# rack scale Published articles for rack scale. This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles. ## Smartbird Adds Supermicro DCBBS Racks to Its Managed Single-Tenant AI Clusters, Two Weeks After WEKA DevFeed: [Smartbird Adds Supermicro DCBBS Racks to Its Managed Single-Tenant AI Clusters, Two Weeks After WEKA]() Original publisher: [Read original article]() Author: Harold Fritts Published: 2026-11-08T19:07:39Z Content type: news Language: en Sources: [StorageReview.com]() Topics: [AI Infrastructure](), [Managed Services](), [digital sovereignty](), [tenant data protection]() Tags: [added](), [agentic](), [agentic-ai](), [ai](), [ai-clusters](), [ai-data-centers](), [ai-infrastructure](), [availability](), [data-center](), [deployment](), [design](), [enterprise](), [managed](), [network](), [operations](), [partner](), [procurement](), [rack](), [rack-scale](), [solutions](), [supermicro]() ### AI overview Smartbird and Supermicro will offer dedicated, single-tenant AI clusters built with Supermicro DCBBS and rack-scale systems. Smartbird is responsible for design, procurement, integration, deployment, and ongoing operations; customers can choose on-premises or colocation deployments tailored to their processor, capacity, performance, and workload needs, including training, fine-tuning, inference, agentic AI, and HPC. The arrangement combines Supermicro's compute, storage, networking, power, cooling, and supplier network with Smartbird's managed service, reducing vendors and handoffs. Single tenancy provides isolation and operational control, while deployment location can support data residency requirements; the announcement does not identify compliance certifications or establish that configurations meet specific regulations. It also gives no configurations, GPU platforms, availability dates, customers, or regions. ### Source excerpt Smartbird, the managed AI infrastructure company that launched in June and trades on Nasdaq as BIRD, has added Supermicro to its partner network. The two will deliver dedicated, single-tenant AI clusters built from Supermicro's Data Center Building Block Solutions and rack-scale systems, with Smartbird handling design, procurement, integration, deployment, and day-to-day operations as the single The post Smartbird Adds Supermicro DCBBS Racks to Its Managed Single-Tenant AI Clusters, Two Weeks After WEKA appeared first on StorageReview.com. ## FastAFD: When Does Attention-FFN Disaggregation Pay on Rack-Scale NVLink? DevFeed: [FastAFD: When Does Attention-FFN Disaggregation Pay on Rack-Scale NVLink?]() Original publisher: [Read original article]() Author: Yichao Fu, Yuxuan Zhang, Minshen Zhang, Ruitian Wang, Junda Chen, Hao Zhang Published: 2026-11-06T05:00:00Z Content type: article Language: en Sources: [arXiv: Distributed Computing](), [arXiv: Machine Learning]() Topics: [Mixture of Experts (MoE)](), [vllm](), [GB200](), [NVLink](), [groq]() Tags: [accelerators](), [attention](), [batch](), [cost](), [cs-dc](), [cs-lg](), [decode](), [disaggregation](), [engine](), [expert-parallelism](), [experts](), [gb200](), [gpus](), [groq](), [lpx](), [mixture-of-experts](), [mixture-of-experts-moe](), [moe](), [nvl72](), [nvlink](), [open-source](), [rack-scale]() ### AI overview FastAFD separates attention from MoE experts on GB200 NVL72, overlapping one microbatch's MoE work with another's attention, and compares against vLLM's best wide expert-parallel setting. It reports up to 1.17x throughput on Qwen3, 1.16x on Qwen3.5, and parity on MiniMax-M2.5. Its model predicts AFD throughput within 3% from colocated vLLM profiles. The model attributes gains to larger batches and overlap, weighed against the GPUs allocated to experts; AFD pays off when the gains exceed that cost. Speculative decoding particularly helps light-expert models, adding 8% to AFD's advantage over wide EP. The abstract projects an ideal 2.48x speedup with speculative decoding if experts use on-chip SRAM accelerators, which would remove most disaggregation cost. ### Source excerpt arXiv:2611.07149v1 Announce Type: cross Abstract: Attention-FFN disaggregation (AFD) separates attention from the experts of a mixture-of-experts (MoE) model and has been reported to raise decode throughput over colocated serving. Whether this advantage holds on rack-scale NVLink systems such as GB200 NVL72 remains unclear: prior results were obtained on other hardware, against narrower colocated baselines, or with analytical models, whereas production engines serve large MoE models on these systems with wide expert parallelism (wide EP). We build and tune FastAFD, an AFD engine for GB200 NVL72 that overlaps the MoE of one microbatch with the attention of the other, and compare it against the best vLLM wide EP setting. Experiments show that AFD has a throughput advantage of up to 1.17x over wide EP on Qwen3 and 1.16x on Qwen3.5, and matches it on MiniMax-M2.5. We derive an analytical model that predicts AFD throughput within 3% using only profiles of colocated vLLM, telling whether switching to AFD would pay off before deploying it. The model yields a concise insight: AFD's speedup factors into a batch gain (larger batches, since attention GPUs hold no expert weights), an overlap gain (shorter steps, since the MoE runs on the expert GPUs behind attention), and a disaggregation cost (the GPUs given to experts). AFD has the advantage when the two gains outweigh the cost. This view explains which models benefit from AFD, and shows that speculative decoding helps light-expert models most, where the enlarged overlap gain outweighs the larger disaggregation cost and adds a further 8% to AFD's advantage over wide EP. Looking ahead, serving the experts on accelerators that keep their weights in on-chip SRAM, such as the Groq LPX accelerators paired with Vera Rubin NVL72, would remove most of the disaggregation cost, raising the ideal speedup to 2.48x with speculative decoding. FastAFD is open source at https://github.com/hao-ai-lab/FastAFD. ## Rack-scale AI: How Dell and AMD are building infrastructure DevFeed: [Rack-scale AI: How Dell and AMD are building infrastructure]() Original publisher: [Read original article]() Author: Mark Albertson Published: 2026-11-02T23:33:24Z Content type: article Language: en Sources: [SiliconANGLE]() Topics: [AI Infrastructure](), [dell](), [rocm](), [on-prem](), [Advanced AI workflows]() Tags: [agentic](), [agentic-ai](), [agents](), [ai](), [ai-infrastructure](), [ai-platform](), [amd](), [amd-epyc](), [amd-epyc-processors](), [amd-instinct-mi350p](), [amd-instinct-mi355x](), [amd-rocm-software](), [bandwidth](), [cube-event-coverage](), [dave-vellante](), [dell](), [dell-ai-data-platform-event-from-ambition-to-ai-at-scale-2026](), [dell-ai-platform-with-amd](), [dell-poweredge-r7725](), [dell-poweredge-xe7745](), [dell-technologies](), [dellaidataplatform](), [graphics-processing-units](), [hbm3e-memory](), [infrastructure](), [memory](), [mike-darby](), [network](), [news](), [otel-ai-initiative](), [partner](), [pcie](), [poweredge](), [poweredge-xe9785](), [powerscale](), [powerswitch](), [pradyumna-narendra](), [rack-scale](), [robert-hormuth](), [rocm](), [server-consolidation](), [sovereign-ai](), [telecom](), [thecube](), [university-of-cambridge]() ### AI overview Dell and AMD are assembling enterprise AI infrastructure around server consolidation, modular expansion and a range of GPU deployments. Dell says AMD-powered PowerEdge systems can deliver 70% generational performance gains and consolidate up to 13 prior-generation servers into one. The Dell AI Platform with AMD combines Instinct accelerators, EPYC processors, PowerEdge servers, PowerSwitch networking, PowerScale storage and ROCm; customers can start with smaller configurations and scale on the same validated framework. Dell's support for the PCIe-based MI350P in two PowerEdge models adds 144 GB of HBM3E memory for inference and an air-cooled option for data centers not ready for liquid cooling. The article also cites Cambridge's Zenith AI supercomputer and an AT&T telecom initiative using Dell servers with MI355X GPUs. These are reported capabilities and