Get the SupremeRAID™ Summer 6-Pack! Buy 5 licenses, get the 6th FREE / Read the Storage Tax Blog
Contact Us
Video
November 4, 2025
Storage Review: RAID for AI on Dell PowerEdge: Faster, Safer, and Slot-Smart
In this webinar, Storage Review breaks down how Dell's latest PowerEdge platforms, dense NVMe storage, and SupremeRAID™ AE work together to keep GPUs fed and data protected without giving up PCIe slots. We cover real configs from the lab, why most inference lives on 1U and 2U systems, how GPU Direct Storage moves data SSD to GPU, and what rebuild performance and thermal realities look like at scale.
Featuring:
• Brian Beeler, StorageReview
• Kevin O'Brien, StorageReview
• Kelly Osburn, Graid Technology
• Seamus Jones, Dell Technologies
What you will learn:
• PowerEdge backplane options and NVMe density, including E3.S and U.2 layouts
• When and why to run RAID on the GPU you already have with SupremeRAID™ AE
• How GPU Direct Storage cuts latency by bypassing CPU and host memory
• Rebuild behavior, degraded mode performance, and uptime impact
• Slot economics, networking headroom, and airflow considerations for AI servers
01:20 PowerEdge Gen17 flexibility: CPU options, backplanes, drive counts, PCIe
03:10 Storage density in practice: 32 to 44 NVMe in 2U, E3.S focus
04:30 AI in real servers: why most inference lands on 1U and 2U with 1 to 4 GPUs
06:10 Single namespace and protection needs for AI datasets
07:10 Graid SupremeRAID AE explained: run on existing GPU, tiny GPU slice, time division multiplexing, no extra slot
09:00 GDS path overview: SSD to GPU direct, bypass CPU and host memory
10:40 Rebuild rates and degraded mode performance
12:40 ROI angle: utilization, avoiding CPU RAID overhead and PCIe lane chokepoints
14:10 Operator view: provisioning workflow and API control
15:20 Lab result callout: ~4 percent TPS drop during active data staging
16:20 Edge and thermal realities: airflow, slot pressure, rugged platforms
18:10 NVMe-oF and scale out: larger drive counts, future erasure coding
19:40 Reference builds and validated configs
20:40 Wrap, links to paper, charts, and the long hardware video
News & Resources
September 14, 2026
Breaking the KV Cache Inference Wall — Solidigm Data Insights Podcast, Featuring Graid Technology
"You've got a four-lane highway, and to get to your storage, you have to go down to a one-lane bridge." Garrett McKibben joins Solidigm on TechArena's Data Insights podcast to unpack why KV cache overflow is choking AI inference.
.jpg)
Graid Technology CEO Leander Yu and VP of Product Development Alven Yen explain how moving RAID processing onto the GPU breaks the bottleneck traditional controllers cannot solve, delivering 36 million GPU-initiated IOPS for AI infrastructure.
.jpg)
Join Graid Technology at Dell Technologies Forum Taipei 2026 and discover how SupremeRAID™ AE transforms local NVMe storage into a high-performance, scalable, and protected extended KV cache—reducing redundant computation and accelerating time to first token (TTFT).