Recent content by johanpmeert

When using 4 sticks of ram, Always look at the QVL memory list of your mobo. On there you'll only find some Kingston 4x32GB sets @ 5200 speed. So that will be the max. that works. You're even lucky that sets of 4 dimms are even supported.

real-world wired 10gbe can do speeds very close to its maximum, and it can do it while all other users are also getting the same speeds. Wireless has great theoretical speeds but if you get to 25% of that in real-world you can consider yourself lucky. And when 50 competing users and 5 competing...

Nobody cares about wifi 7 for a desktop. 10gbe nic and Thunderbolt 4 on the other hand, THAT's what's important.

Someone fixed it. Do: cd repositories/GPTQ-for-LLaMa git reset --hard 468c47c01b4fe370616747b6d69a2d3f48bab5e4 pip install -r requirements.txt python setup_cuda.py install

Yeah, it's broken and that after only a few days. I tried editing line 56 in the python code and just added another parameter which clearly is a integer number. It seems that the parameter must be a power of 2 so you can enter 1, 2, 4, 8,16 and so on. Then that error disappears but the model...

Exactly. And add a GC-MAPLE RIDGE card and you get 2x TB4 ports. Since it has to be slotted in the chipset slot, it's far enough from the graphics card to not impede the cooling. Note to editor: there's a mistake in the specs of the mobo on top: it lists a 2.5Gbe NIC

If I'm not mistaken, I can see dozens of perfectly good 1200W platinum PSU being rolled over. What a waste.