# ServeTheHome

Server and Workstation Reviews

This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles.

## November 2026 Top500 New Systems Are Few and Far Between

DevFeed: [November 2026 Top500 New Systems Are Few and Far Between](<https://devfeed.tech/articles/november-2026-top500-new-systems-are-few-and-far-between-145836.md>)

Original publisher: [Read original article](<https://www.servethehome.com/november-2026-top500-new-systems-analysis/>)

Author: Patrick Kennedy

Published: 2026-11-19T17:48:48Z

Content type: news

Language: en

Sources: [ServeTheHome](<https://devfeed.tech/sources/servethehome.md>)

Topics: [High-Performance Computing](<https://devfeed.tech/topics/high-performance-computing.md>), [systems](<https://devfeed.tech/topics/systems.md>), [rdma](<https://devfeed.tech/topics/rdma.md>)

Tags: [17](<https://devfeed.tech/tags/17.md>), [2021](<https://devfeed.tech/tags/2021.md>), [2022](<https://devfeed.tech/tags/2022.md>), [2023](<https://devfeed.tech/tags/2023.md>), [2026](<https://devfeed.tech/tags/2026.md>), [ai](<https://devfeed.tech/tags/ai.md>), [amd](<https://devfeed.tech/tags/amd.md>), [amd-epyc](<https://devfeed.tech/tags/amd-epyc.md>), [ampere](<https://devfeed.tech/tags/ampere.md>), [arm](<https://devfeed.tech/tags/arm.md>), [blackwell](<https://devfeed.tech/tags/blackwell.md>), [china](<https://devfeed.tech/tags/china.md>), [china-telecom](<https://devfeed.tech/tags/china-telecom.md>), [churn](<https://devfeed.tech/tags/churn.md>), [cloud](<https://devfeed.tech/tags/cloud.md>), [cluster](<https://devfeed.tech/tags/cluster.md>), [clusters](<https://devfeed.tech/tags/clusters.md>), [intel](<https://devfeed.tech/tags/intel.md>), [news](<https://devfeed.tech/tags/news.md>), [november-2026](<https://devfeed.tech/tags/november-2026.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [systems](<https://devfeed.tech/tags/systems.md>), [top500](<https://devfeed.tech/tags/top500.md>), [trends](<https://devfeed.tech/tags/trends.md>)

### AI overview

The November 2026 TOP500 added just 28 systems--the lowest count in the article's tracking series since June 2018--and total list performance grew about 5%. Only five newcomers reached the Top 100; none made the Top 10. The average listed system is now about 3.2 years old. Twenty-three new entries use accelerators, including 14 with NVIDIA Blackwell, while all five CPU-only newcomers use AMD EPYC. China contributed one new system and has 18 systems on the list, down from 173 in November 2021; only two newcomers are classified as Industry. InfiniBand appears in 19 of the 28 new systems. The authors caution that TOP500 submissions reflect declared HPL configurations, not a census of HPC or AI deployments.

### Source excerpt

We check out the new systems in the November 2026 Top500 list and see some of the major trends shaking out The post November 2026 Top500 New Systems Are Few and Far Between appeared first on ServeTheHome.

## Bull Shows Intel Xeon Diamond Rapids Package at SC26

DevFeed: [Bull Shows Intel Xeon Diamond Rapids Package at SC26](<https://devfeed.tech/articles/bull-shows-intel-xeon-diamond-rapids-package-at-sc26-142421.md>)

Original publisher: [Read original article](<https://www.servethehome.com/bull-shows-intel-xeon-diamond-rapids-package-at-sc26/>)

Author: Patrick Kennedy

Published: 2026-11-18T17:21:10Z

Content type: news

Language: en

Sources: [ServeTheHome](<https://devfeed.tech/sources/servethehome.md>)

Topics: [intel](<https://devfeed.tech/topics/intel.md>), [Supercomputing](<https://devfeed.tech/topics/supercomputing.md>), [Xeon 600](<https://devfeed.tech/topics/xeon-600.md>)

Tags: [51-2t](<https://devfeed.tech/tags/51-2t.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-infrastructure](<https://devfeed.tech/tags/ai-infrastructure.md>), [amd](<https://devfeed.tech/tags/amd.md>), [asrock](<https://devfeed.tech/tags/asrock.md>), [asrock-rack](<https://devfeed.tech/tags/asrock-rack.md>), [bull](<https://devfeed.tech/tags/bull.md>), [channel](<https://devfeed.tech/tags/channel.md>), [compute](<https://devfeed.tech/tags/compute.md>), [connectx](<https://devfeed.tech/tags/connectx.md>), [cooling](<https://devfeed.tech/tags/cooling.md>), [copper](<https://devfeed.tech/tags/copper.md>), [cpu](<https://devfeed.tech/tags/cpu.md>), [cpus](<https://devfeed.tech/tags/cpus.md>), [diamond-rapids](<https://devfeed.tech/tags/diamond-rapids.md>), [intel](<https://devfeed.tech/tags/intel.md>), [intel-xeon](<https://devfeed.tech/tags/intel-xeon.md>), [package](<https://devfeed.tech/tags/package.md>), [sc26](<https://devfeed.tech/tags/sc26.md>), [server](<https://devfeed.tech/tags/server.md>), [server-cpus](<https://devfeed.tech/tags/server-cpus.md>), [supercomputing](<https://devfeed.tech/tags/supercomputing.md>)

### AI overview

At SC26, Bull displayed what the excerpt calls the only Diamond Rapids CPU package found at the event, installed in its Sequana XH31600. The blade contains two dual-socket nodes, with 16 memory channels per CPU, and uses liquid cooling. Bull also showed its BXI interconnect: a 64-port 800G OSFP switch and QSFP-DD PCIe Gen5 x16 NICs. An NVIDIA ConnectX-8 NIC was displayed alongside it; the excerpt notes that it supports PCIe Gen6 and contrasts with Bull's Gen5 NIC. The bespoke platform lacks a copper-cabled rear backplane. The author says Diamond Rapids appears to be sampling and estimates a possible Q2 2027 launch, while noting that timing or a split 1P/2P launch could shift the window.

### Source excerpt

Bull shows off an Intel Xeon Diamond Rapids LGA9324 package in its Sequana XH31600 platform at Supercomputing 2026 The post Bull Shows Intel Xeon Diamond Rapids Package at SC26 appeared first on ServeTheHome.

## NVIDIA Announces Rubin HPC: Science-Focused Server GPU to Arrive in 2028

DevFeed: [NVIDIA Announces Rubin HPC: Science-Focused Server GPU to Arrive in 2028](<https://devfeed.tech/articles/nvidia-announces-rubin-hpc-science-focused-server-gpu-to-arrive-in-2028-140184.md>)

Original publisher: [Read original article](<https://www.servethehome.com/nvidia-announces-rubin-hpc-science-focused-server-gpu-to-arrive-in-2028/>)

Author: Ryan Smith

Published: 2026-11-17T15:45:19Z

Content type: news

Language: en

Sources: [ServeTheHome](<https://devfeed.tech/sources/servethehome.md>)

Topics: [High-Performance Computing](<https://devfeed.tech/topics/high-performance-computing.md>), [NVLink](<https://devfeed.tech/topics/nvlink.md>), [CUDA](<https://devfeed.tech/topics/cuda.md>), [Tensor Cores](<https://devfeed.tech/topics/tensor-cores.md>)

Tags: [accelerators](<https://devfeed.tech/tags/accelerators.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [hpc](<https://devfeed.tech/tags/hpc.md>), [news](<https://devfeed.tech/tags/news.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [rubin](<https://devfeed.tech/tags/rubin.md>), [rubin-hpc](<https://devfeed.tech/tags/rubin-hpc.md>), [sc26](<https://devfeed.tech/tags/sc26.md>), [science](<https://devfeed.tech/tags/science.md>), [server](<https://devfeed.tech/tags/server.md>)

### AI overview

NVIDIA says its Rubin HPC GPU, aimed at science and engineering workloads, is planned for release in 2028. The announced chip will have more than 50,000 CUDA cores--about twice as many as the current Rubin server GPU--reflecting an emphasis on vector processing that many traditional HPC workloads need. NVIDIA also cites a balanced approach to compute and data movement, NVLink 7, and "sparse data acceleration," but has not disclosed performance figures, memory details, or how the sparse-data feature works. Support across precisions from NVFP4 to FP64 is promised; actual FP64 performance will depend on the architecture, including how much hardware is allocated to FP64 and tensor operations. The article presents the announcement as NVIDIA's return to dedicated HPC after Blackwell, with AMD's MI430X expected in 2027 and Intel preparing a PCIe GPU D-Series.

### Source excerpt

Marking their return to the HPC market, at SC26 NVIDIA has announced the Rubin HPC chip. The science and engineering-focused GPU will have over 50,000 CUDA cores and will be released in 2028 The post NVIDIA Announces Rubin HPC: Science-Focused Server GPU to Arrive in 2028 appeared first on ServeTheHome.

