AI & ML interests
local inference, llama.cpp, vllm, quantisation, GGUF, Blackwell architecture, workstation GPU, single-GPU workflows
Recent Activity
Organization Card
RTX PRO 4000 Blackwell
Community hub for single-slot workstation GPU owners. Benchmarks, configs, and real-world results.
Hardware Specs
Architecture
Blackwell (GB203)
VRAM
24 GB GDDR7 ECC
Bandwidth
672 GB/s
CUDA Cores
8,960
Tensor Cores
280 (5th Gen)
TDP
145W
Form Factor
Single-slot
PCIe
Gen 5 x16
What this community is for
- Inference benchmarks: tok/s across models, quant levels, and backends (llama.cpp, vLLM, TensorRT-LLM)
- Quantisation compatibility: which GGUF/NVFP4/FP8 quants fit in 24 GB and how they perform
- Workstation configs: systemd services, Docker setups, thermal management, multi-GPU pairing
- Real-world results from owner-operated hardware, not cloud benchmarks
Community resources
Repos published under this org. Contributions welcome.
Contribute
Own an RTX PRO 4000 Blackwell? Share your benchmarks and configs.
Open a discussion on any repo or submit results to the benchmarks dataset. Tag your model cards with rtx-pro-4000 so others can find them.
models 0
None public yet
datasets 0
None public yet