 |
VOOZH | about |
NVIDIA
Operating mode
Use this to bias workload recommendations toward responsiveness, background autonomy, lighter serving, or multi-GPU scale-out.
Current mode
Balanced
Balanced for general local use. Keeps the ranking neutral across personal and serving workflows.
The GTX 1660 Ti 6GB shares the non-RTX Turing designation with the 1660 Super but uses GDDR6 on a narrower bus, resulting in 288 GB/s bandwidth versus the Super's 336 GB/s. The 6 GB VRAM and Turing architecture (compute 7.5) make it functional for Q4 quantized 7B inference via llama.cpp, but the lower bandwidth means it's outperformed by the 1660 Super for AI workloads. At similar used prices, the 1660 Super is the better pick.
Beyond LLMs
What AI tasks this GPU can handle — from text generation to image and video creation.
| Capability | Status | Representative Model | Detail |
|---|---|---|---|
| LLM Chat (7B) | Needs offload | Llama 3.1 8B Q4 | — |
| LLM Coding (30B) | Won’t fit | Qwen 3 30B Q4 | — |
Architecture
Turing is NVIDIA's first-generation RTX architecture, introducing dedicated RT and Tensor Cores to consumer GPUs for the first time. Built on TSMC's 12nm FinFET process.
AI Relevance
The first consumer architecture with Tensor Cores, enabling meaningful acceleration for INT8 and FP16 inference. However, limited VRAM (typically 6-11 GB) restricts modern LLM model sizes.
Chat
AGemma 4 E2B matches Chat and keeps a practical fit profile. It is a recent-generation family, which helps on current local SOTA workloads. It fits natively with comfortable headroom. Context coverage stays within the requested workload envelope. Known distribution channels: huggingface, ollama, lm-studio.
Coding
AGemma 4 E2B is a specialized fit for Coding. It is a recent-generation family, which helps on current local SOTA workloads. It should run, but memory headroom will be limited. Context coverage stays within the requested workload envelope. Known distribution channels: huggingface, ollama, lm-studio.
Agentic Coding
AJust out of reach
High-quality models that need a bit more memory
Image & Video Generation
18 of 52 models can generate images or video on your GTX 1660 Ti 6GB
| Model | Max Resolution | Gen Time | Grade |
|---|---|---|---|
| SD TurboImage | 512×512 | ~5.2s | A |
| Stable Diffusion 1.5Image | 512×768 | ~10.3s | B |
| Realistic Vision v5.1Image | 512×768 | ~10.3s | B |
| DreamShaper 8Image | 512×768 | ~10.3s | B |
| LCM DreamShaper v7 |
Upgrade paths
See what you unlock with more powerful hardware
Upgrade options
Unlocks 38 additional models that do not fit on the current setup.
~$249 MSRP
Unlocks 38 additional models that do not fit on the current setup.
Lifts average decode speed across fitting models by about 70%.
~$499 MSRP
Unlocks 112 additional models that do not fit on the current setup.
Lifts average decode speed across fitting models by about 12%.
~$329 MSRP
Unlocks 193 additional models that do not fit on the current setup.
Lifts average decode speed across fitting models by about 482%.
~$8,000 MSRP
Compare with similar
| Won’t fit |
| Llama 3.1 70B Q4 |
| — |
| Image Gen (SDXL) | Very constrained | SDXL 1.0 FP16 | ~~41.2s per image |
| Image Gen (Flux) | Won't fit | Flux.1 Dev FP16 | ~~3m 5s per image |
| Image Gen (SD 3.5) | Won't fit | SD 3.5 Large FP16 | ~~3m 47s per image |
| Video Short (25f) | Won't fit | LTX Video 2B | ~~35.8s/frame |
| Video Long (100f) | Won't fit | Wan Video 14B | ~~1m 45s/frame |
Gemma 4 E2B is a specialized fit for Agentic Coding. It is a recent-generation family, which helps on current local SOTA workloads. It should run, but memory headroom will be limited. Context coverage stays within the requested workload envelope. Known distribution channels: huggingface, ollama, lm-studio.
Reasoning
AGemma 4 E2B matches Reasoning and keeps a practical fit profile. It is a recent-generation family, which helps on current local SOTA workloads. It should run, but memory headroom will be limited. Context coverage stays within the requested workload envelope. Known distribution channels: huggingface, ollama, lm-studio.