## NVIDIA and Lambda Disclose First Production Co-Packaged Optics Cluster... It is FINALLY real

DevFeed: [NVIDIA and Lambda Disclose First Production Co-Packaged Optics Cluster... It is FINALLY real](<https://devfeed.tech/articles/nvidia-and-lambda-disclose-first-production-co-packaged-optics-cluster-it-is-finally-real-136839.md>)

Original publisher: [Read original article](<https://www.servethehome.com/nvidia-and-lambda-disclose-first-production-co-packaged-optics-cluster-it-is-finally-real/>)

Author: Patrick Kennedy

Published: 2026-11-16T21:24:47Z

Content type: news

Language: en

Sources: [ServeTheHome](<https://devfeed.tech/sources/servethehome.md>)

Topics: [Nvidia](<https://devfeed.tech/topics/nvidia.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [adoption](<https://devfeed.tech/tags/adoption.md>), [chip](<https://devfeed.tech/tags/chip.md>), [cluster](<https://devfeed.tech/tags/cluster.md>), [co-packaged-optics](<https://devfeed.tech/tags/co-packaged-optics.md>), [cpo](<https://devfeed.tech/tags/cpo.md>), [deployment](<https://devfeed.tech/tags/deployment.md>), [first](<https://devfeed.tech/tags/first.md>), [gpus](<https://devfeed.tech/tags/gpus.md>), [industry](<https://devfeed.tech/tags/industry.md>), [innovation](<https://devfeed.tech/tags/innovation.md>), [lab](<https://devfeed.tech/tags/lab.md>), [lambda](<https://devfeed.tech/tags/lambda.md>), [light](<https://devfeed.tech/tags/light.md>), [links](<https://devfeed.tech/tags/links.md>), [modules](<https://devfeed.tech/tags/modules.md>), [networking](<https://devfeed.tech/tags/networking.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [optics](<https://devfeed.tech/tags/optics.md>), [production](<https://devfeed.tech/tags/production.md>), [production-cluster](<https://devfeed.tech/tags/production-cluster.md>), [sc26](<https://devfeed.tech/tags/sc26.md>)

### AI overview

At SC26, NVIDIA and Lambda disclosed a production cluster with more than 10,000 GPUs using co-packaged optics (CPO), showing that the technology has moved beyond the lab. NVIDIA's first CPO implementation is in its Quantum-X switch: 18 serviceable front-mounted light sources feed optical connections at the switch package, which has 144 MPO/MTP ports rated at 800 Gbps each. The article says serviceable light sources address a past reliability and maintenance obstacle; having fewer sources than individual pluggable optical modules may also reduce failure points and ease cooling. Moving optics closer to the switch package avoids driving high-speed copper signals over longer distances to front-panel cages, which the article says can lower power use and leave more power for GPUs in a constrained facility. The excerpt establishes that this deployment is running, but offers no measured power savings or adoption forecast.

### Source excerpt

Lambda is the first to disclose it has co-packaged optics deployed in a production cluster with over 10,000 GPUs The post NVIDIA and Lambda Disclose First Production Co-Packaged Optics Cluster... It is FINALLY real appeared first on ServeTheHome.

## TE Connectivity at OCP Summit 2026 The Interview

DevFeed: [TE Connectivity at OCP Summit 2026 The Interview](<https://devfeed.tech/articles/te-connectivity-at-ocp-summit-2026-the-interview-134872.md>)

Original publisher: [Read original article](<https://www.servethehome.com/te-connectivity-at-ocp-summit-2026-the-interview/>)

Author: Patrick Kennedy

Published: 2026-11-16T07:00:21Z

Content type: news

Language: en

Sources: [ServeTheHome](<https://devfeed.tech/sources/servethehome.md>)

Topics: [communications](<https://devfeed.tech/topics/communications.md>), [Network architectures](<https://devfeed.tech/topics/network-architectures.md>), [CVE-2026-86060](<https://devfeed.tech/topics/cve-2026-86060.md>)

Tags: [2017](<https://devfeed.tech/tags/2017.md>), [2026](<https://devfeed.tech/tags/2026.md>), [3d](<https://devfeed.tech/tags/3d.md>), [3d-printed](<https://devfeed.tech/tags/3d-printed.md>), [448g](<https://devfeed.tech/tags/448g.md>), [800v](<https://devfeed.tech/tags/800v.md>), [about](<https://devfeed.tech/tags/about.md>), [ai](<https://devfeed.tech/tags/ai.md>), [connectivity](<https://devfeed.tech/tags/connectivity.md>), [copper](<https://devfeed.tech/tags/copper.md>), [first](<https://devfeed.tech/tags/first.md>), [interview](<https://devfeed.tech/tags/interview.md>), [more](<https://devfeed.tech/tags/more.md>), [news](<https://devfeed.tech/tags/news.md>), [ocp](<https://devfeed.tech/tags/ocp.md>), [ocp-summit](<https://devfeed.tech/tags/ocp-summit.md>), [optical](<https://devfeed.tech/tags/optical.md>), [power](<https://devfeed.tech/tags/power.md>), [serdes](<https://devfeed.tech/tags/serdes.md>), [summit](<https://devfeed.tech/tags/summit.md>), [te-connectivity](<https://devfeed.tech/tags/te-connectivity.md>)

### AI overview

At OCP Summit 2026, TE Connectivity described a system-level approach to AI infrastructure that combines high-speed data links, power delivery, cooling, serviceability and manufacturing. Payne outlined a newly announced 448G-per-lane front-panel concept developed with NVIDIA and Cisco, designed to balance signal integrity, power and form factor, with provision for liquid cooling as transceiver power rises. TE is advancing copper backplanes and midplanes while investing in co-packaged optics; Payne said copper remains important as routing limits make an eventual shift to optics more pressing. For power, the booth showed a sidecar power rack, liquid-cooled horizontal busbar and prototype rack-to-rack 800V cables, with 800V pathfinding aimed at future megawatt-class racks. Payne tied these developments to customer feedback, iterative design and production at scale, citing 10 DDN manufacturing facilities worldwide.

### Source excerpt

Brian Payne joins us for an interview on 800V DC power, 488G SerDes, copper and optical connectivity and more from OCP Summit 2026 The post TE Connectivity at OCP Summit 2026 The Interview appeared first on ServeTheHome.

## AMD Launches EPYC Embedded 9006: Venice Gets an Early Start in Embedded

DevFeed: [AMD Launches EPYC Embedded 9006: Venice Gets an Early Start in Embedded](<https://devfeed.tech/articles/amd-launches-epyc-embedded-9006-venice-gets-an-early-start-in-embedded-134119.md>)

Original publisher: [Read original article](<https://www.servethehome.com/amd-launches-epyc-embedded-9006-venice-gets-an-early-start-in-embedded/>)

Author: Ryan Smith

Published: 2026-11-14T19:00:44Z

Content type: news

Language: en

Sources: [ServeTheHome](<https://devfeed.tech/sources/servethehome.md>)

Topics: [Embedded Systems](<https://devfeed.tech/topics/embedded-systems.md>), [pcie](<https://devfeed.tech/topics/pcie.md>), [ddr5](<https://devfeed.tech/topics/ddr5.md>)

Tags: [about](<https://devfeed.tech/tags/about.md>), [acceleration](<https://devfeed.tech/tags/acceleration.md>), [adopting](<https://devfeed.tech/tags/adopting.md>), [agentic](<https://devfeed.tech/tags/agentic.md>), [agentic-ai](<https://devfeed.tech/tags/agentic-ai.md>), [agentic-ai-systems](<https://devfeed.tech/tags/agentic-ai-systems.md>), [ai](<https://devfeed.tech/tags/ai.md>), [amd](<https://devfeed.tech/tags/amd.md>), [benefits](<https://devfeed.tech/tags/benefits.md>), [embedded](<https://devfeed.tech/tags/embedded.md>), [epyc](<https://devfeed.tech/tags/epyc.md>), [epyc-9006](<https://devfeed.tech/tags/epyc-9006.md>), [first](<https://devfeed.tech/tags/first.md>), [generation](<https://devfeed.tech/tags/generation.md>), [launches](<https://devfeed.tech/tags/launches.md>), [news](<https://devfeed.tech/tags/news.md>), [next](<https://devfeed.tech/tags/next.md>), [performance](<https://devfeed.tech/tags/performance.md>), [platform](<https://devfeed.tech/tags/platform.md>), [server](<https://devfeed.tech/tags/server.md>), [server-cpus](<https://devfeed.tech/tags/server-cpus.md>), [sp7](<https://devfeed.tech/tags/sp7.md>), [venice](<https://devfeed.tech/tags/venice.md>), [with](<https://devfeed.tech/tags/with.md>), [zen-6](<https://devfeed.tech/tags/zen-6.md>)

### AI overview

AMD launched EPYC Embedded 9006 alongside its commercial Venice server CPUs, bringing the SP7 platform to embedded customers. The lineup supports up to 256 cores, 16-channel DDR5-8000/MRDRIMM-12800 memory, and 96 PCIe Gen6/CXL 3.1 lanes. AMD offers at least seven years of availability and says SP7 designs can be upgraded to the next Florence generation without a full platform redesign or revalidation. AMD targets automatic test equipment, high-end firewalls, and enterprise storage controllers. One tradeoff for some embedded designs: SP7 has 96 PCIe lanes, fewer than SP5's 128, though the new lanes are faster. The 128-core EPYC Embedded 9756 was available at launch; five other chips were expected in late December.

### Source excerpt

With its new Zen 6 "Venice" silicon in hand, AMD has launched its next-generation EPYC Embedded 9006 platform for the embedded market, aiming to bring the performance benefits of the new platform to the long term stability-focused market segment The post AMD Launches EPYC Embedded 9006: Venice Gets an Early Start in Embedded appeared first on ServeTheHome.

