StorageReview.com

Enterprise infrastructure since 1998. Storage, servers, networking, AI. Independent reviews, lab benchmarks, breaking news from the data center.

[StorageReview.com @storagereview]

https://t.co/ZmBjYdxysI

[StorageReview.com @storagereview]

NVIDIA quietly made Omniverse free for development, production, and redistribution, dropping the NVIDIA AI Enterprise requirement that listed at $4,500 per GPU per year. Support moves to a paid add-on.
https://t.co/7nAoyuBfHP
#NVIDIA #Omniverse #DigitalTwins @NVIDIAAI

[StorageReview.com @storagereview]

Intel's Core Ultra 7 265K (20 cores, no Hyper-Threading, up to 5.5GHz) streets around $300 and takes the single-thread lead in Cinebench R23 at 2,267 while trading blows with AMD X3D chips that cost 1.5x to 3x more.
https://t.co/RGX1fP7GfU
#Intel #ArrowLake #PCBuild @intel

[StorageReview.com @storagereview]

Sandisk and Kioxia are sampling BiCS10, their 332-layer 1Tb TLC 3D NAND. It hits over 29 Gb/mm2, 59% denser than BiCS8, with a 4.8 Gb/s interface (33% faster) to feed PCIe Gen6 SSDs. Same JV node.
https://t.co/DGebyT081b
#NAND #SSD #Storage @KIOXIAAmerica @SanDisk

[StorageReview.com @storagereview]

@AMD We'll be there, should be an amazing event as usual! Can't wait to see what's next from AMD.

[StorageReview.com @storagereview]

Couchbase's AI Data Plane is GA, folding agent memory, vector search, caching, and document storage into one governed layer for production AI agents, validated with LangGraph, CrewAI, and LlamaIndex. More...
https://t.co/2grB6HploH
#AI #Database #DataEngineering @couchbase

[StorageReview.com @storagereview]

No tools to pull a GPU, four M.2 Gen5 x4 slots on the board, and two pre-wired 3.5" bays. Part 2 on the HP Z8 Fury G6i gets into storage and serviceability, plus the six-fan cooling that kept it from throttling under load. #HPZ8Fury #RTXPRO6000 #workstation @ZbyHP @HP

[StorageReview.com @storagereview]

Scality and OVHcloud have built a sovereign AI storage stack that keeps sensitive data off US hyperscalers. It pairs Scality RING and ARTESCA object storage with OVHcloud's OPCP and Bare Metal, with GPU Direct support and S3-compatible backup targets for Veeam and Commvault.
https://t.co/76tAdjzrpQ
#SovereignCloud #ObjectStorage #AI @scality @OVHcloud

[StorageReview.com @storagereview]

AMD's Versal Premium Gen 2 Memory-on-Package adaptive SoCs put up to 32GB of LPDDR5X in-package for 288GB/s of bandwidth in up to 60% less board area than external memory. They also pack PCIe 6.0 and CXL 3.1 at 64GT/s.
https://t.co/MUiZ8uYvFx
#AMD #FPGA #Embedded @AMD

[StorageReview.com @storagereview]

@bentleypc @230V @ZbyHP @HP HP supports one full power at 600W or up to 4 of the Max Q at 300W, two different products.

[StorageReview.com @storagereview]

Four RTX PRO 6000 Blackwell Max-Q cards at 300W each is 1,200W of GPU before you add the Xeon and SSDs. That's why the HP Z8 Fury G6i ships with two hot-swap 1,700W (@230V) supplies. We crack it open in Part 1. #HPZ8Fury #RTXPRO6000 #workstation

@ZbyHP @HP

[StorageReview.com @storagereview]

More photos of the Dell XE7745 GPU server liquid retrofit from Jetcool! @jetcool_usa

[StorageReview.com @storagereview]

JetCool's liquid-cooled Dell PowerEdge XE7745 uses SmartPlate direct-to-chip cooling rated for 8kW per server. It cuts GPU temps up to 11%, fan power up to 50%, and total server power up to 30%, while dropping noise by up to 23 dB.
https://t.co/hGXaNFXPbe
#LiquidCooling #DataCenter #AI @jetcool_usa @Dell @NVIDIAAI

[StorageReview.com @storagereview]

Supermicro's expanded Intel edge AI lineup spans fanless units to 1U servers, supporting the full Arc Pro B-series. The B70 hits up to 367 TOPS, while a fanless Core Ultra Series 3 box pairs GPU and NPU for 180 TOPS.
https://t.co/7OpuamnFBq
#EdgeAI #Intel #AI @Supermicro

[StorageReview.com @storagereview]

@tmophoto These have more endurance, 3DWPD, with most workloads it's effectively impossible to write that much. They do also benefit from better write performance as well.

[StorageReview.com @storagereview]

We're pulling 16 DIMMs out of a Dell PowerEdge XE7740 on purpose. With DRAM near a quarter million at list, the question is how much KV cache you can push onto flash instead. Half the RAM, full bandwidth, more stress on the flash, results soon. #KVcache #PowerEdge #SolidigmSSD @solidigm

[StorageReview.com @storagereview]

Supermicro's new DCBBS Blueprint builds HPC clusters from NVIDIA Vera Rubin NVL4 scalable units: eight 52U liquid-cooled racks, 288 nodes, up to 1,152 Rubin GPUs and 576 Vera CPUs, at 362kW per rack.
https://t.co/y9i77sTKXe

#NVIDIA #HPC #AI @Supermicro

[StorageReview.com @storagereview]

NetApp StorageGRID 12.1 adds a Global Federated Namespace scaling to 10 exabytes and up to 400% higher throughput than 12.0, reaching 12 TB/s aggregate for AI workloads.
https://t.co/j4cFu7Fzw7

#NetApp #ObjectStorage #AI @NetApp

[StorageReview.com @storagereview]

IBM says it has built the first sub-1nm chip: a 0.7nm (7-angstrom) node packing nearly 100 billion transistors on a fingernail-sized die, almost double its 2nm density. Its new 3D "nanostack" architecture claims up to 50% more performance or 70% better efficiency.
https://t.co/gpbRf1ak42

#Semiconductors #IBM #ChipDesign @IBM

[StorageReview.com @storagereview]

Dell's new PowerEdge XE8812 brings NVIDIA Vera Rubin NVL4 to market: up to 144 GPUs and over 300kW per rack, fanless direct liquid cooling, and 50% more memory per socket than GB200 NVL4. It's the platform behind NERSC's Doudna supercomputer. GA early 2027.
https://t.co/FJXBWEOP46

#NVIDIA #HPC #Dell @Dell @NVIDIAAI

[StorageReview.com @storagereview]

Qualcomm is going after the data center: the Dragonfly C1000 CPU packs 250+ Oryon cores above 5GHz with PCIe Gen7, and Meta will use it for its next-gen server fleet. It also unveiled the AI300 accelerator and is acquiring AI software firm Modular.
https://t.co/sHBtDydZ60

#Qualcomm #DataCenter #AI @Qualcomm