#dpu — Public Fediverse posts

Eduardo Bolsonaro não comparece a interrogatório no Supremo

https://web.brid.gy/r/https://www.brasildefato.com.br/2026/04/14/eduardo-bolsonaro-nao-comparece-a-interrogatorio-no-supremo/

Ireland’s DPC launches EU data probe into X’s Grok AI for generating sexualized deepfake images, including of children. GDPR compliance questioned. https://english.mathrubhumi.com/technology/irish-eu-probe-x-grok-ai-sexual-deepfakes-hazmsemi?utm_source=dlvr.it&utm_medium=mastodon #Ireland #DPU #EU #GrokAI #Grok

NVIDIA’s new Inference Context Memory Storage Platform reshapes AI inference by treating KV cache as a multi-tier memory hierarchy—from HBM to NVMe SSD. This enables longer context windows, persistent reasoning, and scalable multi-agent inference while keeping hot data in GPU memory and offloading cold context to SSD.
https://www.buysellram.com/blog/nvidia-unveils-the-inference-context-memory-storage-platform/
#NVIDIA #Rubin #AI #Inference #LLM #AIInfrastructure #MemoryHierarchy #HBM #NVMe #DPU #BlueField4 #AIHardware #GPU #DRAM #KVCache #DataCenter #tech

NVIDIA’s new Inference Context Memory Storage Platform reshapes AI inference by treating KV cache as a multi-tier memory hierarchy—from HBM to NVMe SSD. This enables longer context windows, persistent reasoning, and scalable multi-agent inference while keeping hot data in GPU memory and offloading cold context to SSD.
https://www.buysellram.com/blog/nvidia-unveils-the-inference-context-memory-storage-platform/
#NVIDIA #Rubin #AI #Inference #LLM #AIInfrastructure #MemoryHierarchy #HBM #NVMe #DPU #BlueField4 #AIHardware #GPU #DRAM #KVCache #DataCenter #tech

NVIDIA’s new Inference Context Memory Storage Platform reshapes AI inference by treating KV cache as a multi-tier memory hierarchy—from HBM to NVMe SSD. This enables longer context windows, persistent reasoning, and scalable multi-agent inference while keeping hot data in GPU memory and offloading cold context to SSD.
https://www.buysellram.com/blog/nvidia-unveils-the-inference-context-memory-storage-platform/
#NVIDIA #Rubin #AI #Inference #LLM #AIInfrastructure #MemoryHierarchy #HBM #NVMe #DPU #BlueField4 #AIHardware #GPU #DRAM #KVCache #DataCenter #tech

NVIDIA’s Inference Context Memory Storage Platform, announced at CES 2026, marks a major shift in how AI inference is architected. Instead of forcing massive KV caches into limited GPU HBM, NVIDIA formalizes a hierarchical memory model that spans GPU HBM, CPU memory, cluster-level shared context, and persistent NVMe SSD storage.

This enables longer-context and multi-agent inference by keeping the most active KV data in HBM while offloading less frequently used context to NVMe—expanding capacity without sacrificing performance. This shift also has implications for AI infrastructure procurement and the secondary GPU/DRAM market, as demand moves toward higher bandwidth memory and context-centric architectures.

https://www.buysellram.com/blog/nvidia-unveils-the-inference-context-memory-storage-platform/

#NVIDIA #Rubin #AI #Inference #LLM #AIInfrastructure #MemoryHierarchy #HBM #NVMe #DPU #BlueField4 #AIHardware #GPU #DRAM #KVCache #LongContextAI #DataCenter #AIStorage #AICompute #AIEcosystem #tech

NVIDIA’s Inference Context Memory Storage Platform, announced at CES 2026, marks a major shift in how AI inference is architected. Instead of forcing massive KV caches into limited GPU HBM, NVIDIA formalizes a hierarchical memory model that spans GPU HBM, CPU memory, cluster-level shared context, and persistent NVMe SSD storage.

