Morrowmake

Do not pay $12,000 for 2 x DGX Sparks.

For the same price you can buy 4 x 170HXs and a computer to run them in, for FAR better performance.

Should I post complete builds at various price points?

If this is real...

> be me
> buy 8 old NVIDIA mining GPUs, 8GB each
> turns out they unlock to 64GB PCIe 2
> turns out they actually unlock to 80GB PCIe 4
> turns out they're faster than A100s
> my $12,000 purchase is worth $80,000
> 640GB of data centre GPUs now sit in my living room

blog.kate.land/posts/using-vo…

HOLY CRAP LETS GOOOOOOOOOOOO

170HX bros its TIME

I am really starting to feel the hardware bottleneck bite.

I can only work on a single inference project at a time, and I have days worth of work queued and waiting to run on my GPUs. It's never-ending.

I am now leaning towards not dropping a Qwen 4 27b recipe on release day, so that I can keep working on something else that I think will be more valuable in the long run. There are other people making recipes for these cards now so hopefully this won't really be felt.

It's frustrating but unless I can get my hands on more hardware I need to choose what I allocate my resources to.

Anthropic models are simply on another level right now.