PulseAugur
EN
LIVE 13:27:04

Custom rig hits 33.4 tps with Qwen3.8 model on quad P100s

By PulseAugur Editorial · [1 sources] ·

A user has shared telemetry data from a custom-built server rig designed for running large language models. The setup features four NVIDIA Tesla P100 GPUs with a total of 64GB of VRAM, utilizing 58.7 GB for operation. This configuration achieved a speed of 33.4 tokens per second for output generation and 47 tokens per second for prompt prefill while running the 120 billion parameter Qwen3.8 model via the Strata framework. The system demonstrated efficient operation, maintaining a temperature of 72°C with a low power draw of 170W. AI

IMPACT Demonstrates efficient hardware utilization for running large language models locally.

RANK_REASON User-generated telemetry for a custom hardware setup running an LLM.

Read on Mastodon — fosstodon.org →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

Custom rig hits 33.4 tps with Qwen3.8 model on quad P100s

How we ranked this

Signal score
9 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
User-generated telemetry for a custom hardware setup running an LLM.
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
infra, model release
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

Full methodology in our editorial standards.

COVERAGE [1]

  1. Mastodon — fosstodon.org TIER_1 English(EN) · [email protected] ·

    Dawg! 33.4 tps on this quad P100 rig running Strata (120b QWEN3.8). Yeah, it works. Fast. Telemetry for a custom quad NVIDIA Tesla P100 16GB setup (64GB total H

    Dawg! 33.4 tps on this quad P100 rig running Strata (120b QWEN3.8). Yeah, it works. Fast. Telemetry for a custom quad NVIDIA Tesla P100 16GB setup (64GB total HBM2 VRAM) running alongside 128GB of system DDR3 memory. The system shows 58.7 GB of VRAM actively allocated across all …

RELATED ENTITIES

RELATED TOPICS