## NVIDIA CEO and Co-Founder Jensen Huang Resigns, Effective Immediately

DevFeed: [NVIDIA CEO and Co-Founder Jensen Huang Resigns, Effective Immediately](<https://devfeed.tech/articles/nvidia-ceo-and-co-founder-jensen-huang-resigns-effective-immediately-130993.md>)

Original publisher: [Read original article](<https://www.servethehome.com/nvidia-ceo-and-co-founder-jensen-huang-resigns-effective-immediately/>)

Author: Ryan Smith

Published: 2026-11-12T15:08:54Z

Content type: news

Language: en

Sources: [ServeTheHome](<https://devfeed.tech/sources/servethehome.md>)

Topics: [Nvidia](<https://devfeed.tech/topics/nvidia.md>)

Tags: [affair](<https://devfeed.tech/tags/affair.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-industry](<https://devfeed.tech/tags/ai-industry.md>), [automotive](<https://devfeed.tech/tags/automotive.md>), [board](<https://devfeed.tech/tags/board.md>), [board-of-directors](<https://devfeed.tech/tags/board-of-directors.md>), [ceo](<https://devfeed.tech/tags/ceo.md>), [co-founder](<https://devfeed.tech/tags/co-founder.md>), [colette-kress](<https://devfeed.tech/tags/colette-kress.md>), [company](<https://devfeed.tech/tags/company.md>), [cuda](<https://devfeed.tech/tags/cuda.md>), [earnings](<https://devfeed.tech/tags/earnings.md>), [effective](<https://devfeed.tech/tags/effective.md>), [email](<https://devfeed.tech/tags/email.md>), [employee](<https://devfeed.tech/tags/employee.md>), [graphics](<https://devfeed.tech/tags/graphics.md>), [industries](<https://devfeed.tech/tags/industries.md>), [jensen-huang](<https://devfeed.tech/tags/jensen-huang.md>), [news](<https://devfeed.tech/tags/news.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [nvidia-ceo](<https://devfeed.tech/tags/nvidia-ceo.md>)

### AI overview

NVIDIA announced that co-founder and CEO Jensen Huang resigned effective immediately. CFO Colette Kress will lead as interim CEO while the board searches for a permanent successor, considering candidates inside and outside the company; Harvey C. Jones was appointed board chair. The company said operations, customer commitments, and the product roadmap should continue without interruption. The text says the board has not given an official reason for Huang's departure and describes recent allegations and an independent investigation as context, not as an established explanation. It emphasizes the challenge of replacing a long-serving, highly centralized leader and leaves the effects on NVIDIA's strategy and the AI and graphics industries unresolved.

### Source excerpt

NVIDIA's CEO and Co-Founder Jensen Huang has resigned from the company, effective immediately, marking a pivotal moment for the company and the AI and graphics industries at large The post NVIDIA CEO and Co-Founder Jensen Huang Resigns, Effective Immediately appeared first on ServeTheHome.

## This is the Meta MetaRoCE Network Adapter

DevFeed: [This is the Meta MetaRoCE Network Adapter](<https://devfeed.tech/articles/this-is-the-meta-metaroce-network-adapter-130994.md>)

Original publisher: [Read original article](<https://www.servethehome.com/this-is-the-meta-metaroce-network-adapter/>)

Author: Rohit Kumar

Published: 2026-11-12T04:51:39Z

Content type: news

Language: en

Sources: [ServeTheHome](<https://devfeed.tech/sources/servethehome.md>)

Topics: [Protocol (disambiguation)](<https://devfeed.tech/topics/protocol.md>), [Network architectures](<https://devfeed.tech/topics/network-architectures.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [adapter](<https://devfeed.tech/tags/adapter.md>), [clusters](<https://devfeed.tech/tags/clusters.md>), [congestion-control](<https://devfeed.tech/tags/congestion-control.md>), [convergence](<https://devfeed.tech/tags/convergence.md>), [ethernet](<https://devfeed.tech/tags/ethernet.md>), [fast](<https://devfeed.tech/tags/fast.md>), [features](<https://devfeed.tech/tags/features.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [hints](<https://devfeed.tech/tags/hints.md>), [implement](<https://devfeed.tech/tags/implement.md>), [layer](<https://devfeed.tech/tags/layer.md>), [meta](<https://devfeed.tech/tags/meta.md>), [metaroce](<https://devfeed.tech/tags/metaroce.md>), [network](<https://devfeed.tech/tags/network.md>), [networking](<https://devfeed.tech/tags/networking.md>), [ocp](<https://devfeed.tech/tags/ocp.md>), [ocp-summit](<https://devfeed.tech/tags/ocp-summit.md>), [scale](<https://devfeed.tech/tags/scale.md>), [training](<https://devfeed.tech/tags/training.md>)

### AI overview

MetaRoCE is Meta's open RDMA transport proposal for scaling training networks to million-GPU clusters. Its listed features include per-path reliability and congestion control, receiver rate hints for faster incast convergence, tolerance of packet loss and reordering, and multiple streams across network paths on a single queue pair. It is designed for commodity Ethernet using ECN and ECMP without switch changes. Meta previously demonstrated the protocol on programmable AMD Pensando NICs, which the excerpt says could let Meta source hardware from multiple vendors; an OCP NIC 3.0 adapter was displayed at OCP Summit 2026.

### Source excerpt

At OCP Summit 2026, Meta showed off its MetaRoCE adapter designed to scale to million-GPU training clusters The post This is the Meta MetaRoCE Network Adapter appeared first on ServeTheHome.

## AMD EPYC 9006 Results Are Here and AMD Delivers a Knockout Part Deux

DevFeed: [AMD EPYC 9006 Results Are Here and AMD Delivers a Knockout Part Deux](<https://devfeed.tech/articles/amd-epyc-9006-results-are-here-and-amd-delivers-a-knockout-part-deux-127156.md>)

Original publisher: [Read original article](<https://www.servethehome.com/amd-epyc-9006-results-are-here-and-amd-delivers-a-knockout-part-deux/>)

Author: Patrick Kennedy

Published: 2026-11-10T17:00:08Z

Content type: news

Language: en

Sources: [ServeTheHome](<https://devfeed.tech/sources/servethehome.md>)

Topics: [Electrical engineering and computer science (EECS)](<https://devfeed.tech/topics/electrical-engineering-and-computer-science-eecs.md>)

Tags: [agentic](<https://devfeed.tech/tags/agentic.md>), [agentic-workloads](<https://devfeed.tech/tags/agentic-workloads.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-systems](<https://devfeed.tech/tags/ai-systems.md>), [amd](<https://devfeed.tech/tags/amd.md>), [amd-epyc](<https://devfeed.tech/tags/amd-epyc.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [bandwidth](<https://devfeed.tech/tags/bandwidth.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [chiplets](<https://devfeed.tech/tags/chiplets.md>), [cpu](<https://devfeed.tech/tags/cpu.md>), [data-center](<https://devfeed.tech/tags/data-center.md>), [epyc](<https://devfeed.tech/tags/epyc.md>), [epyc-9006](<https://devfeed.tech/tags/epyc-9006.md>), [performance](<https://devfeed.tech/tags/performance.md>), [server](<https://devfeed.tech/tags/server.md>), [server-cpus](<https://devfeed.tech/tags/server-cpus.md>), [venice](<https://devfeed.tech/tags/venice.md>), [zen-6](<https://devfeed.tech/tags/zen-6.md>)

### AI overview

AMD's EPYC 9006 "Venice" introduces Zen 6 for servers, with up to 256 cores per socket, PCIe Gen6, and more than twice the memory bandwidth of the prior generation, alongside claimed 20% higher IPC and higher clocks. The article's launch review benchmarks the flagship EPYC 9996 using AgentSTH V7 and preview material for V8, including agentic workloads; the supplied excerpt does not provide benchmark results. AMD also describes multiple product families, including Venice-X high-cache parts and LPDDR5X Verano. The EPYC 9686F's shipping peak frequency rose from the previously announced 5.0 GHz to 5.3 GHz after final silicon exceeded expectations; the 64-core 9586F received no such bump, so the 9686F is the lineup's highest-clocked option.

### Source excerpt

The day is finally here for the launch of AMD's EPYC 9006 "Venice" platform and the Zen 6 CPU architecture. AMD has been hard at work over the past two years to build a better server CPU, and thanks to its performance, Venice is primed to deliver another knockout blow for AMD The post AMD EPYC 9006 Results Are Here and AMD Delivers a Knockout Part Deux appeared first on ServeTheHome.

## Google TPU 8tx Pairs AMD EPYC Turin Hosts with Confidential Computing while TPU 8i Max Doubles Memory Bandwidth

DevFeed: [Google TPU 8tx Pairs AMD EPYC Turin Hosts with Confidential Computing while TPU 8i Max Doubles Memory Bandwidth](<https://devfeed.tech/articles/google-tpu-8tx-pairs-amd-epyc-turin-hosts-with-confidential-computing-while-tpu-8i-max-doubles-memory-bandwidth-127157.md>)

Original publisher: [Read original article](<https://www.servethehome.com/google-tpu-8tx-8i-max-amd-epyc-turin/>)

Author: Vic A

Published: 2026-11-09T21:37:58Z

Content type: article

Language: en

Sources: [ServeTheHome](<https://devfeed.tech/sources/servethehome.md>)

Topics: [Confidential Computing](<https://devfeed.tech/topics/confidential-computing.md>), [AI Factory](<https://devfeed.tech/topics/ai-factory.md>), [Security & Privacy](<https://devfeed.tech/topics/security-privacy.md>)

