GPU-4
Vision and video on a few streams
- GPU
- Quadro P1000, 4 GB
- vCPU
- 4
- Memory
- 8 GB
- NVMe
- 80 GB
104.76 AZN / month
or 3.88 AZN by the day · 1062.15 AZN a year
Products / GPU servers
GPU servers
A cloud server in Baku with a whole NVIDIA Quadro P1000 passed through to it. Not shared, not sliced.
Available on request: tell us what you'll run and we confirm a card is free.
ubuntu@vision-1
$ nvidia-smi -LGPU 0: Quadro P1000 (UUID: GPU-…)$ docker run --gpus all …
GPU plans
The same card on two server sizes. Any other size works too: the card is 1.80 AZN a day on top.
Ask for GPU-4
GPU-8
More streams, bigger pipelines
- GPU
- Quadro P1000, 4 GB
- vCPU
- 8
- Memory
- 16 GB
- NVMe
- 160 GB
160.92 AZN / month
or 5.96 AZN by the day · 1631.55 AZN a year
The card
- GPU
- NVIDIA Quadro P1000
- Architecture
- Pascal, compute capability 6.1
- Memory
- 4 GB GDDR5
- CUDA cores
- 640
- Video
- NVENC and NVDEC encode and decode
- Access
- The whole card, passed through to your server
Ready when it boots
- Ubuntu 24.04 LTS with NVIDIA's headless server driver
- Docker with the NVIDIA Container Toolkit
- CUDA 12 inside your containers
- The same firewall, key-only SSH and snapshots as every server
Check your framework first. The P1000 is a Pascal card: CUDA 13 and some recent framework builds no longer support it. CUDA 12 and ONNX Runtime work well, and we're happy to check your stack before you rent.
Good for
Computer vision
Object detection and tracking with YOLO-class models on live camera streams.
Video
Decoding and encoding streams on the card's own video engines, leaving the CPU free.
Inference
Small and medium models that fit in 4 GB, served with ONNX Runtime or TensorRT.
CUDA development
A real NVIDIA card to build and test CUDA code on, without buying one.
Not a fit for training large models, or for LLMs that need more than 4 GB of GPU memory.
Tell us what you want to run.
We check it fits the card and confirm one is free.