Ads skipped

The Only Guide You Need for Local AI Hardware

28K views · Sep 30, 2026 · Education

Comments · 28

  • @Hikove · 8 hours ago

    This format should be the norm, including a segment for the ADs and done, not pivot the whole video around a sponsor.

    2

  • @CodeTrix22 · 12 hours ago

    Your content has been on fire lately 🔥

    3

  • @blayral · 5 hours ago

    Memory speed has a twist for real numbers : the memory controler is why the numbers differ

  • @rickyu1978 · 11 hours ago

    really clear and easy to understand, no fluff.

  • @frankhocza6758 · 26 minutes ago

    Awesome and clear information. Thanks :)

  • @Kantuski · 11 hours ago

    Great video and explanation. &nbsp;I build using older technology that I have had for years and the satisfaction I feel is when I see 15-18 t/s on a bigger model and then going to a smaller model gets me 55-62 t/s. I run Qwen3.8-27b at around 26 t/s and am happy, but 3.8-9B runs at 55 t/s, that’s my joy.<br>Your explanation makes sense and helps me understand why I love technology so much. 😁

  • @tomislavvukojevic2479 · 6 hours ago

    Full precision it’s 32 bit it’s about four time more than Q8 <a href="https://www.youtube.com/watch?v=x6siPRkPAeQ&amp;t=63">1:03</a>

    1

  • @pushpendratripathi46 · 11 hours ago

    Macbook air m5 and m1 are all I have currently

    1

  • @CapCrunch45 · 11 hours ago

    Hey Tim, I know this is a bit off-topic, but will there be a new video in the Rust tutorial playlist where you teach about ownership? I really enjoy your coding tutorials, and I’d like to learn more how to code manually in Rust (even in the age of AI).

    2

  • @SuperGrimmy · 5 hours ago (edited)

    The most affordable option right now if you have parts for a linux server. Install Ubuntu 26 and throw in a Radeon AI PRO R97000 32G. Works out of the box with ollama. I think almost anyone can set that up in no time. This assumes 32GB is enough for your use case. Expanding to two cards is an option, but you&apos;ll need to find a 8x/x8 motherboard. That gives you 64GB and that&apos;s the practical cap for that setup. It all depends on your budget, use case and needs. Mine are more or less running 24/7, something that can be a problem for consumer cards.

  • @iPaintCars_215 · 2 hours ago

    For anyone thinking ryzen, look for something with lpddr5 that’s soldered on. Its bandwidth speed is higher than standard ddr5 and you’ll need every bit of it… learned the hard way. Although 15-20 tok/s is usable regardless of what everyone will have u believe

  • @jointdr.hoffman5991 · 3 hours ago

    Hey Bro,<br>Appreciate U 4Great info provided.<br>Let us know on how it behaves in relation with that Mac mini M6 + 32 gb

Up next

LIVE

Understand Local AI in 15 Minutes

Tech With Tim · 52K views

LIVE

The Current State of Siri AI

SAMTIME · 75K views

LIVE

MCP Servers Explained & Built

Tech With Tim · 63K views

LIVE

M5 Ultra Mac Studio Review: Goodbye Claude?

Mark Ellis Reviews · 74K views

LIVE

What If the Antichrist Is AI?

Clue Fog · 57 views

LIVE

How I Actually Use Claude Code to Build Real Apps

Tech With Tim · 74K views

LIVE

How Delivery Apps Charge You More (For Worse Food)

GEN · 2.4K views

LIVE

Why iPhone users never switch

Michael Girdley · 31K views

LIVE

EU Tells USA To BACK OFF! Dutch BAN Microsoft, Bonds COLLAPSE, Trump PANICS

House of El · 219K views

LIVE

My Dream Power Hammer Restoration! Part 10

Alec Steele · 912K views

LIVE

The Closest Thing We Have to Alien Technology

Veritasium · 59M views

LIVE

Bill Gates: A.I. ‘Makes Nuclear Weapons Look Like Nothing’ | The Ezra Klein Show

The Ezra Klein Show and 2 more · 615K views

LIVE

All hope is NOT lost... Affordable PC Gaming?

JayzTwoCents · 141K views

LIVE

How to Build AI Agents in Python - 3 Ways

Tech With Tim · 46K views

LIVE

The SHOCKING Truth About The Pyramids No One Tells You

The Diary Of A CEO Clips · 4.7M views

LIVE

Something BIG Just Happened Over Lithuania… NATO Takes EMERGENCY ACTION

The Frontline Show · 123K views

LIVE

OpenAI PULLS MODEL, HALTS TRAINING As Danger LOOMS

Breaking Points · 208K views

LIVE

I Tested ChatGPT Dot vs GrokBot (Not Even Close)

Jack Roberts · 7.5K views

LIVE

🚨 PLÖTZLICH ESKALIERT DIE RUNDE! Lisa Eckhart sagt EINEN SATZ – PUBLIKUM RASTET KOMPLETT AUS!

Scharfer Blick · 39K views

LIVE

The Global Debt Crisis Has Begun (Why Gold Is Falling First)

ITM TRADING, INC. · 584K views

YouTube, with the door locked.

Aegis plays a clean stream instead of YouTube's player, so pre-roll ads, trackers, and fingerprinting never ride along. Drop Shields any time if you want the official player back.