This enables longer-context and multi-agent inference by keeping the most active KV data in HBM while offloading less frequently used context to NVMe—expanding capacity without sacrificing performance. This shift also has implications for AI infrastructure procurement and the secondary GPU/DRAM market, as demand moves toward higher bandwidth memory and context-centric architectures.

https://www.buysellram.com/blog/nvidia-unveils-the-inference-context-memory-storage-platform/

#NVIDIA #Rubin #AI #Inference #LLM #AIInfrastructure #MemoryHierarchy #HBM #NVMe #DPU #BlueField4 #AIHardware #GPU #DRAM #KVCache #LongContextAI #DataCenter #AIStorage #AICompute #AIEcosystem #tech

NVIDIA’s Inference Context Memory Storage Platform, announced at CES 2026, marks a major shift in how AI inference is architected. Instead of forcing massive KV caches into limited GPU HBM, NVIDIA formalizes a hierarchical memory model that spans GPU HBM, CPU memory, cluster-level shared context, and persistent NVMe SSD storage.

This enables longer-context and multi-agent inference by keeping the most active KV data in HBM while offloading less frequently used context to NVMe—expanding capacity without sacrificing performance. This shift also has implications for AI infrastructure procurement and the secondary GPU/DRAM market, as demand moves toward higher bandwidth memory and context-centric architectures.

https://www.buysellram.com/blog/nvidia-unveils-the-inference-context-memory-storage-platform/

#NVIDIA #Rubin #AI #Inference #LLM #AIInfrastructure #MemoryHierarchy #HBM #NVMe #DPU #BlueField4 #AIHardware #GPU #DRAM #KVCache #LongContextAI #DataCenter #AIStorage #AICompute #AIEcosystem #tech

NVIDIA’s Inference Context Memory Storage Platform, announced at CES 2026, marks a major shift in how AI inference is architected. Instead of forcing massive KV caches into limited GPU HBM, NVIDIA formalizes a hierarchical memory model that spans GPU HBM, CPU memory, cluster-level shared context, and persistent NVMe SSD storage.

This enables longer-context and multi-agent inference by keeping the most active KV data in HBM while offloading less frequently used context to NVMe—expanding capacity without sacrificing performance. This shift also has implications for AI infrastructure procurement and the secondary GPU/DRAM market, as demand moves toward higher bandwidth memory and context-centric architectures.

https://www.buysellram.com/blog/nvidia-unveils-the-inference-context-memory-storage-platform/

#NVIDIA #Rubin #AI #Inference #LLM #AIInfrastructure #MemoryHierarchy #HBM #NVMe #DPU #BlueField4 #AIHardware #GPU #DRAM #KVCache #LongContextAI #DataCenter #AIStorage #AICompute #AIEcosystem #tech

NVIDIA’s Inference Context Memory Storage Platform, announced at CES 2026, marks a major shift in how AI inference is architected. Instead of forcing massive KV caches into limited GPU HBM, NVIDIA formalizes a hierarchical memory model that spans GPU HBM, CPU memory, cluster-level shared context, and persistent NVMe SSD storage.

This enables longer-context and multi-agent inference by keeping the most active KV data in HBM while offloading less frequently used context to NVMe—expanding capacity without sacrificing performance. This shift also has implications for AI infrastructure procurement and the secondary GPU/DRAM market, as demand moves toward higher bandwidth memory and context-centric architectures.

https://www.buysellram.com/blog/nvidia-unveils-the-inference-context-memory-storage-platform/

#NVIDIA #Rubin #AI #Inference #LLM #AIInfrastructure #MemoryHierarchy #HBM #NVMe #DPU #BlueField4 #AIHardware #GPU #DRAM #KVCache #LongContextAI #DataCenter #AIStorage #AICompute #AIEcosystem #tech