RAG
AMinistral 3 3B is viable for RAG, but is not the most specialized choice. It is a recent-generation family, which helps on current local SOTA workloads. It fits natively with comfortable headroom. Context coverage stays within the requested workload envelope. Known distribution channels: huggingface.
| 512×768 |
| ~3.1s |
| B |
| PixArt-SigmaImage | 256×256 | ~41.2s | B |
| FramePack I2VVideo | 256×256 | ~1m 16s/frame | B |
| SDXL TurboImage | 256×256 | ~5.2s | D |
| SDXL LightningImage | 256×256 | ~15.5s | D |
| Stable Diffusion XL 1.0Image | 256×256 | ~41.2s | D |
| Playground v2.5Image | 256×256 | ~1m 2s | D |
| RealVisXL v5.0Image | 256×256 | ~46.4s | D |
| DreamShaper XLImage | 256×256 | ~46.4s | D |
| Juggernaut XL v9Image | 256×256 | ~46.4s | D |
| Animagine XL 3.1Image | 256×256 | ~46.4s | D |
| Pony Diffusion V6 XLImage | 256×256 | ~46.4s | D |
| Animagine XL 4.0Image | 256×256 | ~46.4s | D |
| Illustrious XLImage | 256×256 | ~46.4s | D |
| Wan Video 2.1 1.3BVideo | 256×256 | ~30.1s/frame | F |
| Stable Diffusion 3.5 MediumImage | 256×256 | ~1m 12s | F |
| Flux.2 Klein 4BImage | 256×256 | ~12.4s | F |
| LTX Video 2BVideo | 256×256 | ~35.8s/frame | F |
| KolorsImage | 256×256 | ~1m 22s | F |
| Stable CascadeImage | 256×256 | ~1m 43s | F |
| AuraFlow v0.3Image | 256×256 | ~3m 5s | F |
| Stable Diffusion 3.5 LargeImage | 256×256 | ~3m 47s | F |
| Stable Diffusion 3.5 Large TurboImage | 256×256 | ~41.2s | F |
| CogVideoX 2BVideo | 256×256 | ~35.8s/frame | F |
| HunyuanVideoVideo | 256×256 | ~1m 16s/frame | F |
| ChromaImage | 256×256 | ~41.2s | F |
| Z-Image TurboImage | 256×256 | ~42.5s | F |
| Flux.1 DevImage | 256×256 | ~3m 5s | F |
| Flux.1 SchnellImage | 256×256 | ~36.1s | F |
| LTX Video 13BVideo | 256×256 | ~1m 16s/frame | F |
| Flux.1 Kontext DevImage | 256×256 | ~3m 26s | F |
| AnimateDiff v1.5.3Video | 512×768 | ~18.8s/frame | F |
| Cosmos Diffusion 7BVideo | 256×256 | ~59s/frame | F |
| CogVideoX 5BVideo | 256×256 | ~51.6s/frame | F |
| Wan2.2 TI2V 5BVideo | 256×256 | ~51.6s/frame | F |
| Flux.2 Klein 9BImage | 256×256 | ~20.6s | F |
| Flux.1 Fill DevImage | 256×256 | ~2m 55s | F |
| Mochi 1 PreviewVideo | 256×256 | ~1m 8s/frame | F |
| HunyuanVideo 1.5Video | 256×256 | ~1m 3s/frame | F |
| Helios 14BVideo | 256×256 | ~1m 18s/frame | F |
| SkyReels V2 14BVideo | 256×256 | ~1m 18s/frame | F |
| Wan Video 2.1 14BVideo | 256×256 | ~1m 18s/frame | F |
| Wan Video 2.2 14BVideo | 256×256 | ~1m 18s/frame | F |
| Qwen ImageImage | 256×256 | ~1m 9s | F |
| Qwen Image EditImage | 256×256 | ~1m 9s | F |
| Flux.2 DevImage | 256×256 | ~32m 29s | F |
| MAGI-1Video | 256×256 | ~1m 37s/frame | F |
| HunyuanImage 3.0Image | 256×256 | ~2m 2s | F |
Image models estimated at 1024×1024 (28 steps, FP16). Video models estimated at 768×512 (25 frames, 30 steps, FP16). Actual performance varies with runtime and system load.
Buying advice
Usable for local AI with limits
Can run 4 of 50 top models, mostly smaller ones. Larger models need heavy quantization or won't fit.
6.0 GB
VRAM
$279
MSRP
$47/GB
Cost per GB VRAM
Best models for this GPU
What will limit you first
This setup is broadly balanced for this model.
Very little memory headroom
You can run the model, but there is not much room left for longer context, bigger batches, extra apps, or future model updates.
Older PCIe generation
PCIe 3.0 is workable, but it compounds the penalty when you offload heavily or try to scale across multiple cards.
Best upgrade itinerary
Buy headroom, not only minimum fit
A slightly larger memory tier gives you safer context growth and makes the recommendation more future-proof.
Unlocks 38 additional models that do not fit on the current setup.
Want more headroom? RTX 3050 8GB (8.0 GB VRAM) is the next step up.
