StorageReview.com

The New Standard for Academic AI Infrastructure x.com/i/broadcasts/1…

@largePrawn @DellTech @NVIDIAAI Sorry, inline translations from X aren’t working. Seriously though we give away gear on our Discord.

@storagereview @DellTech @NVIDIAAI whoa so cool of you guys to be announcing your 4x rtx 6000 pro giveaway next and only guys named tony are eligible to win

We now have four RTX Pro 6000 GPUs for the XE7740 server. Kevin walks through the bracket install and a few other behind-the-scenes details. #servers #datacenter #gpu @DellTech @NVIDIAAI

More details on the ARTESCA+ Veeam solution here - storagereview.com/review/artesca…

@MichaelDell One step ahead of you 🤠

Think Big dell.com/en-us/shop/del…

@NVIDIA_AI_PC We’re probably juggling 20, and the list isn’t getting any smaller.

Be honest — how many local models do you have downloaded right now? 👀

@GamingWithDaOpa Keeps the lady friend warm we're told. So essentially nowhere, like a good lab should.

@storagereview Where does the Heat go ?

How organized is your homelab? #servers #tech #homelab

Full breakdown: storagereview.com/news/qnap-intr…

QNAP just announced the QAI-h1290FX, an edge AI storage server built specifically for organizations that want to run LLMs and generative AI workloads on their own hardware. The pitch is straightforward: private AI without the cloud. Run RAG pipelines, LLMs, and generative AI applications locally, keeping sensitive data off public cloud platforms entirely. The system targets enterprises where data privacy, latency, and governance requirements make cloud AI a non-starter. This fits a pattern we are seeing across the industry right now. The conversation has shifted from "can we run AI" to "can we run AI without giving up control of our data." Purpose-built edge AI hardware with local storage is becoming its own product category fast. @QNAP_nas #AI #edgeAI #LLM #storage

@planedrop Well obviously yours is a mess. We should have known. 🙃

@storagereview Sorry I meant MINE is a disaster haha not this one.

@bokiko The answer is usually not a yes or not, it's how many?!? 😇

To RTX 6000 or not ?

@planedrop We'd call this one more chaos, thin line between that and disaster though, will concede that.

@storagereview An absolute disaster, as all homelabs should be.

@AMD We’re thrilled to get more access to AMD accelerated platforms. Just pulled in a Dell 9680 with 8-way MI300X. Our power bill is sobbing but the results should be amazing to prove out a ton of inference workloads.

"Competitive inference performance across the workload profiles." The latest MI350X + ROCm 7.2 shows strong performance, broader framework support, and the momentum building around AMD hardware across the AI ecosystem. Full review via @StorageReview: storagereview.com/review/supermi…

HPE expanded ProLiant for the places data centers can't go. The new EL2000 chassis runs Intel Xeon 6 + RTX PRO 6000 Blackwell at -40 to 55C, 95% humidity, heavy vibration and EMI. storagereview.com/news/hpe-expan… @HPE_Compute @IntelBusiness @nvidia #edge #AI #server

@ajaeger74 @DellTech @NVIDIAAI We'll record it, maybe even livestream it in IG or TT.

@storagereview @DellTech @NVIDIAAI I'd love to see Kevin doing the surgery

The Dell XE7740 is designed to handle any accelerators that customers want to use. We're re-deploying ours right now for a new set of inference work with 4x NVIDIA RTX Pro 6000 GPUs. @DellTech @NVIDIAAI

@teromee We're hearing a lot that customers would rather be able to run two distinct workloads on two servers than have all the cards in one. Maybe for some it's more of a scheduling issue.

@storagereview I go with 2 but I am about redunancy rather than a single point of failure. but if we talkin other stuff that breaks because the differences I am going with a single server w/8GPUs.

Help settle a debate. Budget isn't a constraint. What would you rather have? Two GPU servers with 4 RTX Pro 6000s each or one GPU server with 8 RTX Pro 6000 GPUs?

@planedrop No one can hear you cry in the data center. And man, the hypervisors have fallen so far off for enterprise inference.

@storagereview I think I'd say 2, but it really depends on use case and other restraints. Like, are we going to pretend the 8 in a single chassis won't be way louder? What about vGPU slicing with a hypervisor? Then it doesn't really matter unless you need more CPU. etc...