NVIDIA’s new Inference Context Memory Storage Platform reshapes AI inference by treating KV cache as a multi-tier memory hierarchy—from HBM to NVMe SSD. This enables longer context windows, persistent reasoning, and scalable multi-agent inference while keeping hot data in GPU memory and offloading cold context to SSD.
https://www.buysellram.com/blog/nvidia-unveils-the-inference-context-memory-storage-platform/
#NVIDIA #Rubin #AI #Inference #LLM #AIInfrastructure #MemoryHierarchy #HBM #NVMe #DPU #BlueField4 #AIHardware #GPU #DRAM #KVCache #DataCenter #tech

NVIDIA’s new Inference Context Memory Storage Platform reshapes AI inference by treating KV cache as a multi-tier memory hierarchy—from HBM to NVMe SSD. This enables longer context windows, persistent reasoning, and scalable multi-agent inference while keeping hot data in GPU memory and offloading cold context to SSD.
https://www.buysellram.com/blog/nvidia-unveils-the-inference-context-memory-storage-platform/
#NVIDIA #Rubin #AI #Inference #LLM #AIInfrastructure #MemoryHierarchy #HBM #NVMe #DPU #BlueField4 #AIHardware #GPU #DRAM #KVCache #DataCenter #tech

NVIDIA’s new Inference Context Memory Storage Platform reshapes AI inference by treating KV cache as a multi-tier memory hierarchy—from HBM to NVMe SSD. This enables longer context windows, persistent reasoning, and scalable multi-agent inference while keeping hot data in GPU memory and offloading cold context to SSD.
https://www.buysellram.com/blog/nvidia-unveils-the-inference-context-memory-storage-platform/
#NVIDIA #Rubin #AI #Inference #LLM #AIInfrastructure #MemoryHierarchy #HBM #NVMe #DPU #BlueField4 #AIHardware #GPU #DRAM #KVCache #DataCenter #tech

NVIDIA’s new Inference Context Memory Storage Platform reshapes AI inference by treating KV cache as a multi-tier memory hierarchy—from HBM to NVMe SSD. This enables longer context windows, persistent reasoning, and scalable multi-agent inference while keeping hot data in GPU memory and offloading cold context to SSD.
https://www.buysellram.com/blog/nvidia-unveils-the-inference-context-memory-storage-platform/
#NVIDIA #Rubin #AI #Inference #LLM #AIInfrastructure #MemoryHierarchy #HBM #NVMe #DPU #BlueField4 #AIHardware #GPU #DRAM #KVCache #DataCenter #tech

NVIDIA’s new Inference Context Memory Storage Platform reshapes AI inference by treating KV cache as a multi-tier memory hierarchy—from HBM to NVMe SSD. This enables longer context windows, persistent reasoning, and scalable multi-agent inference while keeping hot data in GPU memory and offloading cold context to SSD.
https://www.buysellram.com/blog/nvidia-unveils-the-inference-context-memory-storage-platform/
#NVIDIA #Rubin #AI #Inference #LLM #AIInfrastructure #MemoryHierarchy #HBM #NVMe #DPU #BlueField4 #AIHardware #GPU #DRAM #KVCache #DataCenter #tech

Oh my, `nl2doca`. "It takes the route and puts it into the ASIC" - turns out if you don't use the NVIDIA User Experience, it takes the route and punts it to software. #bluefield3 #dpu

🎁 Dev Box DPU Upgrade 🎁

More ZFS PoC for vGPU + DOCA blah blah NVMe-oF offloads

Left–to–Right (lotta slots)
- Broadcom CNA 2x10G
- Optane P3608 2x 2TB NVMe AIC
- Nvidia T10 16GB GPU (w/ 90mm blower)
- blank
- Nvidia T10 16GB GPU (w/ 90mm blower)
- Nvidia Bluefield2 DPU 16GB RAM, 2x 25G
- Optane 4x 512GB NVDIMM
- Samsung 4x 32GB ECC Volatile
- Intel Xeon 8370C (Azure) 32C/64T 270W

Chassis cooled by 2x 180mm, 1x 140mm, 1x 120mm. Idles very comfortably below 40dB, wind-tunnel at full rpm is more like whoooooshing than the usual 28K-RPM 40mm axial fan screamers people expect from enterprise hardware. Choose wisely.

