Product Updates
Changelog
Track the latest GPU hardware rollouts, kernel optimizations, and platform improvements across HyperTensor.
v2.4.0
October 15, 2024
LatestNVIDIA Blackwell B200 General Availability
B200 clusters are now available across US-East, EU-Central, and AP-South regions with native FP4 tensor core support.
- FeatureAdded support for NVIDIA B200 8-node and 32-node bare-metal topologies.
- ImprovementOptimized InfiniBand RDMA routing table propagation speed by 40%.
- FixResolved intermittent mount timeouts on high-capacity NVMe ephemeral volumes.
v2.3.1
September 30, 2024
Enhanced Kubernetes CRDs & Autoscaling
Version 2.3.1 introduces advanced queue management and priority-based preemptible GPU scheduling.
- FeatureIntroduced QueueConfig CRD for batch job prioritization.
- ImprovementReduced container snapshot hydration memory footprint by 18%.
- FixPatched potential memory leak in Prometheus metrics sidecar exporter.
v2.2.0
September 10, 2024
Confidential Computing Extensions
Hardware-enforced encryption for memory enclaves on all H100 serverless instances.
- FeatureEnabled NVIDIA Hopper confidential computing mode toggle in CLI and UI.
- ImprovementStreamlined attestation report generation for compliance auditing.