Tags: [accelerators](<https://devfeed.tech/tags/accelerators.md>), [agentic](<https://devfeed.tech/tags/agentic.md>), [agentic-workloads](<https://devfeed.tech/tags/agentic-workloads.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [amd](<https://devfeed.tech/tags/amd.md>), [amd-epyc](<https://devfeed.tech/tags/amd-epyc.md>), [bandwidth](<https://devfeed.tech/tags/bandwidth.md>), [confidential-computing](<https://devfeed.tech/tags/confidential-computing.md>), [epyc](<https://devfeed.tech/tags/epyc.md>), [google](<https://devfeed.tech/tags/google.md>), [google-tpu](<https://devfeed.tech/tags/google-tpu.md>), [google-tpu-8tx](<https://devfeed.tech/tags/google-tpu-8tx.md>), [inference](<https://devfeed.tech/tags/inference.md>), [memory](<https://devfeed.tech/tags/memory.md>), [memory-bandwidth](<https://devfeed.tech/tags/memory-bandwidth.md>), [server](<https://devfeed.tech/tags/server.md>), [tpu](<https://devfeed.tech/tags/tpu.md>), [tpu-8i](<https://devfeed.tech/tags/tpu-8i.md>), [tpu-8i-max](<https://devfeed.tech/tags/tpu-8i-max.md>), [tpu-8tx](<https://devfeed.tech/tags/tpu-8tx.md>), [turin](<https://devfeed.tech/tags/turin.md>)

### AI overview

Google's TPU 8tx and TPU 8i Max target different needs using existing TPU silicon. The 8tx pairs TPU 8t silicon with 5th Gen AMD EPYC Turin hosts and confidential computing, which protects data while it is being processed. Its dual-socket hosts provide 128 PCIe Gen5 lanes for four TPUs and high-speed NICs; Google says its Boardfly network can deliver up to 2x lower latency than TPU 8t with near-zero performance penalty. For inference and serving where data must stay within the confidential computing domain, Google recommends 8tx. The 8i Max combines two TPU 8i chips in one package, doubling compute, SRAM and HBM bandwidth versus one 8i; it provides 384 GB HBM, 17.2 TB/s bandwidth and 768 MB SRAM. Both use Boardfly, while the article distinguishes 8i Max as an inference-focused option for memory-intensive workloads.

### Source excerpt

The new Google TPU 8tx pairs AMD EPYC Turin hosts with confidential computing while TPU 8i Max doubles memory bandwidth for inference The post Google TPU 8tx Pairs AMD EPYC Turin Hosts with Confidential Computing while TPU 8i Max Doubles Memory Bandwidth appeared first on ServeTheHome.

## Dell PowerEdge R770AP Review One Fast Intel Xeon 6 Server

DevFeed: [Dell PowerEdge R770AP Review One Fast Intel Xeon 6 Server](<https://devfeed.tech/articles/dell-poweredge-r770ap-review-one-fast-intel-xeon-6-server-123939.md>)

Original publisher: [Read original article](<https://www.servethehome.com/dell-poweredge-r770ap-review-one-fast-intel-xeon-6-server/>)

Author: Patrick Kennedy

Published: 2026-11-08T23:00:11Z

Content type: article

Language: en

Sources: [ServeTheHome](<https://devfeed.tech/sources/servethehome.md>)

Topics: [systems](<https://devfeed.tech/topics/systems.md>), [dell](<https://devfeed.tech/topics/dell.md>), [intel](<https://devfeed.tech/topics/intel.md>), [ddr5](<https://devfeed.tech/topics/ddr5.md>), [Benchmark](<https://devfeed.tech/topics/benchmark.md>), [M.2 NVMe](<https://devfeed.tech/topics/m-2-nvme.md>)

Tags: [air](<https://devfeed.tech/tags/air.md>), [airflow](<https://devfeed.tech/tags/airflow.md>), [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [boot](<https://devfeed.tech/tags/boot.md>), [broadcom](<https://devfeed.tech/tags/broadcom.md>), [connectx](<https://devfeed.tech/tags/connectx.md>), [data-center](<https://devfeed.tech/tags/data-center.md>), [data-centers](<https://devfeed.tech/tags/data-centers.md>), [ddr5](<https://devfeed.tech/tags/ddr5.md>), [dell](<https://devfeed.tech/tags/dell.md>), [dell-poweredge](<https://devfeed.tech/tags/dell-poweredge.md>), [granite-rapids](<https://devfeed.tech/tags/granite-rapids.md>), [intel](<https://devfeed.tech/tags/intel.md>), [intel-xeon](<https://devfeed.tech/tags/intel-xeon.md>), [intel-xeon-6](<https://devfeed.tech/tags/intel-xeon-6.md>), [memory](<https://devfeed.tech/tags/memory.md>), [performance](<https://devfeed.tech/tags/performance.md>), [poweredge](<https://devfeed.tech/tags/poweredge.md>), [review](<https://devfeed.tech/tags/review.md>), [server](<https://devfeed.tech/tags/server.md>), [server-systems](<https://devfeed.tech/tags/server-systems.md>), [socket](<https://devfeed.tech/tags/socket.md>), [xeon-6](<https://devfeed.tech/tags/xeon-6.md>)

### AI overview

The Dell PowerEdge R770AP is a 2U air-cooled server built for high core density, supporting up to two 128-core Intel Xeon 6980P processors. Its front can be configured for up to 32 E3.S Gen5 NVMe drives, compared with up to 16 U.2 drives, or ordered without a backplane; the reviewed unit has four 1.6TB E3.S SSDs. Rear options include OCP NIC 3.0 networking, with the reviewed configuration using Broadcom BCM57608 and ConnectX-6 Dx controllers and 200Gbps ports; the platform can accommodate 400Gbps networking. The excerpt also describes redundant 1.8kW power supplies, dedicated iDRAC management, and a rear BOSS-N1 boot module with two M.2 drives for RAID 1. It presents one configuration and notes Dell offers multiple storage, power, and riser options; the supplied text covers the external hardware overview and ends before the internal system review.

### Source excerpt

Today we are taking a look at Dell's PowerEdge R770AP. The 2U server goes all-in on performance with a dual-socket Xeon 6 setup and 24 channels of DDR5 memory The post Dell PowerEdge R770AP Review One Fast Intel Xeon 6 Server appeared first on ServeTheHome.

## ASRock Rack DMRD16-2L+ Spotted with 16 DDR5 Slots for Diamond Rapids

DevFeed: [ASRock Rack DMRD16-2L+ Spotted with 16 DDR5 Slots for Diamond Rapids](<https://devfeed.tech/articles/asrock-rack-dmrd16-2l-spotted-with-16-ddr5-slots-for-diamond-rapids-122190.md>)

Original publisher: [Read original article](<https://www.servethehome.com/asrock-rack-dmrd16-2l-plus-diamond-rapids-ocp-summit-2026/>)

Author: John Lee

Published: 2026-11-07T16:00:55Z

Content type: news

Language: en

Sources: [ServeTheHome](<https://devfeed.tech/sources/servethehome.md>)

Topics: [ddr5](<https://devfeed.tech/topics/ddr5.md>)

Tags: [asrock-rack](<https://devfeed.tech/tags/asrock-rack.md>), [cxl-3-0](<https://devfeed.tech/tags/cxl-3-0.md>), [ddr5](<https://devfeed.tech/tags/ddr5.md>), [diamond-rapids](<https://devfeed.tech/tags/diamond-rapids.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [intel](<https://devfeed.tech/tags/intel.md>), [m-2](<https://devfeed.tech/tags/m-2.md>), [memory](<https://devfeed.tech/tags/memory.md>), [motherboard](<https://devfeed.tech/tags/motherboard.md>), [ocp-summit](<https://devfeed.tech/tags/ocp-summit.md>), [pcie](<https://devfeed.tech/tags/pcie.md>), [pcie-5-0](<https://devfeed.tech/tags/pcie-5-0.md>), [pcie-6-0](<https://devfeed.tech/tags/pcie-6-0.md>), [server](<https://devfeed.tech/tags/server.md>), [server-motherboards](<https://devfeed.tech/tags/server-motherboards.md>), [xeon](<https://devfeed.tech/tags/xeon.md>)

### AI overview

Shown at OCP Summit 2026, ASRock Rack's single-socket DMRD16-2L+ is an EEB board for Intel Diamond Rapids Xeons, with sixteen memory channels and one DIMM slot per channel. It supports RDIMMs up to 128GB at 8000MT/s, MRDIMMs up to 128GB at 12800MT/s, and MRDIMM 2U modules up to 256GB at 12800MT/s. Instead of PCIe CEM slots or an OCP NIC 3.0 connector, expansion uses up to sixteen MCIO x8 connectors rated for PCIe 6.0 or CXL 3.0, providing 128 lanes in total. Local storage is limited to one PCIe 5.0 x4 M.2 slot. Management uses an ASPEED AST2700 BMC; networking includes a dedicated IPMI port and two 1GbE ports. The excerpt presents the MCIO-only I/O design as a move toward cabling amid PCIe signaling challenges.

### Source excerpt

We check out the ASRock Rack DMRD16-2L+ which is a 16-channel Intel Diamond Rapids Xeon motherboard with 128x PCIe Gen6 lanes over MCIO The post ASRock Rack DMRD16-2L+ Spotted with 16 DDR5 Slots for Diamond Rapids appeared first on ServeTheHome.

