VQV Signal
According to a post on X by @QwenAIFans, a community report credits @yama_boku with running an NVIDIA V100 capped at 150W. The report describes 63.2 tokens per second when decoding at 128K context and 58.5 at 256K, using the Strata engine tuned for older CUDA. It is one user's own measurement, run with a 3-bit IQ3_S quantization of the model.
According to a post on X by @QwenAIFans, a community report credits @yama_boku with running an NVIDIA V100 capped at 150W. The report describes 63.2 tokens per second when decoding at 128K context and 58.5 at 256K, using the Strata engine tuned for older CUDA. It is one user's own measurement, run with a 3-bit IQ3_S quantization of the model.
AI-assisted summary based on the listed source.
A community report via @yama_boku shows an old NVIDIA V100 capped at 150W reaching 63.2 tokens a second decoding at 128K context and 58.5 at 256K, using the Strata engine tuned for older CUDA. It is one user's own measurement with a 3-bit IQ3_S quantization. Why it matters: Older data center GPUs may still be...
Builders and businesses running open models on a budget may find that older data center GPUs can still handle newer models. Because these figures come from a single user's setup, they are not a broad benchmark. Watch for independent tests of the same model and configuration before relying on these numbers.
Hardware and robotics watchers may want to track whether this becomes a product, benchmark, or deployment signal.
VQV organizes public signals from inspectable sources. It does not independently verify the underlying report.
Signal Strength reflects source quality, relevance, freshness and evidence. Public Interest helps organize discovery; it is not proof of truth.
VQV surfaced this signal because it is recent, relevant to Open Source LLMs, connected to Scobleizer News Pulse (X).
No login, cookies, social SDKs, or automatic posting.
