# 推理 · 日跟踪 · 2026-10-06

> 日期：2026-10-06（Asia/Shanghai，周二）。专题正文：`intel/topics/inference.md`。原料：`/workspace/intel/archive/2026-10-06/`（窗口 10-05 早报后→10-06 早）；对照 2026-10-05 日跟踪。抓取：主跑 Saved 267 / skipped 28 / errors 40（+ Scale 源修复补抓 14）；blocked_hosts：blocksandfiles.com、docs.redhat.com、e.huawei.com、support.huawei.com、www.nextplatform.com、www.reddit.com、www.theregister.com。

## 摘要
- **本日无推理平台首发。**
- 厂商自测 · VMware AI Part 3：MLPerf Inference 5.1 Llama3.1 8B（Interactive）on VCF 9.1——推理对 vCPU/内存不敏感；Llama3 70B（TRT-LLM、TP=4、4×H200）与 HammerDB 同机无可测影响；Part 4（time-slicing 推理吞吐）待发。
- 轻 · RH MLOps edge：KServe 用 Modelcar 挂载模型（"now in MicroShift too"，无版本）；服务端 vLLM / 边缘 GGUF+llama.cpp 格式分野。

## 原文摘录

### VMware · Optimizing AI Deployments with VCF Part 3（厂商自测）
- 归档：`vmware/optimizing-ai-deployments-with-vmware-cloud-foundation-part-3-f4be665b.txt`
- URL：https://blogs.vmware.com/cloud-foundation/2026/10/05/optimizing-ai-deployments-with-vmware-cloud-foundation-part-3/

> Even when we allocated only 25%-50% of available CPU cores to the VM, the inference performance is on par with bare metal.  
> … changing the number of vCPUs while fixing the memory size, and changing the memory size while fixing the number of vCPUs, had little or no impact on the performance of AI workloads that use GPUs for acceleration.  
> The HammerDB workload runs at 1.8 million TPM … The database server VM runs at 75% utilization on 12 vCPUs. The LLM-inference uses 1 vCPU. The results show that neither workload experiences a measurable performance impact.