## Cisco Catalyst C1300-24XTS Review A Neat 24-port 10GbE Switch

DevFeed: [Cisco Catalyst C1300-24XTS Review A Neat 24-port 10GbE Switch](<https://devfeed.tech/articles/cisco-catalyst-c1300-24xts-review-a-neat-24-port-10gbe-switch-122191.md>)

Original publisher: [Read original article](<https://www.servethehome.com/cisco-catalyst-c1300-24xts-review-a-neat-24-port-10gbe-switch-marvell/>)

Author: Rohit Kumar

Published: 2026-11-06T20:30:45Z

Content type: article

Language: en

Sources: [ServeTheHome](<https://devfeed.tech/sources/servethehome.md>)

Topics: [10 Gigabit Ethernet](<https://devfeed.tech/topics/10gbe-networking.md>), [Cisco](<https://devfeed.tech/topics/cisco.md>)

Tags: [10gbase-t](<https://devfeed.tech/tags/10gbase-t.md>), [10gbe](<https://devfeed.tech/tags/10gbe.md>), [ac-power](<https://devfeed.tech/tags/ac-power.md>), [catalyst](<https://devfeed.tech/tags/catalyst.md>), [cisco](<https://devfeed.tech/tags/cisco.md>), [console](<https://devfeed.tech/tags/console.md>), [cooling](<https://devfeed.tech/tags/cooling.md>), [fiber](<https://devfeed.tech/tags/fiber.md>), [hardware](<https://devfeed.tech/tags/hardware.md>), [independent](<https://devfeed.tech/tags/independent.md>), [interface](<https://devfeed.tech/tags/interface.md>), [management](<https://devfeed.tech/tags/management.md>), [metal](<https://devfeed.tech/tags/metal.md>), [networking](<https://devfeed.tech/tags/networking.md>), [out-of-band](<https://devfeed.tech/tags/out-of-band.md>), [overview](<https://devfeed.tech/tags/overview.md>), [panel](<https://devfeed.tech/tags/panel.md>), [perforated](<https://devfeed.tech/tags/perforated.md>), [port](<https://devfeed.tech/tags/port.md>), [review](<https://devfeed.tech/tags/review.md>), [switch](<https://devfeed.tech/tags/switch.md>)

### AI overview

The C1300-24XTS provides 24 independent 10GbE data ports: 12 10GBase-T copper ports and 12 SFP+ cages, rather than combo ports. This split supports environments with both copper-connected systems and fiber-connected storage or uplinks; the spaced SFP+ layout also makes optics and DACs easier to fit and aids cooling. The 1U metal chassis has a dedicated 1GbE RJ45 out-of-band management port, separate from the data ports and console connections. It has a single AC power input, so redundant A+B power requires an external automatic transfer switching PDU. The excerpt describes the hardware but does not provide test results or a full performance assessment.

### Source excerpt

In our Cisco Catalyst C1300-24XTS review, we take the switch apart to see how it works and test it under full load to see if it holds up The post Cisco Catalyst C1300-24XTS Review A Neat 24-port 10GbE Switch appeared first on ServeTheHome.

## Supermicro Super AI Station Review: The AI Server That Doubles as a Workstation

DevFeed: [Supermicro Super AI Station Review: The AI Server That Doubles as a Workstation](<https://devfeed.tech/articles/supermicro-super-ai-station-review-the-ai-server-that-doubles-as-a-workstation-119750.md>)

Original publisher: [Read original article](<https://www.servethehome.com/supermicro-super-ai-station-review-the-ai-server-that-doubles-as-a-workstation/>)

Author: Ryan Smith

Published: 2026-11-05T21:15:28Z

Content type: article

Language: en

Sources: [ServeTheHome](<https://devfeed.tech/sources/servethehome.md>)

Topics: [DGX Station](<https://devfeed.tech/topics/dgx-station.md>), [RTX PRO](<https://devfeed.tech/topics/rtx-pro.md>)

Tags: [10gbase-t](<https://devfeed.tech/tags/10gbase-t.md>), [10gbe](<https://devfeed.tech/tags/10gbe.md>), [ai](<https://devfeed.tech/tags/ai.md>), [blackwell](<https://devfeed.tech/tags/blackwell.md>), [blackwell-ultra](<https://devfeed.tech/tags/blackwell-ultra.md>), [compute](<https://devfeed.tech/tags/compute.md>), [connectx](<https://devfeed.tech/tags/connectx.md>), [convertible](<https://devfeed.tech/tags/convertible.md>), [cooling](<https://devfeed.tech/tags/cooling.md>), [cpu](<https://devfeed.tech/tags/cpu.md>), [data-center](<https://devfeed.tech/tags/data-center.md>), [design](<https://devfeed.tech/tags/design.md>), [desktop](<https://devfeed.tech/tags/desktop.md>), [dgx](<https://devfeed.tech/tags/dgx.md>), [dgx-station](<https://devfeed.tech/tags/dgx-station.md>), [gb300](<https://devfeed.tech/tags/gb300.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [review](<https://devfeed.tech/tags/review.md>), [server](<https://devfeed.tech/tags/server.md>), [supermicro](<https://devfeed.tech/tags/supermicro.md>), [tower](<https://devfeed.tech/tags/tower.md>), [workstation](<https://devfeed.tech/tags/workstation.md>)

### AI overview

Supermicro's Super AI Station (ARS-511GD-NB-LCC) pairs NVIDIA Grace Blackwell GB300 hardware with a chassis that can serve as a desktop tower or a 5U rack system. Its 455 mm height fits between rack posts, while its 701 mm depth and 88 lb weight reflect the dual-purpose design. The 1.6 kW system uses liquid cooling, with an externally accessible coolant tank, and an 80 Plus Titanium power supply. Rear connectivity includes two 400 Gbps QSFP112 ports through NVIDIA's ConnectX-8 SuperNIC, 1Gb Ethernet for BMC management, 10Gb Ethernet, and four 10Gbps USB ports; there are no USB-C ports. The excerpt notes that the early system's rear I/O labels are absent and its front USB-A ports are slower than those on some other DGX Station systems. Its BMC provides the only graphics output unless a separate graphics card is used; this unit includes an RTX PRO 2000 Blackwell card.

### Source excerpt

Today we are looking at Supermicro's Super AI Station, the company's neat take on a DGX Station System with a design that can stand as a tower or be rackmounted as a 5U server The post Supermicro Super AI Station Review: The AI Server That Doubles as a Workstation appeared first on ServeTheHome.

## Qualcomm Dragonwing Nabs a Big SpaceX Starlink Mobile Nationwide US 5G Network Win

DevFeed: [Qualcomm Dragonwing Nabs a Big SpaceX Starlink Mobile Nationwide US 5G Network Win](<https://devfeed.tech/articles/qualcomm-dragonwing-nabs-a-big-spacex-starlink-mobile-nationwide-us-5g-network-win-118636.md>)

Original publisher: [Read original article](<https://www.servethehome.com/qualcomm-dragonwing-nabs-a-big-spacex-starlink-mobile-nationwide-us-5g-network-win/>)

Author: Vic A

Published: 2026-11-04T21:32:22Z

Content type: news

Language: en

Sources: [ServeTheHome](<https://devfeed.tech/sources/servethehome.md>)

Topics: [5G](<https://devfeed.tech/topics/5g.md>), [Qualcomm](<https://devfeed.tech/topics/qualcomm.md>)

Tags: [5g](<https://devfeed.tech/tags/5g.md>), [5g-edge](<https://devfeed.tech/tags/5g-edge.md>), [accelerate](<https://devfeed.tech/tags/accelerate.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-compute](<https://devfeed.tech/tags/ai-compute.md>), [antenna](<https://devfeed.tech/tags/antenna.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [base](<https://devfeed.tech/tags/base.md>), [build](<https://devfeed.tech/tags/build.md>), [capacity](<https://devfeed.tech/tags/capacity.md>), [carrier](<https://devfeed.tech/tags/carrier.md>), [cellular](<https://devfeed.tech/tags/cellular.md>), [customer](<https://devfeed.tech/tags/customer.md>), [mobile](<https://devfeed.tech/tags/mobile.md>), [network](<https://devfeed.tech/tags/network.md>), [qualcomm](<https://devfeed.tech/tags/qualcomm.md>), [qualcomm-dragonwing](<https://devfeed.tech/tags/qualcomm-dragonwing.md>), [spacex](<https://devfeed.tech/tags/spacex.md>), [starlink](<https://devfeed.tech/tags/starlink.md>), [starlink-mobile](<https://devfeed.tech/tags/starlink-mobile.md>), [us](<https://devfeed.tech/tags/us.md>)

### AI overview

SpaceX selected Qualcomm Dragonwing platforms for Starlink Mobile's planned U.S. terrestrial 5G network, with rollout expected to begin in 2027 under a multi-year agreement covering tens of thousands of sites. The QRU100 supports 64T64R Massive MIMO radio units, while the QDU100 handles latency-sensitive Layer 1 processing; the excerpt suggests Qualcomm may also provide the L1 software, but says this is not certain. Qualcomm says the platforms are designed to improve performance per watt across radio and distributed-unit processing, which could help deployment economics and sites with power constraints. The excerpt does not identify who supplies Layer 2 and higher functions.

### Source excerpt

Qualcomm Dragonwing just landed a new US customer as part of the SpaceX Starlink mobile nationwide build-out The post Qualcomm Dragonwing Nabs a Big SpaceX Starlink Mobile Nationwide US 5G Network Win appeared first on ServeTheHome.