#gpu #dpu #ai #dev #llm #nvidia #intel #optane #zfs #linux #freebsd

PG&E outage for all of Sunnyvale, UPS resources kept the home office and homelab online most of the hours. In the interim via LTE, decided to resume efforts on spec'ing the latest network refresh.

Big decisions; mostly around scale model design principles, some blandness re: OFED drivers and SPDK version parity for specific NIC/DPU SKUs, and analysis paralysis when focusing on critical nerdatorium protocol support.

Otherwise to say... I'm leaning towards additional Bluefield DPUs, where offload options are more important than bandwidth parity for production port speeds (also lower total port count using 4x25 breakouts), and there's no sense in paying more per unit for a CX6 25G vs BF2 25G... so... yep.

#networking #homelab #mellanox #nvidia #dpu #zfs #distributedsystems #ai #hpc

Die Zahlen sind irgendwie zu gut, um glaubwürdig zu sein: Ich schaffe auf einer #DPU mit nur einem Thread und mit fast gar nicht optimiertem #Code einfach 20 Gbps #iperf-Traffic mit #DPDK zu forwarden 🤔

[Перевод] Подробное руководство по облачной инфраструктуре для ИИ-проектов

Как вычислительные ресурсы, системы хранения, сетевые технологии и AI-фреймворки влияют на производительность AI-проектов. Почему понимание этой инфраструктуры даёт конкурентное преимущество

https://habr.com/ru/articles/896194/

#облачная_инфраструктура #искусственный_интеллект #gpu #tpu #dpu #хранение_данных #сетевые_технологии #nvlink #rdma #ai_фреймворки

The report from this summer's Dagstuhl seminar on Programmable Host Networking has been published: https://doi.org/10.4230/DagRep.14.7.35! #ebpf #DPDK #DPU

Je m'inscris au #camp d'été des #DPU ... la charte est inspirante je trouve alors je partage, pour d'autres organisateur-ices : https://dansesdelapaixuniverselle.fr/wp-content/uploads/2025/01/Charte-vie-du-camp-2025-Fr-OK.pdf

#DUP #Meeting #SummerMeeting #Universités d'été etc. #OrganisationCollective #IntelligenceCollective #mercredicollectif toujours (cf. les attentions concernant les enfants, presque toujours & partout bienvenu-es chez nous soufi-es...)

#Paix #Chant #Spiritualité et #Danse surtout bien sûr einh

#AMD Moves Up Instinct #MI350/#MI355X To Now Release By Mid-2025, #MI400 Lineup Slated For 2026 As #Datacenter Biz Hits Records
Last year the datacenter business at AMD, which includes #EPYC #CPU, Instinct #GPU, #Pensando #DPU, and #Xilinx #FPGA accelerators, accounted for $12.58 billion in sales, or 48.8% of revenues. In both Q3 and Q4 of 2024, datacenter drove in excess of half of AMD’s revenues and nearly 60% of its operating income.
https://www.nextplatform.com/2025/02/04/amd-moves-up-instinct-355x-launch-as-datacenter-biz-hits-records/

#AMD Moves Up Instinct #MI350/#MI355X To Now Release By Mid-2025, #MI400 Lineup Slated For 2026 As #Datacenter Biz Hits Records
Last year the datacenter business at AMD, which includes #EPYC #CPU, Instinct #GPU, #Pensando #DPU, and #Xilinx #FPGA accelerators, accounted for $12.58 billion in sales, or 48.8% of revenues. In both Q3 and Q4 of 2024, datacenter drove in excess of half of AMD’s revenues and nearly 60% of its operating income.
https://www.nextplatform.com/2025/02/04/amd-moves-up-instinct-355x-launch-as-datacenter-biz-hits-records/

