SendTech Times
News
INFRASTRUCTURE CONSTRAINT:

AMD Unveils Threadripper Halo Station With Dual MI350P Accelerators

Newsroom brief

ServeTheHome reports that AMD’s IFA demo pairs a 96-core Threadripper Pro platform with two MI350P accelerators, 288GB of GPU memory and liquid cooling for high-end local AI development.

Verified against source materialEdited by SendTech Times Chips & Compute DeskSource: ServeTheHome
AMD Unveils Threadripper Halo Station With Dual MI350P Accelerators
Image source: ServeTheHome

AMD has used its IFA keynote in Berlin to put a larger local-AI developer machine beside its smaller Ryzen AI Halo box, with ServeTheHome’s coverage identifying the new Threadripper Halo Station as a tower workstation built around workstation-class CPUs and Instinct accelerators.

The system moves the Halo line from mini-PC experimentation into a much heavier local inference and development configuration.

The demonstration unit pairs a Threadripper Pro platform with discrete MI350P accelerator cards, replacing the Ryzen AI Max+ processor used in the earlier Strix Halo device.

That hardware choice defines the workstation’s role.

AMD’s demonstration specifications listed a 96-core Ryzen Threadripper PRO 9995WX, a Zen 5 workstation processor with 192 threads, eight channels of DDR5 memory and 128 lanes of PCIe Gen5 connectivity.

The key-specification table lists CPU memory support at up to 2TB, giving the host side far more room than compact AI developer boxes.

The accelerator configuration carries the main AI workload.

AMD’s base Threadripper Halo Station setup includes two Instinct MI350P PCIe cards, each with 144GB of HBM3E memory.

That creates a 288GB GPU-memory floor for the system, while a four-card path would raise local accelerator memory to 576GB if supported in shipping configurations.

Cooling is part of the workstation design.

The demo machine used liquid-cooled MI350P cards and liquid cooling across the major processors, while the original MI350P configuration was an air-cooled PCIe accelerator for standardized servers.

A tower workstation does not provide the same forced-air environment, so higher-capacity cooling becomes part of making 450-watt-class cards usable at a desk-side scale.

The tower also adds networking headroom compared with the smaller Ryzen AI Halo box.

ServeTheHome noted that the showcase unit looked to use an ASUS Pro WS WRX90E-SAGE SE motherboard and an unidentified SFP networking card.

The earlier mini-PC topped out at a 10Gb Ethernet jack, while a workstation handling larger model and data movement may need faster ingress and egress.

Microsoft appeared during the same keynote to discuss Project Zenith, a curated developer-toolkit package planned for pre-installation on Ryzen AI Halo systems.

Its availability on Threadripper Halo Station remains unspecified.

The market comparison is Nvidia’s DGX Station line.

Both products try to bring server-grade AI development hardware into a desktop form factor, but their architectures differ: Nvidia’s design centers on a B300 accelerator and Grace CPU module, while AMD is leaning on multiple PCIe GPUs attached to a Threadripper Pro workstation platform.

The display unit was assembled from existing commercial components rather than as a fully customized platform.

It used separate closed-loop radiators for the processors instead of one integrated cooling system.

AMD described a possible path to four MI350P cards.

Public specifications still omit the operating system, storage configuration, power supply and final networking details.

They also leave open whether the four-card layout will ship as a factory configuration or remain a customer expansion.

Those gaps separate the demonstrated hardware from a shipping workstation specification.

AMD has yet to announce pricing or availability for Threadripper Halo Station.

Share this article
inXf

Related articles

More
Lenovo Whitsett AI Server Line Doubles Plant To 890,000 Square Feet
Chips & Semiconductors

Lenovo Whitsett AI Server Line Doubles Plant To 890,000 Square Feet

A Tom's Hardware tour details Lenovo's expanded Whitsett AI server line, including 890,000 square feet, up to 20 MW of power and liquid-cooling tests for GB300 and Vera Rubin racks.

Micron 9650 SSD Promises 28GB/Second Reads Before PCIe Gen6 Servers Ship
Chips & Semiconductors

Micron 9650 SSD Promises 28GB/Second Reads Before PCIe Gen6 Servers Ship

Micron’s 9650 enterprise SSD is rated for up to 28GB/second reads and 5.5M random-read IOPS, with liquid cooling expected for the E1. S form factor. Public PCIe Gen6 server platforms were not yet available for a Computex demo.

AMD Prices 256-Core EPYC 9996 At $14,904 For Server Buyers
Chips & Semiconductors

AMD Prices 256-Core EPYC 9996 At $14,904 For Server Buyers

TechRadar reports that AMD’s 6th Gen EPYC 9006 “Venice” lineup includes a 256-core EPYC 9996 with 512 threads, 1GB of L3 cache, a 600W default power rating and a $14,904 list price for 1,000-unit orders.

AMD’s EPYC 9006 Reset Pushes Venice Toward Denser AI Racks
Chips & Semiconductors

AMD’s EPYC 9006 Reset Pushes Venice Toward Denser AI Racks

AMD’s forthcoming EPYC 9006 Venice family adds Zen 6 cores, higher memory bandwidth, PCIe 6, CXL 3.1 and cache-coherent CPU-GPU links for AI-oriented servers.

AMD Deal For Taalas Targets Model-Specific AI Inference
Chips & Semiconductors

AMD Deal For Taalas Targets Model-Specific AI Inference

ServeTheHome wrote that AMD will acquire Taalas, adding model-specific inference silicon to a wider AI infrastructure stack built around Instinct, Helios, EPYC and ROCm.

