ZW LLC demonstrated up to 111% higher throughput in selected ongoing AWS GPU cluster validation tests using Z-infDC.
Public source
Publisher name
Public post
Can AI Data Centers Significantly Increase Inference Capacity Without Adding More GPU Infrastructure? ZW has developed Z-infDC™, a turnkey supervisory and control softwa…
Company
ZW LLC
Z-infDC: Double AI Inference Throughput Without Adding GPU Hardware or Expanding Infrastructure
- Location
- Chapel Hill, US
- Company size
- 11–50 employees
About ZW LLC
ZW is an innovative AI/ML software startup based in Chapel Hill, North Carolina. We are transforming how data centers deliver AI inference services through Z-infDC®, a turnkey, SCADA-inspired supervisory and control platform. In selected AWS GPU validation tests, Z-infDC® demonstrated up to 110% higher AI inference throughput* using the same underlying GPU hardware and infrastructure. Our core team brings decades of real-world experience in AI/ML, LLMs, high-performance computing, software development, optical networking, high-power heating and cooling systems, and complex industrial SCADA architectures for managing massive, continuous manufacturing operations. Our Flagship Product: Z-infDC® Z-infDC® integrates advanced LLM optimization techniques, GPU systems, AI serving engines, orchestration software, and real-time supervisory control into one cohesive platform. It gives data center operators a faster, more effective path to expand AI inference capacity, meet rapidly growing customer demand, and increase revenue and profit potential without the time, cost, and complexity of developing and integrating fragmented software systems themselves. *Based on selected internal validation tests conducted on AWS GPU infrastructure. Performance varies by model, workload, hardware topology, concurrency, parallelism, and serving configuration.
See moreDiscover more
Similar signals
Similar public activity from other companies.
Research & Knowledge
AI Accelerator Institute
AI Accelerator Institute published a report breaking down the 5 most common reasons why GPU utilization is below 70% at peak load, with a focus on storage setup.
Research & Knowledge
Amazon Web Services (AWS)
Amazon Web Services (AWS) benchmarked model loading on p5.48xlarge and found the bottleneck flips with model size.
Research & Knowledge
FarmGPU
FarmGPU published results for the fastest 8-node WEKA cluster using Solidigm D7-PS1010 and MLPerf Storage v3.0
Research & Knowledge
CodeZero
CodeZero explored how compute workloads behave using the Liquid Immersed Z4R system.
Research & Knowledge
CoreWeave