sabareesh

Joined 9 years ago

Karma

323

Submissions

221

Member since

Apr 12

CTO @ guidedchoice.com , 3nickels.com . Making finance easy for everyone

View on Hacker News

Recent submissions

Comment · 4 months ago

I am pretty much doing the same but running the coolant at 40 deg C instead of 45 as my pumps are rated for 45 C max temp. Here is bit more about my setup https://sabareesh.com/posts/blackwell-waterblock/

Comment · 4 months ago

Yes this has been on my mind as well. But this was built one at a time but still overall very happy with them

Comment · 4 months ago

Appreciate the feedback. I have improved the article.

Comment · 4 months ago

Appreciate the feedback. I have improved the article.

Comment · 4 months ago

I am primarily experimenting on post training stack. As of now working on training a model that is natively RLM https://github.com/alexzhang13/rlm

Comment · 4 months ago

It is basically on 2 different circuits/breakers. Asus wrx90e supports 2 psu as well. You may need to synchronize both psu and several adapter for this is available in Amazon. Soon planning to upgrade it to 240V

Comment · 4 months ago

Not sure what really happened but some force or bad solder caused it.

Comment · 4 months ago

Most of the training i am working on is with post training. You can do so much with a system that is running 24/7

Comment · 4 months ago

Sure 140mm fans you may call little but it does need enough static pressure for the radiators. This setup is already several times quieter than stock setup

Comment · 4 months ago

Converting four RTX PRO 6000 Blackwell cards to waterblocks, finding a VRM choke loose on the workbench, and getting back to 41k tok/s.

4× RTX Pro 6000 Blackwell on Water, and the One Card That Wouldn't Behave sabareesh.com

45 sabareesh 4 months ago 53 comments

Comment · 6 months ago

Sounds like speculative decoding but for KV cache

Comment · 6 months ago

Based on last few attemts on claude code to address a docker build issue this feels like a downgrade

Comment · 7 months ago

Codex usually dont add itself as contributor so this is misleading .

Comment · 7 months ago

I wish it provided benchmark comparing Direct RAM offload vs CPU offload vs Full VRAM