#AMD Moves Up Instinct #MI350/#MI355X To Now Release By Mid-2025, #MI400 Lineup Slated For 2026 As #Datacenter Biz Hits Records
Last year the datacenter business at AMD, which includes #EPYC #CPU, Instinct #GPU, #Pensando #DPU, and #Xilinx #FPGA accelerators, accounted for $12.58 billion in sales, or 48.8% of revenues. In both Q3 and Q4 of 2024, datacenter drove in excess of half of AMD’s revenues and nearly 60% of its operating income.
https://www.nextplatform.com/2025/02/04/amd-moves-up-instinct-355x-launch-as-datacenter-biz-hits-records/

#AMD Moves Up Instinct #MI350/#MI355X To Now Release By Mid-2025, #MI400 Lineup Slated For 2026 As #Datacenter Biz Hits Records
Last year the datacenter business at AMD, which includes #EPYC #CPU, Instinct #GPU, #Pensando #DPU, and #Xilinx #FPGA accelerators, accounted for $12.58 billion in sales, or 48.8% of revenues. In both Q3 and Q4 of 2024, datacenter drove in excess of half of AMD’s revenues and nearly 60% of its operating income.
https://www.nextplatform.com/2025/02/04/amd-moves-up-instinct-355x-launch-as-datacenter-biz-hits-records/

#AMD Moves Up Instinct #MI350/#MI355X To Now Release By Mid-2025, #MI400 Lineup Slated For 2026 As #Datacenter Biz Hits Records
Last year the datacenter business at AMD, which includes #EPYC #CPU, Instinct #GPU, #Pensando #DPU, and #Xilinx #FPGA accelerators, accounted for $12.58 billion in sales, or 48.8% of revenues. In both Q3 and Q4 of 2024, datacenter drove in excess of half of AMD’s revenues and nearly 60% of its operating income.
https://www.nextplatform.com/2025/02/04/amd-moves-up-instinct-355x-launch-as-datacenter-biz-hits-records/

"
Kleinsatelliten: Leistungsfähiges Datenmanagement im Orbit
"
"Erdbeobachtungssatelliten sammeln große Datenmengen, während die Übertragungskapazitäten zur Erde begrenzt sind. Spezielle Computer, sogenannte Data Processing Units (DPUs), speichern, werten und komprimieren diese Daten direkt im Satelliten – effizient und zuverlässig."

https://www.emi.fraunhofer.de/de/aktuelles/aktuelles-presse/kleinsatelliten--leistungsstarkes-datenmanagement-im-orbit.html

16.1.2025

#Computer #constellr #COTS #Erdbeobachtung #EO #EMI #Fraunhofer #Daten #DPU #Raumfahrt #Satelliten #SkyBee1 #SpaceFlight

Применение аппаратных ускорителей в SDN: как мы добились скорости передачи пакетов на уровне Bare Metal

Каждый, кто имеет дело с более-менее крупной IT-инфраструктурой, знает, что в мире сетей есть две прямо противоположных реальности. В одной — виртуализация с эффективным дроблением ресурсов, но потерей скоростей, в другой — Bare Metal с высокой скоростью и мощностью, но слабой гибкостью в вопросах выделения ресурсов. И если вы уже задались вопросом: «А можно без крайностей?», я инженер R&D-команды Узнать больше

https://habr.com/ru/companies/cloud_ru/articles/861434/

#виртуализация #bare_metal #rnd #openstack #asap #nvidia #ускоряем_сеть #dpu

Microsoft unveils the Azure Boost DPU, its first data processing unit, and the Azure Integrated Hardware Security Module, a new in-house security chip! 🔒⚡ Big strides in cloud performance and security. #Microsoft #Azure #DPU #CloudComputing #TechNews #Cybersecurity #Innovation

https://medium.com/@samlamucf/high-performance-data-center-storage-solutions-with-dpu-integration-c6e639893531
#DataCenter #GPU #IOBottleneck #Cloud #AIComputing #DPU

