StorageReview.com

Windows Server finally gets native NVMe. By removing the legacy SCSI translation layer, Windows Server 2025 lets modern NVMe SSDs run more efficiently with lower latency and better performance. storagereview.com/review/windows… #NVMe #WindowsServer #Storage @Microsoft

Sizing math: 128GB SRAM per rack across 256 chips. GPT-OSS 120B? One rack. DeepSeek R1? ~5 racks. GLM-5 in BF16? 10+ racks. The front-panel C2C ports chain racks for models that exceed single-rack capacity. Shipping H2 2026. Full deep-dive with diagrams and topology analysis: storagereview.com/news/nvidia-gr… @NVIDIAAI

This also solved Groq's pre-acquisition bottleneck. Standalone Groq struggled with long-context models because KV cache ate into LPU capacity. Split the workload: attention on Rubin's 288GB HBM4, FFN on Groq's 150 TB/s SRAM. The LPU never touches the KV cache.

🧵 GTC 2026 is in the books. Here's everything you need to know about the most interesting piece of the Vera Rubin platform: NVIDIA Groq 3 LPX.

For the first time, end-to-end hardware-based in-flight encryption covers the entire Fibre Channel data path from server to array. Emulex SecureHBA is now integrated into Everpure FlashArray. storagereview.com/review/emulex-… @Everpure @Broadcom #FibreChannel #SAN #storage #security

We tested the Dell Precision 7875 with a 96-core Threadripper PRO 9995WX and dual RTX PRO 6000 Blackwell Max-Q GPUs across rendering, AI inference, and LLM serving. 192GB of GDDR7 changes what a single tower can do. storagereview.com/review/dell-pr… @Dell @AMD @NVIDIAAI

We reviewed the Ubiquiti Flex 2.5G 8 PoE. Eight 2.5GbE PoE++ ports, 10GbE uplink, and runs entirely off a single PoE+++ cable or a 210W AC adapter for up to 196W of PoE budget. Solid edge switch for $199. storagereview.com/review/ubiquit… @Ubiquiti #networking #UniFi #PoE

NVIDIA DGX Rubin NVL8 supports Intel Xeon 6 as a host CPU option for x86-based deployments. 8TB memory, 2.3x MRDIMM bandwidth, PCIe 5.0, TDX confidential computing, and NVIDIA Dynamo support. storagereview.com/news/nvidia-dg… @Intel @NVIDIA #AI #Xeon #GTC2026

NVIDIA Groq 3 LPX at GTC 2026: each LPU hits 150 TB/s memory bandwidth via SRAM vs ~22 TB/s on Rubin's HBM4. That gap is why LPUs offload FFN layers from Rubin GPUs to boost token throughput. 8 LPUs per 1U unit, 315 PFLOPS across the full 256-chip rack. Available 2H 2026. @NVIDIA #AI #Groq #GTC2026

NVIDIA pulled two cold plates at GTC 2026 to show what's underneath the Vera CPU board. Each board carries 2 Vera CPUs. Per CPU: 88 custom Olympus cores with SMT, 1.2 TB/s memory bandwidth, 1.5TB SOCAMM LPDDR5X, and 1.8 TB/s NVLink-C2C. The full Vera CPU Rack is a liquid-cooled MGX system scaling to 256 Vera CPUs, 400TB of LPDDR5X memory, 64 BlueField-4 DPUs, Spectrum-X Ethernet, and 2x performance per watt versus x86. Purpose-built for reinforcement learning workloads. @NVIDIAAI #AI #VeraCPU #datacenter #GTC2026

KIOXIA's GP Series uses XL-FLASH to let GPUs directly access flash as an HBM extension via NVIDIA Storage-Next. Also: a 25.6TB CM9 PCIe 5.0 E3.S drive for KV cache in AI inference systems. Both due by end of 2026. storagereview.com/news/kioxia-gp… @KIOXIA @NVIDIA #storage #SSD #AI

HPE at GTC 2026: Vera Rubin NVL72 for frontier-scale AI, a new Cray GX240 blade with up to 640 Vera CPUs per rack, and the XD700 doubling GPU density at 128 Rubin GPUs per rack. RTX PRO 6000 Blackwell available now. storagereview.com/news/hpe-cray-… @HPE @NVIDIA #AI #HPC #GTC2026

Lenovo expanded Hybrid AI Advantage with NVIDIA at GTC 2026 across client, enterprise, and cloud. Highlights: 8x lower cost per token vs. cloud IaaS, RTX PRO Blackwell inference servers, and Vera Rubin NVL72 as a launch partner. storagereview.com/news/lenovo-ex… @Lenovo #AI #GTC2026

Dell launched the AI Data Platform at GTC 2026: GPU-accelerated data engines, Lightning File System, and Exascale Storage. Built to stop GPUs from sitting idle waiting on data. storagereview.com/news/dell-expa… @DellTech #storage #AI #GTC2026

Micron is now in volume production on HBM4 (36GB 12-high), a 192GB SOCAMM2, and the 9650 PCIe Gen6 data center SSD, all targeting NVIDIA Vera Rubin. The post-Blackwell memory and storage stack is taking shape. storagereview.com/news/micron-be… @MicronTech @NVIDIA #storage #AI

Everpure is aligning FlashBlade//EXA with NVIDIA AI Factory and previewing Data Stream, an automation layer for moving data through AI pipelines. storagereview.com/news/everpure-… @PureStorage @NVIDIA #storage #AI

NVIDIA GTC 2026: Jensen Huang unveils Vera Rubin NVL72, Groq 3 LPX decode acceleration, Vera CPU Rack, and Dynamo 1.0, a full-stack AI factory platform built for agentic AI at scale. storagereview.com/news/nvidia-gt… @NVIDIAAI #gtc26 #ai #nvidia

Gen5 SSD performance in a 2242 form factor. We tested the Corsair MP700 MICRO 4TB, delivering ~10GB/s reads and 4TB capacity in a compact M.2 2242 drive. Benchmarks include FIO, LLM load times, and GPU Direct Storage. storagereview.com/review/corsair… #ssd #corsair #storage @CORSAIR