3 ms·

Indeed. RIP Anandtech. I've seen bandwidth tests since then that showed similar for newer generations, but not the m4. Not sure if the common LLM tools on m

by sliken 2y ago

Indeed. RIP Anandtech. I've seen bandwidth tests since then that showed similar for newer generations, but not the m4. Not sure if the common LLM tools on mac can use CPU (vector instructions), AMX, and Neural engine in parallel to make use of the full bandwidth.