## AMD Q3 2026 Earnings Data Center Growth Drives Record Quarter

DevFeed: [AMD Q3 2026 Earnings Data Center Growth Drives Record Quarter](<https://devfeed.tech/articles/amd-q3-2026-earnings-data-center-growth-drives-record-quarter-117159.md>)

Original publisher: [Read original article](<https://www.servethehome.com/amd-q3-2026-earnings-data-center-growth/>)

Author: Cliff Robinson

Published: 2026-11-04T05:24:08Z

Content type: article

Language: en

Sources: [ServeTheHome](<https://devfeed.tech/sources/servethehome.md>)

Topics: [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [x86](<https://devfeed.tech/topics/x86.md>), [ddr5](<https://devfeed.tech/topics/ddr5.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [acceleration](<https://devfeed.tech/tags/acceleration.md>), [aerospace](<https://devfeed.tech/tags/aerospace.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-infrastructure](<https://devfeed.tech/tags/ai-infrastructure.md>), [amd](<https://devfeed.tech/tags/amd.md>), [anthropic](<https://devfeed.tech/tags/anthropic.md>), [data-center](<https://devfeed.tech/tags/data-center.md>), [data-center-growth](<https://devfeed.tech/tags/data-center-growth.md>), [data-center-revenue](<https://devfeed.tech/tags/data-center-revenue.md>), [defense](<https://devfeed.tech/tags/defense.md>), [demand](<https://devfeed.tech/tags/demand.md>), [earnings](<https://devfeed.tech/tags/earnings.md>), [embedded](<https://devfeed.tech/tags/embedded.md>), [growth](<https://devfeed.tech/tags/growth.md>), [leadership](<https://devfeed.tech/tags/leadership.md>), [news](<https://devfeed.tech/tags/news.md>), [q3](<https://devfeed.tech/tags/q3.md>), [q3-2026-earnings](<https://devfeed.tech/tags/q3-2026-earnings.md>), [venice](<https://devfeed.tech/tags/venice.md>), [x86](<https://devfeed.tech/tags/x86.md>)

### AI overview

AMD reported Q3 2026 revenue of $13.377 billion, up 45% year over year, with Data Center revenue nearly doubling to $8.546 billion. AMD attributes the segment's growth to strong EPYC demand and the launch of Instinct MI450 GPUs; it says major clients are running workloads on Helios rack-scale systems. The article compares AMD's Data Center growth of 97% with Intel's 78% on $7.3 billion in quarterly revenue. Client and Gaming revenue fell 8% to $3.722 billion, while Embedded rose 29% to $1.109 billion. AMD forecast Q4 revenue of about $17 billion, plus or minus 3%, implying roughly 66% year-over-year growth at the midpoint. The article's expectations about Venice's future x86 performance leadership and Intel's Diamond Rapids timing are projections, not reported results.

### Source excerpt

AMD Q3 2026 earnings were massive with the data center and embedded segments seeing big uplift with the clear Venice x86 leadership looming The post AMD Q3 2026 Earnings Data Center Growth Drives Record Quarter appeared first on ServeTheHome.

## Marvell Proposes Acquiring Lumentum to Combine Optical Circuit Switching with AI Connectivity Silicon

DevFeed: [Marvell Proposes Acquiring Lumentum to Combine Optical Circuit Switching with AI Connectivity Silicon](<https://devfeed.tech/articles/marvell-proposes-acquiring-lumentum-to-combine-optical-circuit-switching-with-ai-connectivity-silicon-113893.md>)

Original publisher: [Read original article](<https://www.servethehome.com/marvell-to-acquire-lumentum-115-billion-optical-ai-connectivity/>)

Author: Cliff Robinson

Published: 2026-11-02T22:30:09Z

Content type: news

Language: en

Sources: [ServeTheHome](<https://devfeed.tech/sources/servethehome.md>)

Topics: [communications](<https://devfeed.tech/topics/communications.md>), [Network architectures](<https://devfeed.tech/topics/network-architectures.md>), [Chip design](<https://devfeed.tech/topics/chip-design.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>)

Tags: [acquisition](<https://devfeed.tech/tags/acquisition.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-clusters](<https://devfeed.tech/tags/ai-clusters.md>), [ai-connectivity](<https://devfeed.tech/tags/ai-connectivity.md>), [ai-infrastructure](<https://devfeed.tech/tags/ai-infrastructure.md>), [co-design](<https://devfeed.tech/tags/co-design.md>), [co-packaged-optics](<https://devfeed.tech/tags/co-packaged-optics.md>), [coherent](<https://devfeed.tech/tags/coherent.md>), [compute](<https://devfeed.tech/tags/compute.md>), [connectivity](<https://devfeed.tech/tags/connectivity.md>), [copper](<https://devfeed.tech/tags/copper.md>), [cpo](<https://devfeed.tech/tags/cpo.md>), [custom-silicon](<https://devfeed.tech/tags/custom-silicon.md>), [data-center](<https://devfeed.tech/tags/data-center.md>), [lumentum](<https://devfeed.tech/tags/lumentum.md>), [marvell](<https://devfeed.tech/tags/marvell.md>), [networking](<https://devfeed.tech/tags/networking.md>), [optical](<https://devfeed.tech/tags/optical.md>), [optical-connectivity](<https://devfeed.tech/tags/optical-connectivity.md>), [silicon](<https://devfeed.tech/tags/silicon.md>), [switching](<https://devfeed.tech/tags/switching.md>)

### AI overview

Marvell proposes acquiring Lumentum in an all-stock deal valued at about $115 billion, combining Marvell's interconnect and custom silicon capabilities with Lumentum's lasers, compound semiconductor manufacturing, and optical circuit switching (OCS). The stated aim is closer co-design across the signal chain as rising lane speeds bring copper's distance limits closer and AI clusters demand more connectivity. The combination could support tighter integration such as co-packaged optics: near-package optical engines shorten electrical paths, replacing long PCB traces with fiber runs and pluggable light sources. Lumentum's R300 OCS uses MEMS mirrors to establish direct optical paths, avoiding intermediate packet processing and optical-electrical-optical conversion on switched paths; Lumentum says this lowers latency and network power while improving signal integrity. The article describes those benefits as company claims. Marvell intends to bring more of the supply chain in-house and develop integrated, pre-validated solutions. The deal is expected to close in the first half of 2028; afterward, Marvell shareholders would own 67% and Lumentum shareholders 33%.

### Source excerpt

Marvell announced its intention to acquire Lumentum in a $115B acquisition to further optical connectivity in the AI era The post Marvell Proposes Acquiring Lumentum to Combine Optical Circuit Switching with AI Connectivity Silicon appeared first on ServeTheHome.

## Qualcomm Validates HBC Stacked Memory Ahead of 2027 Dragonfly AI250 Sampling

DevFeed: [Qualcomm Validates HBC Stacked Memory Ahead of 2027 Dragonfly AI250 Sampling](<https://devfeed.tech/articles/qualcomm-validates-hbc-stacked-memory-ahead-of-2027-dragonfly-ai250-sampling-113894.md>)

Original publisher: [Read original article](<https://www.servethehome.com/qualcomm-validates-hbc-stacked-memory-ahead-of-2027-dragonfly-ai250-sampling/>)

Author: Ryan Smith

Published: 2026-11-02T14:00:14Z

Content type: news

Language: en

Sources: [ServeTheHome](<https://devfeed.tech/sources/servethehome.md>)

Topics: [Qualcomm](<https://devfeed.tech/topics/qualcomm.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>), [DRAM price increase](<https://devfeed.tech/topics/dram-price-increase.md>)

Tags: [accelerators](<https://devfeed.tech/tags/accelerators.md>), [advanced-packaging](<https://devfeed.tech/tags/advanced-packaging.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-accelerators](<https://devfeed.tech/tags/ai-accelerators.md>), [ai-inference](<https://devfeed.tech/tags/ai-inference.md>), [architecture](<https://devfeed.tech/tags/architecture.md>), [balance](<https://devfeed.tech/tags/balance.md>), [bandwidth](<https://devfeed.tech/tags/bandwidth.md>), [capacity](<https://devfeed.tech/tags/capacity.md>), [chip](<https://devfeed.tech/tags/chip.md>), [chiplet](<https://devfeed.tech/tags/chiplet.md>), [compute](<https://devfeed.tech/tags/compute.md>), [cost](<https://devfeed.tech/tags/cost.md>), [dragonfly](<https://devfeed.tech/tags/dragonfly.md>), [dram](<https://devfeed.tech/tags/dram.md>), [hbc](<https://devfeed.tech/tags/hbc.md>), [high-bandwidth-compute](<https://devfeed.tech/tags/high-bandwidth-compute.md>), [memory](<https://devfeed.tech/tags/memory.md>), [qualcomm](<https://devfeed.tech/tags/qualcomm.md>), [sampling](<https://devfeed.tech/tags/sampling.md>), [server](<https://devfeed.tech/tags/server.md>)

### AI overview

Qualcomm says initial validation of High Bandwidth Compute (HBC) succeeded, advancing the stacked-memory design toward commercialization in its Dragonfly AI250 accelerator. HBC places LPDDR5X DRAM above a small compute die, which handles bandwidth-intensive work near memory and sends results to the larger processor. Qualcomm argues this reduces data movement and energy use while requiring less costly processor connectivity than HBM; it still depends on challenging DRAM stacking with TSVs. In tests of an early two-layer stack, the company reports over five times HBM4's bandwidth per watt and bandwidth 18 times higher than external LPDDR5X, without disclosing absolute bandwidth. The test stack had fewer layers than the planned four-layer product, and Qualcomm's efficiency result was below its initial sixfold target. Thermal behavior was included in validation, but details were not provided. AI250 is scheduled to begin sampling in the first half of 2027; final product performance and broader deployment remain uncertain.

### Source excerpt

Qualcomm today is announcing that they have completed initial validation on their High Bandwidth Compute technology. The DRAM-stacked-upon-compute architecture moves on to commercialization, with Dragonfly AI250 set to sample in the first half of 2027 The post Qualcomm Validates HBC Stacked Memory Ahead of 2027 Dragonfly AI250 Sampling appeared first on ServeTheHome.

