Is the GTX 1660 S suitable for LLM inference or fine-tuning?

With 6 GB of VRAM and about 277 GB/s of memory bandwidth, it is limited to running very compact, heavily quantized language models for basic inference. For fine-tuning or running larger models, you should select a GPU with a larger memory capacity.

Does this Turing GPU support modern CUDA software?

Yes, the GTX 1660 S supports compute capability 7.5, which is compatible with standard PyTorch/CUDA environments. However, as a consumer GTX-series card, it focuses on standard single-precision compute at about 3.6 TFLOPS rather than specialized datacenter features.

Can I use the GTX 1660 S for image generation in ComfyUI?

Yes, you can deploy a ready ComfyUI template on Cloudj to generate standard-resolution images. Complex pipelines with multiple control networks or larger foundation models will require upgrading to a card with more than 6 GB of VRAM.

How do billing and access work for this instance?

Cloudj provides per-hour billing so you only pay for the time your instance is active. You can connect directly via SSH access or launch pre-configured templates including PyTorch/CUDA, vLLM, Ollama, and ComfyUI.

GPU Cloud · NVIDIA · Turing

Rent NVIDIA GTX 1660 S GPU by the hour

The NVIDIA GTX 1660 S is an entry-level consumer GPU built on the Turing architecture, featuring 6 GB of VRAM and a 125 W power envelope. It provides about 3.6 TFLOPS of single-precision compute and about 277 GB/s of memory bandwidth with compute capability 7.5. Rent this instance on Cloudj with per-hour billing, SSH access, and ready templates for PyTorch/CUDA, vLLM, Ollama, and ComfyUI.

Rent GTX 1660 S All GPUs

Specifications

VRAM per GPU

6 GB

Architecture

NVIDIA Turing

Compute capability

7.5

FP32 compute

≈ 3.6 TFLOPS

Memory bandwidth

≈ 277 GB/s

Power limit

125 W

Values reported by the infrastructure provider for machines currently listed; real performance varies by machine.

Available configurations

Live stock and prices, updated continuously. The final price is confirmed before you order.

Rent now

What the GTX 1660 S is good for

Code Development and Debugging

Test and validate PyTorch and CUDA pipelines on a Turing architecture card before deploying to larger compute instances.

Small Language Model Inference

Run compact quantized language models using Ollama or vLLM within the 6 GB VRAM limit.

Basic Image Generation

Execute standard resolution image synthesis workflows in ComfyUI that fit comfortably inside a 6 GB memory buffer.

Lightweight Model Training

Train small neural networks, classical machine learning models, and introductory deep learning projects.

Is it the right GPU for you?

Choose the GTX 1660 S for budget-conscious development, code testing, and lightweight inference tasks that fit within 6 GB of VRAM. Opt for a GPU with more memory and higher compute throughput if you need to fine-tune modern foundation models or run large unquantized workloads, as this is already a compact entry-level option.

Compare with similar GPUs

GPU	VRAM per GPU	FP32 compute	Power limit
NVIDIA GTX 1660 S	6 GB	≈ 3.6 TFLOPS	125 W
NVIDIA GTX 1070 Ti	8 GB	≈ 7.4 TFLOPS	180 W	Details →
NVIDIA GTX 1080	8 GB	≈ 7.8 TFLOPS	130 W	Details →
NVIDIA GTX 1080 Ti	11 GB	≈ 11 TFLOPS	250 W	Details →
NVIDIA GTX 1650	4 GB	≈ undefined TFLOPS	W	Details →

Frequently asked questions

Rent the GTX 1660 S by the hour

Create an account, choose a configuration and connect over SSH. Billing stops when you delete the GPU.

Rent GTX 1660 S All GPUs