I built my 2x3090 with parts from eBay… MB (x299 Giga), i9 CPU, 64G RAM and 2 3090s… I did spring for a new, heavy duty PS and case with big fans.

All in, I spent about $2k.

System runs 70B models like Llama-2-70B-Orca-200k just fine at 11 T/s…

I feel like there’s not a ton of downside - I think the 3090s will be valuable for a while yet, and that’s over half the value of the system.

Having the hardware right here means I can have thing running all the time - when I read about a new model, I can download, play, etc. in minutes. Spinning up a runpod feels frustratingly slow to me. I went that route for a while, but found that the friction involved meant I tried fewer things. Having a system that might be slower, but is always available just works for my way of working.

So no “regerts” here.

Nope. Have a 2x3090 system and planning to buy another 3090 system to be able to do SD LoRA training while being able to use Dolphin 70b.

120 thousand rubles.

I was an idiot when assembling the PC and somehow inexplicably focused on the processor when assembling the PC, and the video card is quite weak. However, after a while, I realized that this was for the better. I can use the 70B model with 1 token per second. Maybe in the future I will buy another video card so that more layers will help with data processing

3060 12 & 13600K

Winter has enveloped us in its chilly embrace. In my quest for warmth, I realized I needed a heater. But then, a memory dawned on me – I had bought one before! It was during those long nights of training a model with a 100k dataset, which made the room toastier. Now, thanks to that, everyone in this house can enjoy a peaceful and warm winter.