## Liquid-Cooled NVIDIA Servers at the ASRock Rack OCP 2026 Booth

DevFeed: [Liquid-Cooled NVIDIA Servers at the ASRock Rack OCP 2026 Booth](<https://devfeed.tech/articles/liquid-cooled-nvidia-servers-at-the-asrock-rack-ocp-2026-booth-112200.md>)

Original publisher: [Read original article](<https://www.servethehome.com/liquid-cooled-nvidia-servers-at-the-asrock-rack-ocp-2026-booth/>)

Author: Patrick Kennedy

Published: 2026-11-02T04:00:03Z

Content type: news

Language: en

Sources: [ServeTheHome](<https://devfeed.tech/sources/servethehome.md>)

Topics: [Vera Rubin NVL72](<https://devfeed.tech/topics/vera-rubin-nvl72.md>), [NVIDIA Vera Rubin](<https://devfeed.tech/topics/nvidia-vera-rubin.md>), [Nvidia](<https://devfeed.tech/topics/nvidia.md>), [NVLink](<https://devfeed.tech/topics/nvlink.md>), [RTX PRO](<https://devfeed.tech/topics/rtx-pro.md>), [NVIDIA Spectrum-X Ethernet](<https://devfeed.tech/topics/nvidia-spectrum-x-ethernet.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [ai](<https://devfeed.tech/tags/ai.md>), [asrock-rack](<https://devfeed.tech/tags/asrock-rack.md>), [blackwell](<https://devfeed.tech/tags/blackwell.md>), [bluefield-4](<https://devfeed.tech/tags/bluefield-4.md>), [bluefield-4-dpus](<https://devfeed.tech/tags/bluefield-4-dpus.md>), [board](<https://devfeed.tech/tags/board.md>), [chip](<https://devfeed.tech/tags/chip.md>), [compute](<https://devfeed.tech/tags/compute.md>), [connectx](<https://devfeed.tech/tags/connectx.md>), [cooling](<https://devfeed.tech/tags/cooling.md>), [cpus](<https://devfeed.tech/tags/cpus.md>), [dc](<https://devfeed.tech/tags/dc.md>), [liquid](<https://devfeed.tech/tags/liquid.md>), [liquid-cooling](<https://devfeed.tech/tags/liquid-cooling.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [nvl72](<https://devfeed.tech/tags/nvl72.md>), [nvl72-racks](<https://devfeed.tech/tags/nvl72-racks.md>), [ocp](<https://devfeed.tech/tags/ocp.md>), [rack](<https://devfeed.tech/tags/rack.md>), [rubin](<https://devfeed.tech/tags/rubin.md>), [server](<https://devfeed.tech/tags/server.md>), [server-systems](<https://devfeed.tech/tags/server-systems.md>), [servers](<https://devfeed.tech/tags/servers.md>), [spectrum-x](<https://devfeed.tech/tags/spectrum-x.md>), [switches](<https://devfeed.tech/tags/switches.md>), [vera](<https://devfeed.tech/tags/vera.md>)

### AI overview

At OCP 2026, ASRock Rack showed NVIDIA systems spanning a 48RU Vera Rubin NVL72 rack, HGX Rubin NVL8 servers, RTX PRO systems, a Vera CPU server, and a liquid-cooled Spectrum-X switch. Liquid cooling was a central feature: the NVL72 compute trays and HGX systems use direct-to-chip cooling. The NVL72 rack combines 18 compute trays and nine NVLink 6 switch trays, totaling 36 Vera CPUs and 72 Rubin GPUs; ASRock Rack lists several cooling-distribution options, including in-rack and liquid-to-air sidecar CDUs. The 2U HGX Rubin NVL8 system fits eight Rubin GPUs in a fanless chassis and liquid-cools its drives, CPUs, and GPUs. It uses a dual-socket Intel Xeon 6 host, a 54V DC busbar, eight 800Gbps ConnectX-9 ports, and eight hot-swap E1.S NVMe bays. The excerpt gives NVIDIA performance figures for both platforms, but does not detail the other systems mentioned.

### Source excerpt

We checked out the ASRock Rack booth at OCP 2026 and saw servers ranging from fully integrated NVIDIA NVL72 racks to liquid-cooled switches The post Liquid-Cooled NVIDIA Servers at the ASRock Rack OCP 2026 Booth appeared first on ServeTheHome.

## NVIDIA Updates Rubin GPU Specs as First Production Hardware Arrives

DevFeed: [NVIDIA Updates Rubin GPU Specs as First Production Hardware Arrives](<https://devfeed.tech/articles/nvidia-updates-rubin-gpu-specs-as-first-production-hardware-arrives-110694.md>)

Original publisher: [Read original article](<https://www.servethehome.com/nvidia-updates-rubin-gpu-specs-as-first-production-hardware-arrives/>)

Author: Ryan Smith

Published: 2026-11-01T03:00:28Z

Content type: news

Language: en

Sources: [ServeTheHome](<https://devfeed.tech/sources/servethehome.md>)

Topics: [NVIDIA Vera Rubin](<https://devfeed.tech/topics/nvidia-vera-rubin.md>), [NVLink](<https://devfeed.tech/topics/nvlink.md>)

Tags: [accelerators](<https://devfeed.tech/tags/accelerators.md>), [ai](<https://devfeed.tech/tags/ai.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [hbm4](<https://devfeed.tech/tags/hbm4.md>), [nvidia](<https://devfeed.tech/tags/nvidia.md>), [nvl72](<https://devfeed.tech/tags/nvl72.md>), [nvl8](<https://devfeed.tech/tags/nvl8.md>), [rubin](<https://devfeed.tech/tags/rubin.md>), [rubin-gpu](<https://devfeed.tech/tags/rubin-gpu.md>), [server](<https://devfeed.tech/tags/server.md>), [shipping](<https://devfeed.tech/tags/shipping.md>), [specifications](<https://devfeed.tech/tags/specifications.md>), [updated](<https://devfeed.tech/tags/updated.md>)

### AI overview

NVIDIA's shipping Rubin specifications reduce HBM4 and NVLink bandwidth while increasing scale-out networking bandwidth. Per Rubin GPU, memory bandwidth falls from 22 to 19.2 TB/s and NVLink from 3.6 to 3.0 TB/s; networking rises from 400 to 450 GB/s. Vera Rubin NVL72 reflects those changes, with GPU memory capacity unchanged at 20.7 TB. HGX Rubin NVL8 also loses HBM4 capacity, from 2.3 to 2.0 TB, and sees a larger 26% memory-bandwidth reduction; its FP4 and FP8 training figures are also lower. Compute figures for a single Rubin GPU remain unchanged. NVIDIA has not explained the revisions, and the excerpt says performance effects depend on workload bottlenecks.

### Source excerpt

As production hardware starts shipping, NVIDIA has updated some of the specifications for its Rubin GPU platforms The post NVIDIA Updates Rubin GPU Specs as First Production Hardware Arrives appeared first on ServeTheHome.

## Intel Q3 2026 Earnings Data Center Soars While PC Faces Memory Pressure

DevFeed: [Intel Q3 2026 Earnings Data Center Soars While PC Faces Memory Pressure](<https://devfeed.tech/articles/intel-q3-2026-earnings-data-center-soars-while-pc-faces-memory-pressure-109887.md>)

Original publisher: [Read original article](<https://www.servethehome.com/intel-q3-2026-earnings-data-center-foundry-client/>)

Author: Cliff Robinson

Published: 2026-10-30T22:43:29Z

Content type: article

Language: en

Sources: [ServeTheHome](<https://devfeed.tech/sources/servethehome.md>)

Topics: [memory price increases](<https://devfeed.tech/topics/memory-price-increases.md>), [DRAM price increase](<https://devfeed.tech/topics/dram-price-increase.md>), [intel](<https://devfeed.tech/topics/intel.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [client](<https://devfeed.tech/tags/client.md>), [data-center](<https://devfeed.tech/tags/data-center.md>), [earnings](<https://devfeed.tech/tags/earnings.md>), [intel](<https://devfeed.tech/tags/intel.md>), [memory](<https://devfeed.tech/tags/memory.md>), [memory-pressure](<https://devfeed.tech/tags/memory-pressure.md>), [news](<https://devfeed.tech/tags/news.md>), [pc](<https://devfeed.tech/tags/pc.md>), [q3](<https://devfeed.tech/tags/q3.md>), [q3-2026-earnings](<https://devfeed.tech/tags/q3-2026-earnings.md>)

### AI overview

Intel reported Q3 2026 revenue of $16.5 billion, up 21% year over year, with non-GAAP gross margin of 44.1% and EPS of $0.52. Data Center and AI revenue rose 78% to $7.3 billion, with a 41.5% operating margin; demand is strong across customer types, but substrate availability is limiting Xeon supply. The excerpt says AMD's 192-core Turin parts address a segment Intel lacks a direct competitor for, while Xeon 6 is selling amid broad demand. Client Computing and Physical AI revenue fell 5% to $8.1 billion as DRAM and NAND shortages and higher prices raise PC system costs and weigh on demand; Intel expects the slowdown to continue through at least the first half of next year, with memory price pressure into 2027. Intel Foundry revenue rose 58% to $6.7 billion, including intersegment transactions, while its operating loss narrowed to $1.9 billion; improved yields and cycle times are reducing wafer costs.

### Source excerpt

Intel's Q3 2026 earnings highlighted how well the data center segment is doing and how rough the client segment is due to DRAM and NAND pricing The post Intel Q3 2026 Earnings Data Center Soars While PC Faces Memory Pressure appeared first on ServeTheHome.

