Nav

COO at Digital Storm | High-Performance Workstations, AI Compute & PC Hardware Infrastructure

[Nav @DigitalNav]

@orenoa7 I'm curious, how much are you paying per model?

[Nav @DigitalNav]

NVIDIA and AMD are taking different approaches to AI infrastructure, but both are moving toward open hardware standards through OCP.

NVIDIA is contributing its Vera Rubin rack and compute tray designs, focusing on cooling, power delivery, and easier deployment.

AMD's Helion targets agentic AI with CPU-heavy and hybrid CPU+GPU rack configurations, including designs with up to 32,768 EPYC cores.

As AI infrastructure scales, the focus is expanding beyond GPU performance to how entire systems are designed, built, and maintained.

[Nav @DigitalNav]

@Hikari_07_jp 20% faster on the same 43k-token task is a nice improvement. Did the change affect answer quality or just decoding speed?

[Nav @DigitalNav]

@TeksEdge I’d love to know how it feels with everyone using it at once especially latency under load.

[Nav @DigitalNav]

@jun_song I hear this every cycle, but the useful question is what you’d actually build with the hardware.

[Nav @DigitalNav]

@Hikari_07_jp A GPU is an oddly effective way to turn curiosity into a project.

[Nav @DigitalNav]

@seanhighness How are you planning to handle airflow across the cards?

[Nav @DigitalNav]

@mfranz_on I’d want to see model speed at long context before calling it the best choice.

[Nav @DigitalNav]

@_mctrinh Hopefully a domain that doesn’t make me explain why the site is about AI.

[Nav @DigitalNav]

@techguy_x Codex for me.

[Nav @DigitalNav]

@TheAhmadOsman Even the robots need a timeout when they’re taking too long.

[Nav @DigitalNav]

@ivanfioravanti Getting a model to run is one thing; keeping the whole setup secure and reliable is another.

[Nav @DigitalNav]

@LottoLabs Feels that way when I catch myself reaching for AI before I’ve even finished forming the question.

[Nav @DigitalNav]

@omkarships The tools can get you to a prototype but knowing what’s worth building still matters.

[Nav @DigitalNav]

@not_sacke The bigger question is whether I’d still know how to debug the stuff I let AI write.

[Nav @DigitalNav]

@_mctrinh Knowing when an answer is wrong.

[Nav @DigitalNav]

@johnHarry1018 You need a solid business to keep building the AI.

[Nav @DigitalNav]

@LottoLabs I’m trying to keep my expectations in check until we see how it runs.

Are you hoping it actually improves the most?

[Nav @DigitalNav]

@plotarmordev 24GB on a 5080 would be a really interesting option for local AI. Curious what the memory bandwidth looks like, if this leak pans out.

[Nav @DigitalNav]

@jun_song An old machine you already own is a pretty great place to start. You don't need to always buy the newest.