deployments; the excerpt does not provide independent comparative performance results. ### Source excerpt Dell Technologies Inc. and its partner Advanced Micro Devices Inc. face similar competitive pressures. Both tech powerhouses operate in a business climate where success is now measured across central processing units, graphics processing units, rack-scale systems, memory capacity, network bandwidth and tokens per dollar. The pressure to compete across those measures underscores what theCUBE Research [...] The post Rack-scale AI: How Dell and AMD are building infrastructure appeared first on SiliconANGLE. ## Taiwan's AI Ambitions Run into Talent Gap DevFeed: [Taiwan's AI Ambitions Run into Talent Gap]() Original publisher: [Read original article]() Author: Sally Ward-Foxton Published: 2026-10-16T10:55:42Z Content type: news Language: en Sources: [EE Times]() Topics: [ai-on-mobile](), [AI Infrastructure](), [ASUS](), [tsmc]() Tags: [advanced-packaging](), [ai](), [ai-pc](), [analog](), [analysis](), [business](), [community](), [designline](), [digital](), [education](), [ee-times](), [ee-times-university](), [eelife](), [eelive-electronic-design](), [eet](), [electrical-engineering](), [electronic-components](), [electronic-engineering-times](), [electronics](), [embedded-systems](), [fundamentals-courses](), [hardware](), [learning](), [mixed-signal](), [news](), [opinion](), [part-search](), [physical-ai](), [product-search](), [rack-scale](), [semiconductors](), [software](), [taiwan](), [talent](), [talent-shortage](), [tech-papers](), [technology](), [video]() ### AI overview At an industry conference, AMD executive David Wang identified a shortage of systems and software talent as a challenge to Taiwan's ambitions in AI. He discussed opportunities in advanced packaging, rack-scale AI, physical AI and AI PCs, alongside the need for public-private collaboration and open standards. ### Source excerpt AMD's David Wang told EE Times that while Taiwan has huge opportunities in rack-scale AI, physical AI, and AI PC markets, the sticking point may well be talent. The post Taiwan's AI Ambitions Run into Talent Gap appeared first on EE Times. ## AMD Helion Agentic AI Rack Packs Up to 32,768 EPYC Cores in an Open OCP Blueprint DevFeed: [AMD Helion Agentic AI Rack Packs Up to 32,768 EPYC Cores in an Open OCP Blueprint]() Original publisher: [Read original article]() Author: Harold Fritts Published: 2026-10-12T15:14:28Z Content type: news Language: en Sources: [StorageReview.com]() Topics: [Agentic AI](), [Vera Rubin NVL72]() Tags: [agentic-ai](), [ai](), [ai-infrastructure](), [amd](), [data-center](), [enterprise](), [epyc](), [ocp](), [rack-scale](), [reference-design]() ### AI overview AMD introduced Helion, an open rack reference design for agentic AI, in CPU-only hyperscale and CPU/GPU enterprise configurations. The hyperscale version supports up to 32,768 Zen 6 cores, while the enterprise rack combines EPYC blades with up to 16 Instinct MI350P GPUs. AMD plans to contribute the specification to the Open Compute Project; configurations are under development with partners and are planned for November 2026. ### Source excerpt AMD introduced Helion today, an open rack reference design for agentic AI that builds on its Helios rack-scale platform. Helion comes in two versions: a CPU-only rack for hyperscalers with up to 32,768 Zen 6 cores from 6th Gen EPYC 9006 Series processors, and an all-in-one enterprise rack that mixes EPYC CPU blades with up The post AMD Helion Agentic AI Rack Packs Up to 32,768 EPYC Cores in an Open OCP Blueprint appeared first on StorageReview.com. ## Giga Computing Opens GAIFA, a 700 kW Liquid-Cooled AI Factory Test Bed With Two GB300 NVL72 Racks DevFeed: [Giga Computing Opens GAIFA, a 700 kW Liquid-Cooled AI Factory Test Bed With Two GB300 NVL72 Racks]() Original publisher: [Read original article]() Author: Harold Fritts Published: 2026-09-23T16:34:09Z Content type: news Language: en Sources: [StorageReview.com]() Topics: [AI Factory](), [AI Infrastructure](), [GB300 NVL72](), [Computing](), [Nvidia]() Tags: [ai](), [ai-factory](), [ai-infrastructure](), [amd-epyc](), [capacity](), [computing](), [deployment](), [enterprise](), [gb300-nvl72](), [liquid-cooling](), [nvl72](), [power](), [rack-scale]() ### AI overview Giga Computing has opened GAIFA, a six-rack, direct-liquid-cooled AI computing center in Taiwan. The facility is designed to validate rack-scale infrastructure, including compute, power, thermal management, networking, and system compatibility, with customers before deployment. Its first phase includes two NVIDIA GB300 NVL72 racks and a planned 700 kW power envelope. ### Source excerpt Giga Computing has opened GAIFA (GIGABYTE AI Factory Accelerator), its first proprietary AI computing center, in New Taipei City, Taiwan: a six-rack, direct-liquid-cooled facility built to validate rack-scale AI infrastructure with customers before it ships. The first phase carries a planned 700 kW power envelope, with more capacity planned in later phases as operational needs The post Giga Computing Opens GAIFA, a 700 kW Liquid-Cooled AI Factory Test Bed With Two GB300 NVL72 Racks appeared first on StorageReview.com. ## Dreamforce 2026 Highlights Salesforce's AIforce and Headless 360; Podcast Analyzes Nscale's IPO Filing DevFeed: [Dreamforce 2026 Highlights Salesforce's AIforce and Headless 360; Podcast Analyzes Nscale's IPO Filing]() Original publisher: [Read original article]() Author: Devony Hof Published: 2026-09-21T16:03:58Z Content type: news Language: en Sources: [SiliconANGLE]() Topics: [Artificial Intelligence](), [AI Strategy](), [AI Development](), [Claude](), [Security](), [Slack]() Tags: [2026](), [agentforce](), [agentic](), [agentic-ai](), [agents](), [ai](), [ai-agents](), [aiforce](), [announcement](), [artificial-intelligence](), [coreweave](), [cube-event-coverage](), [dave-vellante](), [dreamforce](), [dreamforce-2026](), [enterprise-ai](), [enterprise-ai-harness](), [headless](), [headless-enterprise](), [john-furrier](), [news](), [nscale](), [nscale-ipo](), [nvidia](), [qualitate](), [rack-scale](), [salesforce](), [system-of-intelligence](), [the-latest](), [thecube](), [thecube-podcast](), [thecube-research](), [thecubepod](), [thecubepodcast](), [video-exclusive]() ### AI overview A theCUBE podcast discusses Salesforce's Dreamforce 2026 announcements, including the AIforce interface layer, Headless 360, and an Enterprise AI Harness for agents. It also analyzes Nscale Global Holdings' initial public offering filing. ### Source excerpt Salesforce Inc. keeps putting all of its eggs in the agent basket -- but the strategy appears to be paying off so far. At Dreamforce 2026, Marc Benioff, chief executive officer of Salesforce, bantered with Nvidia Corp.'s chief executive officer, Jensen Huang, and led the announcement of a new artificial intelligence interface layer, AIforce. The [...] The post On theCUBE Pod: Dreamforce unveils the agentic dream, and Nscale goes public appeared first on SiliconANGLE. ## How Oxide Cuts Data Center Power Consumption in Half DevFeed: [How Oxide Cuts Data Center Power Consumption in Half]() Original publisher: [Read original article]() Author: Kevin Spring Published: 2024-11-08T12:00:00Z Content type: article Language: en Sources: [Oxide Computer Company Blog]() Topics: [Cloud and systems](), [Software-Defined Data Center](), [clean energy data centers](), [control-plane](), [Sustainable Computing]() Tags: [cloud](), [compute](), [control-plane](), [cooling](), [data-center](), [emissions](), [energy](), [hardware](), [power-consumption](), [rack-scale]() ### AI overview Oxide describes a rack-scale cloud computer that treats an entire rack as one system, integrating hardware and software to reduce data center power use. Its design includes a DC busbar, larger and more efficient fans, less restrictive airflow, and a control plane with system-wide observability. The article also describes potential future features such as dynamic power capping and efficiency-based workload placement. ### Source excerpt A rack-scale design that drives data center power efficiency.