https://debene.dev/tags/benchmark/ Recent content in Benchmark on debene.dev https://debene.dev/images/og-default.png https://debene.dev/images/og-default.png Hugo -- 0.152.2 en-us Thu, 14 May 2026 00:00:00 +0000 https://debene.dev/posts/apple-silicon-vs-power8/ Thu, 14 May 2026 00:00:00 +0000 https://debene.dev/posts/apple-silicon-vs-power8/ <h1 id="apple-silicon-vs-ibm-power8-a-tale-of-two-architectures-running-llms-in-2026">Apple Silicon vs IBM POWER8: A Tale of Two Architectures Running LLMs in 2026</h1> <p>Last week I published benchmarks of <a href="https://debene.dev/posts/power8-llm-2026/">running Qwen 2.5 7B on a 2016 IBM POWER8</a>. The results were surprisingly good — 6.81 tokens/s on CPU-only inference with 80 threads hammering away.</p> <p>But then came the inevitable question: <strong>How does it compare to modern hardware?</strong></p> <p>So I ran the same benchmarks on my daily driver: a Mac Studio with Apple M2 Max. Same model (Qwen 2.5 7B Q4_K_M), same quantization, different decade. Here&rsquo;s what I found.</p>