
How Much GPU Memory Do I Need?
So far in this series we've met the two big things that live in a GPU's memory. In Part 1 it was the...
Tag archive

So far in this series we've met the two big things that live in a GPU's memory. In Part 1 it was the...

One 4.5 GB model hears, sees and speaks on about 5 GB of GPU. The traps: silent video, a text-only fast server, and numbers read aloud.

LLM...
Billing granularity, tiers, storage and idle time can change a GPU rental bill more than the headline hourly price. Examples from provider pages, 15 November 2026.

No. Each vendor exposes a different counter set and defines GPU load differently, so the same reading...
kimi-k3-in-c proves a giant model can run on a CPU. It also proves, in the negative, why nothing replaces a GPU: the limit is physical, not software.

Separate stock PyTorch layers from custom kernels, then check numerics and speed against the run you already trust.

Building camera analytics for retail taught me that the hard part was everything around the model, from what retailers needed to why our GPU sat at its power cap.

Nanbeige4.3-33B ran at about 42 tokens/s on one RTX 4090 with llama.cpp. Why offload fails for dense models, and the empty-reply bug to watch.
The Genesis of 'Rent = Root': Unpacking CVE-2026-47591 In high-density neocloud...
Vertex AI alternatives for custom training: one 24-hour 8 x H100 run priced on seven platforms, what you lose by leaving, and how to keep GCP credits.
A practical Runpod to Nodus migration for training: what each Runpod concept becomes, the commands, checkpoint carry-over, cost and what to leave behind.