AMD Makes $8.2 Billion World Labs Bet On Physical AI Compute
Chips & Semiconductors

AMD Makes $8.2 Billion World Labs Bet On Physical AI Compute

AMD agreed to acquire World Labs in an all-stock transaction valued at about $8.2 billion, bringing Fei-Fei Li’s spatial-intelligence research team into its AI hardware, software and systems roadmap.

Google TPU Rumor Points to AMD CPU IP Role
Chips & Semiconductors

Google TPU Rumor Points to AMD CPU IP Role

Tom’s Hardware reported that Google may be working with AMD on a 10th-generation TPU project, with CPU IP and advanced packaging emerging as possible reasons for the custom accelerator collaboration.

US Export-Control Shift Opens UAE AI Chip Access, DCD Reports
Chips & Semiconductors

US Export-Control Shift Opens UAE AI Chip Access, DCD Reports

Data Center Dynamics reported that the US government has moved the UAE into a lower-restriction export-control category, giving UAE companies broader access to advanced AI chips from Nvidia and AMD. The account names G42 and hyperscaler data-centre projects, but it does not list chip volumes, approved licences or specific UAE projects.

Keep Reading

More Stories

Latest
Ethereum Testnet Update Targets 200 Million-Gas BlocksCrypto/Web3Oct 6, 2026Ethereum Testnet Update Targets 200 Million-Gas BlocksEthereum developers released Prysm 7.2.1 so the Sepolia trial of Glamsterdam can test 200 million-gas blocks, more than three times the prior 60 million setting, before any main-network change.Kepler Targets 2027 Production for HBM Replacement MemoryCloud & Data CentersOct 6, 2026Kepler Targets 2027 Production for HBM Replacement MemoryEE Times reports that Kepler Computing is preparing 3D ferroelectric memory for 2027 production, promising higher capacity and bandwidth per watt while limiting reliance on advanced-node lithography.Yokogawa Opens Singapore Hub For Industrial Cyber ResilienceCapital & PolicyOct 6, 2026Yokogawa Opens Singapore Hub For Industrial Cyber ResilienceYokogawa Engineering Asia has launched a Singapore center focused on OT cyber resilience, training, response planning and recovery coordination for Southeast Asia, Oceania and Taiwan.ClickFix Attack Uses Browser Cache To Hide Malware PayloadCybersecurityOct 6, 2026ClickFix Attack Uses Browser Cache To Hide Malware PayloadMicrosoft Threat Intelligence traced a ClickFix cache-smuggling method that preloads malware into browser caches, then uses file size checks and a pasted Run command to launch later credential-theft stages.VOA Tests Six-Month Startup Buildout Before Funding DecisionsFintech & Digital PaymentsOct 6, 2026VOA Tests Six-Month Startup Buildout Before Funding DecisionsTechCabal’s interview with VOA Venture Partners founder Victoria Olayide Adesanya describes a six-month build programme that lets the firm work inside African financial-infrastructure startups before deciding whether to invest.Bitcoin Holds $86,000 As Dollar Index Hits 18-Month HighCrypto/Web3Oct 6, 2026Bitcoin Holds $86,000 As Dollar Index Hits 18-Month HighCoinDesk reported that bitcoin stayed near $86,000 while the U.S. Dollar Index reached about 102.5, with U.S. rate expectations and European political risks strengthening the dollar backdrop.Google Freezes OSS Bug Bounty Reports After AI Submission FloodCybersecurityOct 6, 2026Google Freezes OSS Bug Bounty Reports After AI Submission FloodGoogle has stopped accepting new product vulnerability reports in its OSS VRP after invalid automated submissions swamped reviewers, while older reports and some Cloud VRP routes remain open.Fleuret AI Raises €4M For Continuous AI Pentesting PlatformCybersecurityOct 6, 2026Fleuret AI Raises €4M For Continuous AI Pentesting PlatformTech.eu reported that French startup Fleuret AI raised €4 million in pre-seed funding to develop an agentic-AI platform that turns penetration testing into a continuous security process.GFT Analysis Says AI Documentation Can Cut Maintenance Work 30%Fintech & Digital PaymentsOct 6, 2026GFT Analysis Says AI Documentation Can Cut Maintenance Work 30%A GFT Technologies analysis says AI-linked software documentation can cut maintenance effort and speed developer onboarding when knowledge assets stay synchronized with code changes.Schneider Electric Lines Up $22.6 Billion PTC DealAIOct 5, 2026Schneider Electric Lines Up $22.6 Billion PTC DealSchneider Electric plans to buy PTC in a cash transaction valuing the US engineering software provider’s equity at about $22.6 billion, adding product-lifecycle software to its industrial AI push.Aggarwal Pledges Ola Electric Stake To Fund ₹1,000 Cr Rights IssueCapital & PolicyOct 5, 2026Aggarwal Pledges Ola Electric Stake To Fund ₹1,000 Cr Rights IssueOla Electric founder Bhavish Aggarwal pledged 20 Cr shares to finance his participation in a rights issue that forms part of a larger ₹1,500 Cr fundraising plan.Natrona Schools AI Review Puts Student Privacy Ahead Of Classroom Tool UseAIOct 5, 2026Natrona Schools AI Review Puts Student Privacy Ahead Of Classroom Tool UseNatrona County trustees questioned whether teacher AI tools expose student data, even as existing district rules already ban unauthorized generative AI use by students.