memory for
maximum uptime

Zefr eliminates over 90% of memory reliability failures

Answering the call to workload-driven computing, powerful, High Performance Computing (HPC) platforms are processing massive amounts of data to unlock value.

This trend demands maximum uptime.

Industry Standard
Memory Reliability
isn’t Sufficient

Current Defective Parts per Million (DPPM) simply aren’t low enough, contributing to:

  • Delayed platform bring up
  • Inflated maintenance costs
  • Reduced system yield rate

The Zefr Process
Simulates Real-World Conditions

Zefr (Zero Failure Rate) is a screening process performed on OEM original memory modules or SMART Modular built memory modules to deliver ultra-high reliability for demanding workloads.

The Zefr Process
Simulates Real-World Conditions

Zefr (Zero Failure Rate) is a screening process performed on OEM original memory modules or SMART Modular built memory modules to deliver ultra-high reliability for demanding workloads.

Zefr is a Proven Process with Over 3 Million Memory Modules Shipped

With Zefr Memory, our High Performance Computing System came up fast, completed the acceptance-criteria period flawlessly, and has consistently operated reliably, running mission-critical jobs without disruption.

William Wu
VP Products, Penguin Computing

Zefr Benefits

  • Increase ROI
  • Maximize System Yield Rate
  • Accelerate the "Time-to-Insight"

Case Study

An HPC System Integrator built identical systems with standard and Zefr memory

Standard Memory

Purchase 18,384
Standard RDIMMs

Zefr™ Memory

Purchase 18,384
Standard RDIMMs

Build Cluster A:

  • 1,532 Nodes
  • Twelve 16GB RDIMMs per Node

Build Cluster B:

  • 1,532 Nodes
  • Twelve 16GB RDIMMs per Node

Field Failures since Platform Bring up:

360 Failures

Field Failures since Platform Bring up:

0 Failures

Zefr™ Memory

Purchase 18,384
Standard RDIMMs

Build Cluster B:

  • 1,532 Nodes
  • Twelve 16GB RDIMMs per Node

Build Cluster B:

  • 1,532 Nodes
  • Twelve 16GB RDIMMs per Node

Standard Memory

Purchase 18,384
Standard RDIMMs

Build Cluster A:

  • 1,532 Nodes
  • Twelve 16GB RDIMMs per Note

Field Failures since Platform Bring up:

0 Failures

SMART Products with Zefr Processing