Virtual GPU (vGPU) stories
Enterprises can now keep AI workloads and sensitive data on the same private cloud, as Broadcom adds governance and security controls.
Interrupted batch workloads can now resume where they stopped, potentially cutting wasted compute and delays for Google Cloud customers.
Certification could help cloud operators share costly GPUs more efficiently, with k0rdent AI coming within 5% of bare-metal performance.
The deal could let enterprises serve more AI models on fewer GPUs by shifting cache data into DRAM, SSDs and remote storage.
Distributed sites will get tighter controls as HPE adds AI prompt filtering, recovery and encryption updates to guard against data leakage and attacks.
vGPU software creates virtual GPUs that can be shared and allocated between multiple virtual machines for remote workers.
Businesses running AI workloads may now keep costly GPU servers online longer, as Park Place extends third-party maintenance to 60 data centres worldwide.