Does PCIe matter much for inference?

<!-- SC_OFF --><div class="md"><p>Motherboard: Asus x370 Taichi</p> <p>PCIE1: PCIe 3.0 x16 (runs at x16 single, or x8 when PCIE3 is also populated)</p> <p>PCIE2: PCIe 3.0 x16

<!-- SC_OFF --><div class="md"><p>Used to to do eth and other cryptomining where all attached GPUs with a 1x pcie cable, powered pcb adapter was sufficient as it was just data

<!-- SC_OFF --><div class="md"><p>I am currently on an m3 mbp with 24gb ram. For regular python and django work the machine is perfect and i have no need to upgrade for speed. but

<!-- SC_OFF --><div class="md"><p>Hey everyone, I've been working on getting Multi-Token Prediction (MTP) working with quantized GGUFs for Qwen3-27B and the results are pretty

<!-- SC_OFF --><div class="md"><p>I saw a couple of these motherboards with 4x pcie 3.0 x16 slots paired with a 2nd gen threadripper 2920x/2950x for like $350 and I was wondering

<!-- SC_OFF --><div class="md"><p>I thought you guys would be likely to know a direction for me to go on this issue. </p> <p>I have a cheap Frankenstein build, Lenovo p520 with

<!-- SC_OFF --><div class="md"><p>So I got an MI-50 that I use for my local inference but I recently discovered that it is connected with PCIe 3 speeds to my motherboard. Even