 W
Power Connectors
2x 6-pin
1x 6-pin + 1x 8-pin
Architecture
Architecture
Kepler
Kepler
GPU Name
GK104
GK180
Generation
GeForce 600
Tesla Kepler (Kxx)
Process Size
28 nm
28 nm
Transistors
3,540 million
7,080 million
Die Size
294 mm²
561 mm²
Foundry
TSMC
TSMC
Density
12.0M / mm²
12.6M / mm²
API Support
DirectX
12 (11_0)
12 (11_0)
OpenGL
4.6
4.6
Vulkan
1.2.175
1.2.175
OpenCL
3.0
3.0
CUDA
3.0
3.5
Shader Model
6.5 (5.1)
5.1
Physical
Slot Width
Dual-slot
Dual-slot
Length
241 mm 9.5 inches
267 mm 10.5 inches
Outputs
2x DVI1x HDMI 1.4a1x DisplayPort 1.2
No outputs
Bus Interface
PCIe 3.0 x16
PCIe 3.0 x16
Other
Launch Price
299 USD
7,699 USD
Production
End-of-life
End-of-life
Predecessor
GeForce 500
Tesla Fermi
Successor
GeForce 700
Tesla Maxwell
View GeForce GTX 660 Ti Details View Tesla K40c Details