Widescreen Wonder: #LasVegasSphere
54,000 m2 (~3.67 acre) interior LED display (16x16K) and an exterior LED display (‘Exosphere’) consisting out of 1.23 million LED ‘pucks’. Driving all these pixels are around 150 #NVidia RTX #A6000 #GPU, installed in computer systems which are networked using NVidia #BlueField data processing units (#DPU) and NVidia #ConnectX6 NICs (up to 400 Gb/s), with visual content transferred from Sphere Studios in Cali. All this hardware uses 45kW.
https://blogs.nvidia.com/blog/sphere-las-vegas/

💡 AMD collaborerà con Samsung per la produzione di chip a 3nm GAAFET

https://gomoot.com/amd-collaborera-con-samsung-per-la-produzione-di-chip-a-3nm-gaafet

#AI #amd #blog #cpu #dpu #GAAFET #gpu #news #picks #Samsung #tech #tecnologia #TMSC

There's going to be a Dagstuhl Seminar on Programmable Host Networking this summer: https://www.dagstuhl.de/24291.
I'm looking forward to it! #ebpf #dpdk #DPU

Vi har #danskfagetsdag på #dpu. Marte Blikstad-Ballas holder en veloplagt keynote. Jo mere it i undervisningen, deres mindre interaktion... Fordi tiden går på PowerPoint og individuelt arbejde på "læringsbræt".

#skolechat

"Die Danube Private University (DPU) in Krems, Niederösterreich, freut sich über ein nächstes spannendes Projekt, denn die Europäische Weltraumbehörde ESA hat eine Machbarkeitsstudie für ein innovatives Luftreinigungssystem (CLAIS – Clean Air in Space) für Weltraumstationen beauftragt. Eine Presseaussendung der DPU."

https://www.raumfahrer.net/weltraumbehoerde-beauftragt-dpu-clean-air-in-space/

#ACLS #Airbus #CLAIS #DPU #ESA #Friedrichshafen #ISS #Krems #Luftreinigungssystem #Niederösterreich #Österreich #Raumfahrt #Raumstation

15.3.2023

@sfoskett Not until they’re licensed as part of the system, just like a standard NIC or HBA. When you have to independently license whatever is running on the #DPU, it’s still its own thing.

With most #DPU vendors now acquired and integrated into server platforms, is it safe to say that processing power on network adapters is a system feature not an independent product?

Microsoft buys Fungible, Inc. Cool tech, but the "Composable Infrastructure" market never developed. https://www.techradar.com/news/microsoft-snaps-up-fungible-for-azure-cloud-and-data-center-boost #Fungible #DPU #Composable

#Microsoft boosts future #Azure networking and storage smarts
@[email protected] has announced the acquisition of #composableinfrastructure and digital processing unit (#DPU, aka #SmartNIC) vendor #Fungible.
https://bit.ly/3vOqFVY

#skolechat kom til formidlingskonference om tværvidenskabeligt og banebrydende forskningsprojekt om tidlig skrivning.

ATEL er nok det vildeste forskningsprojekt jeg har været med til. Det er lykkes os på empirisk grundlag at beskrive indskolingselevers typiske skriveudvikling på en række dimensioner OG at automatisere analysen, så vi kan give lærere viden om hvor i processen deres elever befinder sig, så vi kan give forslag til hvad de skal undervise i.
#DTU #DPU #NVL
https://events.au.dk/atel

What better #introduction to #fosstodon than a look at what I think was my first #foss contribution. I’d like to apologize now for the unbounded memory allocation bug I introduced. Oops.

https://github.com/php/php-src/commit/72f675be5005f09c048155b85b5d6efd7341f6e0

Since then I’ve done a bunch of #sysadmin work (mostly #Solaris) followed by OS (#solaris #illumos #smartos #zfs) development and more recently various storage stuff (#nvme #nvmeof #roce #spdk #bluefield #dpu #smb).

I love #hiking through the #wilderness and other natural areas.

#AMD Makes a #Big #DPU Move with $1.9B Bid for #Pensando

https://www.nextplatform.com/2022/04/04/amd-makes-a-big-dpu-move-with-1-9-billion-bid-for-pensando/