NVIDIA GeForce GTX 660 OEM

NVIDIA GeForce GTX 660 OEM

NVIDIA GeForce GTX 660 OEM in 2025: Nostalgia or Practicality?

Let’s figure out who this graphics card is suitable for 13 years after its release.


1. Architecture and Key Features

Kepler Architecture: The Foundation of Stability

The GTX 660 OEM is based on the Kepler architecture, which debuted in 2012. This was NVIDIA's first generation optimized for energy efficiency, but it falls short compared to modern counterparts in computing power. The manufacturing process is 28 nm, which is considered outdated by 2025 (new cards use 4–5 nm).

Lack of Modern Technologies

The card does not support ray tracing (RTX), DLSS, FidelityFX, or similar features. It's purely a "raster" GPU designed for basic tasks. The only highlight is support for DirectX 11 and OpenGL 4.6, allowing for smooth operation of older games, but it limits compatibility with newer APIs like Vulkan 2.0 or DirectX 12 Ultimate.


2. Memory: Modest Specifications

GDDR5 and Narrow Bus

The GTX 660 OEM is equipped with 1.5–2 GB of GDDR5 memory with a 192-bit bus. The bandwidth is 144.2 GB/s (6 GHz effective frequency). In comparison, even budget cards of 2025, such as the RTX 4050, offer a 128-bit bus but with GDDR6X and speeds up to 18 GHz, resulting in 288 GB/s.

Gaming Limitations

The memory capacity is sufficient for running older titles at 1080p (e.g., The Witcher 3 on medium settings), but in modern games, even with low textures, frame drops are possible due to insufficient VRAM. For instance, Hogwarts Legacy requires a minimum of 4 GB VRAM to run.


3. Gaming Performance: Realities of 2025

1080p — The Limit of Capabilities

In less demanding games (CS:GO, Dota 2, Rocket League), the GTX 660 OEM delivers 60–90 FPS at low settings. However, in titles like Cyberpunk 2077 or Starfield, even at 720p and minimum settings, the FPS rarely exceeds 20–25 frames.

1440p and 4K — Not for This Card

Due to weak memory and computational resources, resolutions above 1080p are impractical. In Fortnite (Performance mode, 1440p), the average FPS barely reaches 30.

Ray Tracing: No Support Available

The Kepler architecture lacks dedicated RT cores and tensor cores. Technologies like DLSS or FSR are also unavailable, making the card unsuitable for ray tracing games.


4. Professional Tasks: Minimal Applicability

CUDA Cores: Only for Simple Tasks

With 960 CUDA cores, the GTX 660 OEM can handle basic editing in Premiere Pro or DaVinci Resolve, but rendering 4K video will take 3–4 times longer compared to modern GPUs. In 3D modeling (Blender), complex scenes will cause lag.

Scientific Calculations: Outdated Option

For computations based on CUDA/OpenCL, the card is only suitable for educational projects or simple simulations. Compared to modern GPUs (e.g., RTX 4060), its performance in machine learning tasks is virtually negligible.


5. Power Consumption and Heat Generation

TDP 140W: Modest but Not Ideal

The card consumes less than top models from 2025 (e.g., RTX 4090 with TDP of 450W), but 140W is an average figure for its class.

Cooling Recommendations

- Cooling System: At least 2 fans or a tower cooler.

- Case: Good ventilation with 2–3 intake fans and 1 exhaust fan.

- Thermal Paste: Replacement every 2–3 years (temperature under load — up to 75°C).


6. Comparison with Competitors

AMD Radeon HD 7870: Main Competitor of 2012

- Memory: 2 GB GDDR5, 256-bit bus (bandwidth — 153.6 GB/s).

- Performance: 10–15% higher in DirectX 11, but worse optimization for modern APIs.

Modern Analogues (2025): NVIDIA RTX 3050

- Price: $200–250 (new).

- Advantages: 8 GB GDDR6, support for DLSS 3.5, ray tracing, and 4–5 times the performance.

Conclusion: The GTX 660 OEM lags behind even budget GPUs of 2025 but can serve as an alternative for ultra-cheap builds.


7. Practical Tips

Power Supply: Minimum 450W

A power supply with an 80+ Bronze certification and detachable cables is recommended (for example, EVGA 450 BR).

Platform Compatibility

- PCIe 3.0 x16 slot (backward compatible with PCIe 2.0).

- Motherboards: Even older models based on Intel H61 or AMD A75 will work.

Drivers: Limited Support

Official drivers from NVIDIA will be updated until 2023. Compatibility mode may be required for Windows 11/12.


8. Pros and Cons

Pros:

- Low price: $80–100 (new OEM supplies).

- Quiet operation under moderate loads.

- Sufficient for office tasks and older games.

Cons:

- No support for modern technologies (RTX, DLSS).

- Limited memory capacity.

