Mid-range build · mid tower · New + used parts
Sleeper 3090
Updated single-GPU NVIDIA RTX 3090 workstation
Built by Llama Builds · Updated
- Total cost
- $1,804
- VRAM
- 24GB
- GPUs
- 1
- System RAM
- 32GB
What can this build run?
With 24GB of VRAM across one GPU, this build comfortably runs models up to Qwen3.6 35B-A3B (Q3_K_M, ~160 tokens/sec). Our top all-round pick for it is Qwen3.8 27B at Q4_K_M (~39 tokens/sec).
Parts list
Every part in the Sleeper 3090
We may earn a commission when you buy through links on this page. It never changes what we recommend. How we make money
| Part | Price | Buy |
|---|---|---|
GPU NVIDIA GeForce RTX 3090 Bought used | $850 | eBay |
CPU AMD Ryzen 7 7700 3.6 GHz 8-Core | $284 | Amazon |
Motherboard MSI PRO B650-P WIFI ATX AM5 | $220 | Amazon |
Memory TEAMGROUP RGB 32GB DDR5-6000 | $110 | Amazon |
Storage PNY CS3140 1TB M.2-2280 NVMe SSD | $60 | Amazon |
Case HYTE Y40 ATX Mid Tower | $130 | Amazon |
Power Supply be quiet! Pure Power 12 M 850W 80+ Gold | $140 | Amazon |
CPU Cooler Thermalright Frozen Notte ARGB Liquid Cooler | $60 | Amazon |
| Total | $1,804 |
Prices at time of build (2025); check current prices. Individual parts add up to $1,854.
Capability
Models this build runs well
Estimated for a single user with 8K context. Best quantization that fits, ranked by our capability score.
Best overall · also coding, reasoning, vision
Qwen3.8 27BQ4_K_M · ~39 tokens/sec
Runner-up
Gemma 4 31BQ4_K_M · ~33 tokens/sec
Fastest good model · also largest that fits
Qwen3.6 35B-A3BQ3_K_M · ~160 tokens/sec
| Model | Quant | Speed | Fit |
|---|---|---|---|
| Qwen3.8 27B | Q4_K_M | ~39 tok/s | Runs well |
| Gemma 4 31B | Q4_K_M | ~33 tok/s | Fits, tight |
| Gemma 4 26B A4B | Q5_K_M | ~110 tok/s | Runs well |
| Qwen3.6 35B-A3B | Q3_K_M | ~160 tok/s | Runs well |
| gpt-oss-20b | MXFP4 | ~120 tok/s | Runs well |
| Qwen3.5 9B | Q8_0 | ~59 tok/s | Runs well |
| Gemma 4 E4B | Q8_0 | ~66 tok/s | Runs well |
| Phi-4 | Q8_0 | ~41 tok/s | Runs well |
The build
About this build
A single used RTX 3090 on a modern AM5 platform in a HYTE Y40 case, with all the new parts kept under $1,300. A strong price-to-performance pick for small and medium language models.
Why this build
This is a budget 2025 solo RTX 3090 build. I built it as a dedicated AI inference rig, with the goal of keeping all the new components under $1,300. It’s optimized for running small to medium language models with an excellent price-to-performance ratio.
Notes and caveats
- The RTX 3090 is the only used part; everything else was bought new.
- The cost could have been even lower if I had opted for a used HP or Dell barebones PC with an Intel Xeon CPU.
Specifications
Full specs
- GPU
- RTX 3090
- CPU
- AMD Ryzen 7 7700 (8-core, 3.6 GHz)
- Motherboard
- MSI PRO B650-P WIFI (ATX, AM5)
- Memory
- 32GB DDR5-6000 (TEAMGROUP RGB)
- Storage
- PNY CS3140 1TB NVMe SSD
- Power supply
- be quiet! Pure Power 12 M 850W 80+ Gold
- Case
- HYTE Y40 (ATX mid tower)
- Cooling
- Thermalright Frozen Notte ARGB liquid cooler
- Form factor
- mid tower
- GPU power
- 350W rated
Keep browsing
