Everpure demonstrated near liner growth with MLPerf Storage v3.0, topping both checkpoint write and KV cache read tests.
Public source
Publisher name
Public post
New GPUs, new bottlenecks, new benchmark. MLPerf Storage v3.0 simulates B200 accelerators rather than H100s, and adds checkpointing and KV cache tests. Everpure //EXA to…
Company
Everpure
- Industry
- Technology, Information and Internet
- Company size
- 5,001–10,000 employees
About Everpure
We are Everpure. We don’t just store data—we bring it to life. As AI becomes core to business, infrastructure must evolve, yet too many organizations are still relying on rigid, siloed systems built for the past. The Everpure Platform replaces static storage with a unified, living platform that evolves with you—agile, resilient, and built to scale—so you can eliminate friction and unleash the power of your data. Learn more at www.everpuredata.com #Everpuredata
See moreLatest activity
Latest activity from Everpure
54 signals
Presence & Recognition
Everpure is hosting an Ask Us Everything webinar exploring AI and automation capabilities built into Pure1 for storage operations.
Presence & Recognition
Everpure is exhibiting at Booth #1254 at the Santa Clara Convention Center from September 15 to 17, 2026, to showcase AI infrastructure and help organizations move faster, perform better, and scale AI.
Partnerships
Everpure and Microsoft formed a partnership to deliver a trusted foundation for data.
Discover more
Similar signals
Similar public activity from other companies.
Products & Services
DataCore Software
DataCore Software released v3.0 of MLPerf's storage benchmark, adding tests for checkpointing, vector databases, and KV cache on top of original training workloads.
Products & Services
MLCommons
MLCommons released MLPerf Storage v3.0, the first industry benchmark to measure AI storage across four workloads including training, checkpointing, vector database retrieval, and KV cache serving.
Products & Services
Nirvana Labs
Nirvana Labs benchmarked quirq against GKE and E2B, recovered all 5,501 committed checkpoints after an ungraceful kill, and released idle compute in 1.46 seconds.
Products & Services
Lightbits Labs
Lightbits Labs optimized the kvcache with Inferra, achieving zero hardware changes and breaking the GPU memory wall with prefetch instead of refetch.
Products & Services
Pure Storage