- High risk of overheating in poorly ventilated cases.


9. Final Conclusion: Who is the GTX 660 OEM Suitable For?

This graphics card is a choice for:

1. Budget Builds: If you need a PC for internet browsing, office programs, or running games up to 2015.

2. Retro Enthusiasts: Those who collect hardware or experiment with old systems.

3. Temporary Solutions: Until you save up for a modern GPU.

Why You Shouldn’t Buy It?

For gaming in 2025, professional editing, or working with AI, this card is useless. Even a budget RTX 3050 or AMD RX 6500 XT ($150–180) will offer significantly more capabilities.


Conclusion

The NVIDIA GeForce GTX 660 OEM in 2025 is a relic that only has a place in niche scenarios. If your budget is strictly limited and your tasks are simple, it may suffice. In all other cases, it is better to pay extra for modern solutions.

Basic

Label Name
NVIDIA
Platform
Desktop
Launch Date
August 2012
Model Name
GeForce GTX 660 OEM
Generation
GeForce 600
Base Clock
823MHz
Boost Clock
888MHz
Bus Interface
PCIe 3.0 x16
Transistors
3,540 million
TMUs
?
Texture Mapping Units (TMUs) serve as components of the GPU, which are capable of rotating, scaling, and distorting binary images, and then placing them as textures onto any plane of a given 3D model. This process is called texture mapping.
96
Foundry
TSMC
Process Size
28 nm
Architecture
Kepler

Memory Specifications

Memory Size
1536MB
Memory Type
GDDR5
Memory Bus
?
The memory bus width refers to the number of bits of data that the video memory can transfer within a single clock cycle. The larger the bus width, the greater the amount of data that can be transmitted instantaneously, making it one of the crucial parameters of video memory. The memory bandwidth is calculated as: Memory Bandwidth = Memory Frequency x Memory Bus Width / 8. Therefore, when the memory frequencies are similar, the memory bus width will determine the size of the memory bandwidth.
192bit
Memory Clock
1400MHz
Bandwidth
?
Memory bandwidth refers to the data transfer rate between the graphics chip and the video memory. It is measured in bytes per second, and the formula to calculate it is: memory bandwidth = working frequency × memory bus width / 8 bits.
134.4 GB/s

Theoretical Performance

Pixel Rate
?
Pixel fill rate refers to the number of pixels a graphics processing unit (GPU) can render per second, measured in MPixels/s (million pixels per second) or GPixels/s (billion pixels per second). It is the most commonly used metric to evaluate the pixel processing performance of a graphics card.
21.31 GPixel/s
Texture Rate
?
Texture fill rate refers to the number of texture map elements (texels) that a GPU can map to pixels in a single second.
85.25 GTexel/s
FP64 (double)
?
An important metric for measuring GPU performance is floating-point computing capability. Double-precision floating-point numbers (64-bit) are required for scientific computing that demands a wide numeric range and high accuracy, while single-precision floating-point numbers (32-bit) are used for common multimedia and graphics processing tasks. Half-precision floating-point numbers (16-bit) are used for applications like machine learning, where lower precision is acceptable.
85.25 GFLOPS
FP32 (float)
?
An important metric for measuring GPU performance is floating-point computing capability. Single-precision floating-point numbers (32-bit) are used for common multimedia and graphics processing tasks, while double-precision floating-point numbers (64-bit) are required for scientific computing that demands a wide numeric range and high accuracy. Half-precision floating-point numbers (16-bit) are used for applications like machine learning, where lower precision is acceptable.
2.087 TFLOPS

Miscellaneous

Shading Units
?
The most fundamental processing unit is the Streaming Processor (SP), where specific instructions and tasks are executed. GPUs perform parallel computing, which means multiple SPs work simultaneously to process tasks.
1152
L1 Cache
16 KB (per SMX)
L2 Cache
384KB
TDP
130W
Vulkan Version
?
Vulkan is a cross-platform graphics and compute API by Khronos Group, offering high performance and low CPU overhead. It lets developers control the GPU directly, reduces rendering overhead, and supports multi-threading and multi-core processors.
1.1
OpenCL Version
3.0
OpenGL
4.6
DirectX
12 (11_0)
CUDA
3.0
Power Connectors
1x 6-pin
Shader Model
5.1
ROPs
?
The Raster Operations Pipeline (ROPs) is primarily responsible for handling lighting and reflection calculations in games, as well as managing effects like anti-aliasing (AA), high resolution, smoke, and fire. The more demanding the anti-aliasing and lighting effects in a game, the higher the performance requirements for the ROPs; otherwise, it may result in a sharp drop in frame rate.
24
Suggested PSU
300W

Benchmarks

FP32 (float)
Score
2.087 TFLOPS

Compared to Other GPU

FP32 (float) / TFLOPS
2.193 +5.1%
2.015 -3.4%
1.976 -5.3%