Trending

#rdma

Latest posts tagged with #rdma on Bluesky

Latest Top
Trending

Posts tagged #rdma

#rdma is getting some love because all is currently AI. I suspect it could also be a major win for database engineering. See also #genomics where we seem to need 128GiB of ram per core for some reason.

1 0 0 0
Preview
Intellectual Highwayが新たに100Gbps RDMA IPを投入!低遅延でストリーム転送を実現 Intellectual Highwayが提供する新しい100Gbps RDMA IPは、低遅延なデータ転送を可能にし、AIや映像分野への応用が期待されています。

Intellectual Highwayが新たに100Gbps RDMA IPを投入!低遅延でストリーム転送を実現 #神奈川県 #横浜市 #AI #Intellectual_Highway #RDMA

Intellectual Highwayが提供する新しい100Gbps RDMA IPは、低遅延なデータ転送を可能にし、AIや映像分野への応用が期待されています。

0 0 0 0
Post image

Apple's Thunderbolt 5 and RDMA in macOS Tahoe 26.2 are revolutionizing AI research, enabling efficient Mac clustering for massive models. #AI #MachineLearning #Apple #Thunderbolt5 #RDMA Link: thedailytechfeed.com/apples-thund...

1 0 0 0
Preview
1,5 To de VRAM sur un Mac Studio - Le RDMA Thunderbolt 5 qui change la donne Vous rêvez de faire tourner des modèles d'IA de 600 milliards de paramètres sur votre bureau sans avoir à vendre vos enfants ? Hé bien Jeff Geerling vient de tester un truc qui va vous faire baver, je pense. En tout cas, moi ça m'énerve (dans le bon sens du terme hein...) ! Apple lui a prêté 4 Mac Studios M3 Ultra pour tester une nouvelle fonctionnalité qui débarque avec macOS 26.2 et qui s'appelle le **RDMA over Thunderbolt 5**. En gros, c'est une techno qui permet à plusieurs Macs de partager leur mémoire unifiée comme si c'était un seul gros pool de RAM et du coup, au lieu d'avoir 4 machines séparées avec chacune leur mémoire, vous vous retrouvez avec 1,5 To de VRAM partagée accessible par toutes les machines.
1 1 0 0

Overview: Hacker News discussed RDMA over Thunderbolt 5 on Mac Studios for AI/ML. Users explored its potential, limitations, and compared it to alternatives, speculating on Apple's server strategy & cost-effectiveness for compute-intensive tasks. #RDMA 1/6

0 0 1 0

🌐macOS 26.2ではRDMA over Thunderboltによる高速AIクラスターが実現
developer.apple.com/documentation/macos-rele...
via #HackerNews

0 0 0 0
Post image

Just saw how RDMA slashes CPU load on S3‑compatible object storage, letting NVIDIA GPUs focus on AI inference. Less network I/O, more speed—game changer for ML pipelines. Dive into the benchmarks! #RDMA #S3Storage #AIInference

🔗 aidailypost.com/news/rdma-cu...

0 0 0 0
Preview
RDMA Point-to-Point Communication for LLM Systems Emerging Large Language Model (LLM) system patterns, such as disaggregated inference, Mixture-of-Experts (MoE) routing, and asynchronous reinforcement fine-tuning, require flexible point-to-point c…

RDMA Point-to-Point Communication for LLM Systems

#CUDA #RDMA #LLM #Package

hgpu.org?p=30339

0 0 0 0

📰 Broadcom Perkenalkan Thor Ultra — NIC Ethernet AI 800G Pertama di Dunia

👉 Baca artikel lengkap di sini: ahmandonk.com/2025/10/16/broadcom-thor...

#800g #ethernet #ai #networking #broadcom #jericho #4 #nic #rdma #thor #ultra #tomahawk #6 #uec

0 0 0 0
Reversible Offloading Boosts RDMA Write Performance

Reversible Offloading Boosts RDMA Write Performance

Researchers propose “unloading,” a reversible offload that shifts part of RDMA write processing back to the CPU, yielding up to 31% performance gains over full offload. Read more: getnews.me/reversible-offloading-bo... #rdma #offloading #datacenter

0 0 0 0
λScale Boosts Serverless LLM Inference with Fast Model Scaling

λScale Boosts Serverless LLM Inference with Fast Model Scaling

λScale speeds up serverless LLM inference by using RDMA multicast and an execute‑while‑load strategy, delivering up to 5× lower tail latency and about 31.3% cost reduction. Read more: getnews.me/lscale-boosts-serverless... #serverless #llm #rdma

0 0 0 0
A decorated farewell party wall for USAID’s Regional Development Mission for Asia (RDMA). String lights and rows of photos form a glowing backdrop. Silver balloons spell “FOREVER” with red and blue hearts on each side, and below, a balloon spells “love.”

