ainewsblitz.com

Breaking

NVIDIA Slashes DeepSeek-V4 Pro Startup From 8 Minutes to Under 2 Using GPU-to-GPU RDMA

NVIDIA Slashes DeepSeek-V4 Pro Startup From 8 Minutes to Under 2 Using GPU-to-GPU RDMA
  • Infra & Chips
  • Foundation Models
  • Open Source

NVIDIA says it has cut the worker startup time for the massive DeepSeek-V4 Pro model from about eight minutes to under two by streaming model weights directly between GPUs over RDMA, bypassing traditional storage bottlenecks. The improvement was achieved with NVIDIA ModelExpress (MX), the weight-distribution and cache-management service inside the company's open-source distributed inference framework, NVIDIA Dynamo.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.

$20
Read this article
$29/month
Unlimited — all 27,934 articles, the full archive, and comprehension quizzes
Save 72%
$98/year
≈ $8.17/month
Unlimited, billed once a year