## ASRock Industrial AI Box-A395 Review: Taking Strix Halo to Work

DevFeed: [ASRock Industrial AI Box-A395 Review: Taking Strix Halo to Work](<https://devfeed.tech/articles/asrock-industrial-ai-box-a395-review-taking-strix-halo-to-work-108360.md>)

Original publisher: [Read original article](<https://www.servethehome.com/asrock-industrial-ai-box-a395-review-taking-strix-halo-to-work/>)

Author: Ryan Smith

Published: 2026-10-29T18:00:14Z

Content type: article

Language: en

Sources: [ServeTheHome](<https://devfeed.tech/sources/servethehome.md>)

Topics: [Frontier AI](<https://devfeed.tech/topics/frontier-ai.md>), [renesas](<https://devfeed.tech/topics/renesas.md>)

Tags: [10gbe](<https://devfeed.tech/tags/10gbe.md>), [2-5gbe](<https://devfeed.tech/tags/2-5gbe.md>), [2d](<https://devfeed.tech/tags/2d.md>), [3](<https://devfeed.tech/tags/3.md>), [added](<https://devfeed.tech/tags/added.md>), [agents](<https://devfeed.tech/tags/agents.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-chip](<https://devfeed.tech/tags/ai-chip.md>), [ai-inference](<https://devfeed.tech/tags/ai-inference.md>), [ai-workstation](<https://devfeed.tech/tags/ai-workstation.md>), [aluminum](<https://devfeed.tech/tags/aluminum.md>), [amd](<https://devfeed.tech/tags/amd.md>), [antenna](<https://devfeed.tech/tags/antenna.md>), [apple](<https://devfeed.tech/tags/apple.md>), [asrock-industrial](<https://devfeed.tech/tags/asrock-industrial.md>), [audio](<https://devfeed.tech/tags/audio.md>), [mini-pc](<https://devfeed.tech/tags/mini-pc.md>), [portable](<https://devfeed.tech/tags/portable.md>), [review](<https://devfeed.tech/tags/review.md>), [ryzen-ai](<https://devfeed.tech/tags/ryzen-ai.md>), [ryzen-ai-max](<https://devfeed.tech/tags/ryzen-ai-max.md>), [strix-halo](<https://devfeed.tech/tags/strix-halo.md>), [workstation](<https://devfeed.tech/tags/workstation.md>)

### AI overview

The ASRock Industrial AI Box-A395 is a portable, self-contained Ryzen AI Max+ 395 mini-PC aimed more at industrial and headless use than a conventional workstation. Its upright aluminum micro-tower saves floor space, includes an internal power supply and optional carrying handle, and has a distinctive, vented design. Connectivity is concentrated on the rear: two HDMI 2.1 outputs and three USB-C ports can support up to four displays; two USB-C ports provide 40Gbps USB4, alongside 10Gbps USB-A and USB-C ports. Networking includes 10GbE, 2.5GbE and Wi-Fi 7 with external antenna connectors. The two top-facing USB-A ports are only USB 2.x, making them suitable for basic peripherals rather than high-speed devices. The excerpt describes the hardware exterior and ports; it does not provide performance or thermal test results.

### Source excerpt

Today we are taking a look at ASRock Industrial's highly portable Ryzen AI Max+ 395 system, the AI Box-A395. The barebones tower system packs a punch, as well as a handle to quickly pack up the PC The post ASRock Industrial AI Box-A395 Review: Taking Strix Halo to Work appeared first on ServeTheHome.

## Microsoft Maia 300 Shown at OCP 2026

DevFeed: [Microsoft Maia 300 Shown at OCP 2026](<https://devfeed.tech/articles/microsoft-maia-300-shown-at-ocp-2026-102717.md>)

Original publisher: [Read original article](<https://www.servethehome.com/microsoft-maia-300-shown-at-ocp-2026/>)

Author: Patrick Kennedy

Published: 2026-10-27T21:28:01Z

Content type: news

Language: en

Sources: [ServeTheHome](<https://devfeed.tech/sources/servethehome.md>)

Topics: [Microsoft](<https://devfeed.tech/topics/microsoft.md>), [Vera Rubin NVL72](<https://devfeed.tech/topics/vera-rubin-nvl72.md>)

Tags: [2026](<https://devfeed.tech/tags/2026.md>), [acceleration](<https://devfeed.tech/tags/acceleration.md>), [accelerators](<https://devfeed.tech/tags/accelerators.md>), [ai](<https://devfeed.tech/tags/ai.md>), [ai-accelerators](<https://devfeed.tech/tags/ai-accelerators.md>), [amd](<https://devfeed.tech/tags/amd.md>), [amd-epyc](<https://devfeed.tech/tags/amd-epyc.md>), [asics](<https://devfeed.tech/tags/asics.md>), [azure](<https://devfeed.tech/tags/azure.md>), [bandwidth](<https://devfeed.tech/tags/bandwidth.md>), [hbm4](<https://devfeed.tech/tags/hbm4.md>), [maia](<https://devfeed.tech/tags/maia.md>), [maia-300](<https://devfeed.tech/tags/maia-300.md>), [microsoft](<https://devfeed.tech/tags/microsoft.md>), [microsoft-maia-300](<https://devfeed.tech/tags/microsoft-maia-300.md>), [news](<https://devfeed.tech/tags/news.md>), [ocp](<https://devfeed.tech/tags/ocp.md>), [ocp-summit](<https://devfeed.tech/tags/ocp-summit.md>), [platform](<https://devfeed.tech/tags/platform.md>), [summit](<https://devfeed.tech/tags/summit.md>), [tray](<https://devfeed.tech/tags/tray.md>)

### AI overview

At OCP Summit 2026, Microsoft showed Maia 300 in a four-accelerator node. Each chip has eight HBM4 stacks, 288GB of memory, 18TB/s memory bandwidth, 3.2TB/s bidirectional scale-up bandwidth, and a 950W TDP. Microsoft says a rack can hold 96 Maia accelerators, with roughly 12,000 in the scale-up domain. The node combines liquid-cooled cold plates for the accelerators with a large fan partition cooling the CPUs, SSDs, NICs, and other components. Its design uses extensive cabled connections and multiple accelerator PCBs; the author contrasts this with NVIDIA's stated push toward fewer cables in Vera Rubin NVL72. The article notes the chassis is wider than a standard OU but narrower than AMD Helios, and that the tray's components exceed 4kW TDP.

### Source excerpt

At OCP Summit 2026, we saw the new Microsoft Maia 300 platform with four 288GB HBM4 950W accelerators per tray The post Microsoft Maia 300 Shown at OCP 2026 appeared first on ServeTheHome.

## ASRock Creator Radeon AI Pro R9700 Review: Workstation Inference, AMD Style

DevFeed: [ASRock Creator Radeon AI Pro R9700 Review: Workstation Inference, AMD Style](<https://devfeed.tech/articles/asrock-creator-radeon-ai-pro-r9700-review-workstation-inference-amd-style-100674.md>)

Original publisher: [Read original article](<https://www.servethehome.com/asrock-creator-radeon-ai-pro-r9700-review-workstation-inference-amd-style/>)

Author: Ryan Smith

Published: 2026-10-26T18:07:04Z

Content type: article

Language: en

Sources: [ServeTheHome](<https://devfeed.tech/sources/servethehome.md>)

Topics: [Local AI](<https://devfeed.tech/topics/local-ai.md>), [quantization](<https://devfeed.tech/topics/quantization.md>), [AI Infrastructure](<https://devfeed.tech/topics/ai-infrastructure.md>)

Tags: [ai](<https://devfeed.tech/tags/ai.md>), [ai-accelerator](<https://devfeed.tech/tags/ai-accelerator.md>), [ai-accelerators](<https://devfeed.tech/tags/ai-accelerators.md>), [ai-inference](<https://devfeed.tech/tags/ai-inference.md>), [ai-workstation](<https://devfeed.tech/tags/ai-workstation.md>), [amd](<https://devfeed.tech/tags/amd.md>), [asrock](<https://devfeed.tech/tags/asrock.md>), [gpu](<https://devfeed.tech/tags/gpu.md>), [inference](<https://devfeed.tech/tags/inference.md>), [radeon](<https://devfeed.tech/tags/radeon.md>), [radeon-pro](<https://devfeed.tech/tags/radeon-pro.md>), [rdna-4](<https://devfeed.tech/tags/rdna-4.md>), [review](<https://devfeed.tech/tags/review.md>), [workstation](<https://devfeed.tech/tags/workstation.md>), [workstation-processors](<https://devfeed.tech/tags/workstation-processors.md>)

### AI overview

ASRock's Creator Radeon AI Pro R9700 pairs AMD's RDNA 4 Navi 48 GPU with 32GB of GDDR6 and a 300W dual-slot blower design aimed at workstation inference. Its 64 compute units include 128 AI accelerators; FP8 matrix throughput is rated at 382.7 TFLOPS, or twice that with sparsity. FP8 support improves low-precision throughput, while FP6 and FP4 weights are supported but processed at FP8; the article cites FP6 as a way to fit an approximately 30-billion-parameter model in 32GB. The card supports soft ECC only on Linux, with manual enabling. Its 32GB capacity is below the previous-generation Radeon Pro W7900's 48GB, and the excerpt gives no performance benchmark results. The card uses a vapor chamber, rear-mounted 12V-2x6 power connector, and four DisplayPort 2.1a outputs.

### Source excerpt

Today we are taking a look at ASRock's high-end Radeon AI inference card for workstations, the Creator Radeon AI Pro R9700. With 32GB of RAM and a $1700 price tag, the card is a highly affordable way to fill out a workstation for local AI inference The post ASRock Creator Radeon AI Pro R9700 Review: Workstation Inference, AMD Style appeared first on ServeTheHome.

[Next page](<https://devfeed.tech/sources/servethehome.md?cursor=WyIyMDI2LTEwLTI2VDE4OjA3OjA0KzAwOjAwIiwgImI5OWI4MWYxLTE3OTAtNDBiMC04MTU5LTg4Y2M0Mjk3MTQxOCJd>)