A decorated farewell party wall for USAID’s Regional Development Mission for Asia (RDMA). String lights and rows of photos form a glowing backdrop. Silver balloons spell “FOREVER” with red and blue hearts on each side, and below, a balloon spells “love.”

A month since USAID, they reflect on lessons beyond titles. At the RDMA farewell, they became the “Art Director,” turning limited resources into joy. Proof that creativity, planning, and teamwork endure—skills can travel further than roles. #USAIDForever #RDMA

16 3 0 0

Von InfiniBand-Performance zu Ethernet-Kompatibilität: Ultra Ethernet revolutioniert Rechenzentren. Schritt-für-Schritt vom Konzept zur Produktivumgebung.

admindocs.de/grundlagen/n...

#UltraEthernet #RDMA #RoCEv2 #ConnectX #LinuxNetworking #HPC #AI_ML #Datacenter

0 0 0 0
Preview
Simplifying InfiniBand on AKS Learn the what, the whys, and the hows of configuring InfiniBand networking for high performance compute (HPC) workloads on AKS

Need to run really large LLMs that require the power of RDMA with InfiniBand?
Check out this blog to see how to get it working on Azure Kubernetes Service:
azure.github.io/AKS/2025/04/...

#AKS #Kubernetes #InfiniBand #AI #LLM #RDMA #Azure

3 3 0 0
Preview
Simplifying InfiniBand on AKS Learn the what, the whys, and the hows of configuring InfiniBand networking for high performance compute (HPC) workloads on AKS

Interested in running workloads on #AKS that require #RDMA over #Infiniband? 🤔

Take a read through the latest AKS blog post on the topic and the linked guide which will walk you through installing #Nvidia's #NetworkOperator and #GPUOperator

azure.github.io/AKS/2025/04/...

0 0 0 0

Very first details on Ultra Ethernet, via a pull request. Software mock-up over UDP, for this in-progress new #rdma over Ethernet effort.
dub.sh/ultra-ethern...

0 0 0 0
Preview
GitHub - deepseek-ai/3FS: A high-performance distributed file system designed to address the challenges of AI training and inference workloads. A high-performance distributed file system designed to address the challenges of AI training and inference workloads. - GitHub - deepseek-ai/3FS: A high-performance distributed file system design...

I've waited a long time & no real hope this will change, but maybe maybe if really good fast #RDMA projects emerge maybe RDMA will start showing up in more hardware. Ultra Ethernet Consortium, where are you?

Such deep respect for DeepSeek being so so attentive cross-stack
github.com/deepseek-ai/...

1 0 0 0
Preview
GitHub - deepseek-ai/3FS: A high-performance distributed file system designed to address the challenges of AI training and inference workloads. A high-performance distributed file system designed to address the challenges of AI training and inference workloads. - GitHub - deepseek-ai/3FS: A high-performance distributed file system design...

#DeepSeek just dropped a file system for distributed #AI training over SSDs
on #RDMA networks! 🧠 Check it out here:
buff.ly/3EZzh3S

Does their system support clairvoyant prefetching 🎇 ?

3 0 0 1

What adapters are those? Good luck! Report back! Neat project! #rdma

1 0 1 0
Preview
Introduce Valkey Over RDMA protocol by pizhenwei · Pull Request #123 · valkey-io/valkey-doc RDMA is the abbreviation of remote direct memory access. It is a technology that enables computers in a network to exchange data in the main memory without involving the processor, cache, or operat...

#valkey (le fork de Redis) explore le #rdma et leurs tests annoncent un gain de x2.5 en QPS, et de meilleures latences.

github.com/valkey-io...

TCP est un peu un boulet, cf mon billet http://blog.garambro

0 0 0 0

Tweet to vote for #BTSArmy for #SquadGoals! @radiodisney #RDMA @BTS_twt

0 0 0 0

I'm starting to think that Britney and Jamie Lynn have the same surgeon. They are looking straight up like twins. #RDMA

0 0 0 0