NETINT Technologies Inc.’s Post

"Practically linear growth" (scaling) with 128-core #Ampere Altra Max CPUs, for video encoding, according to NETINT Technologies Inc. This is one of the benefits of #arm architecture compared to x86. #neoverse

The tech world is rapidly evolving, and one of the most significant shifts is the adoption of ARM architecture, renowned for its efficiency and scalability. This evolution is particularly impactful for professionals in streaming, broadcasting, and CDN infrastructure. If you're in these industries, understanding the benefits of ARM could be pivotal for your technology strategy. G&L Geißendörfer & Leschinsky GmbH initial interest in ARM technology was sparked by the impressive performance of Qualcomm-based ARM hardware encoders, which soon expanded as we noted industry giants like AWS adopting ARM-based setups such as Graviton. The key advantage driving this shift is the significantly lower power consumption that ARM offers, a benefit that's been crucial in the mobile sector and equally valuable in streaming technology. The efficiency of ARM architecture not only makes it ideal for energy-intensive operations but also facilitates considerable cost savings and enhances sustainability. These insights highlight why transitioning to ARM is becoming a strategic move for those looking to optimize technology operations in today's energy-conscious environment. ARM processors, especially those with higher core counts such as the 128-core machines from Ampere, exhibit nearly linear performance scaling, unlike traditional processors like AMD EPIC which often struggle in this area. This linear scalability is crucial for applications that demand reliable live transcoding, where predictable performance scaling is necessary. The key benefits of this scalability include predictable performance as the workload increases and more efficient resource utilization, which helps avoid the performance bottlenecks often seen in systems that do not scale linearly. In G&L Geißendörfer & Leschinsky GmbH's CDN infrastructure developed in Germany, the switch to ARM processors has demonstrated clear advantages. This is particularly true for handling specific streaming protocols not supported by standard servers, such as ancient Icecast streaming or MPEG-TS over TCP, where ARM's efficiency in managing diverse workloads is highly beneficial. For those considering a similar transition, it is advisable to first assess the compatibility of your current infrastructure with ARM technology and then plan a phased integration. This approach allows for monitoring performance improvements and managing transition risks effectively. Adopting ARM architecture represents a strategic decision. Staying informed about the latest developments in ARM technology is crucial to ensure your infrastructure remains competitive and cost-effective. 🌍💡 Watch this insightful video: https://netint.biz/4ac6aV1 #VPU #streaming #video #Encoding #ARM Ampere

"AMD, Broadcom, Cisco, #Google, HPE, #Intel, #Meta, and #Microsoft have joined forces to develop Ultra Accelerator Link (UALink), a new industry standard to enable high-speed, low-latency interconnection for datacenter grade AI and HPC accelerators" as competition to #NVIDIA's NVLink. While it is another example of #AI being at least good to push hardware development, maybe we need to step back and look at what our priorities are? It used to be to create new content/information, understand and abstract it (reduce volume for exchange).... now we compete in the fastest way to transfer raw information between nodes (without understanding it?). https://lnkd.in/e9pc5CHT?

We have 3 tenets at Hedgehog: 1. AI needs a new network 2. Ethernet wins (again) 3. Network disaggregation happens (finally) NVIDIA has been proving the first tenet for a while now. Looks like this development confirms the second. Writing is on the wall for the third. #hedgehog #ainetwork #cloudnetwork

Last week we had a 🌟 LOT 🌟 of announcements coming off Cisco Live Amsterdam. A couple big announcements were centered around helping our customers with their own Generative AI adoption so if you missed it here were the big announcements last week: 1. We announced our new SAAS offering for GenAI deployments, Motific 🎇 . This is a game changer if you are looking to accelerate your GenAI deployment and add security controls, manage for shadow AI in your organization, control usage for budgeting purposes, and so so much more. Even better, its VENDOR AGNOSTIC. Link to Announcement: https://lnkd.in/gFCze5ec 2. Cisco and Nvidia's Partnership Announcement 👁 . Not only do we have the newest Nvidia Tensor Core GPUs available in Cisco’s M7 generation of UCS rack and blade servers, BUT this announcement showed Nvidia's partnership with Cisco to deploy AI on our Ethernet based (thats right InfiniBand) networking solutions. We believe Ethernet is the future of GenAI, if you want to know more read up here: https://lnkd.in/gFCze5ec Excited for our customers and for the future we get to build together! Reach out if you want to talk more about what we are doing with GenAI.

Another great article at Next Platform, discussing #AI needs from a networking perspective. https://lnkd.in/dgVz6gMr NVIDIA's InfiniBand has been the default (bundled) choice with Ampere (A100) and Hopper (H100) chips; however, Ethernet has a few tricks up its sleeve 🤹 With first-gen products from the Ultra Ethernet Consortium supposedly arriving this year, and its already mid-May, is concerning 😰 #arista #juniper #cisco #sonic

Nvidia’s recent vision on RAN & AI is really thought provoking: Infrastructure Utilization: Currently average 5G gNB utilization is less than 30%, 70% of the resources are either not at all used or used very rarely. One may argue that this is current trend and may change in future. Yes!! for such scenarios resource pooling is the best strategy. That’s where we used to have RNCs & BSCs in 2G/3G. Resource pooling  is again pointing towards centralised infra i.e. CRAN/VRAN and openness. #vRAN #cRAN #ORAN Infrastructure sharing: With the advent of AI, we need more EDGE sites, parallel processing. Can we leverage this infrastructure for edge AI workloads? Can one save on edge data centre by putting some of the workloads on existing infra? Answer is yes!  if infrastructure is generic and can host any type of workload. #RAN #AI Future readiness: Most important aspect of choosing an infrastructure is if its future ready? Can it host upcoming 6G and AI together? Can Telco monetize this infra or provide IaaS (Infrastructure as a service). Yes!! it could be. 

𝐓𝐫𝐚𝐧𝐬𝐚𝐜𝐭 𝐒𝐦𝐚𝐫𝐭𝐞𝐫 𝐰𝐢𝐭𝐡 𝐄𝐪𝐮𝐢𝐧𝐢𝐱 𝐚𝐧𝐝 𝐍𝐕𝐈𝐃𝐈𝐀 Every presentation about AI that I have attended always talks about the first adopters being #FinancialServices and Health Care. That's why I'm thrilled to hear about the joint partnership announced today between Equinix and NVIDIA to offer managed DGX pods. With Equinix's liquid cooling capabilities and Nvidia's prioritization of supply GPUs to this offering, the problems of on-prem data centers often lacking the power and cooling necessary for the GPUs, and the challenge of getting access to GPUs, are solved. We've already been having a number of conversations with financial service customers who have discovered that running private AI infrastructure has its challenges. Check out the details of this exciting partnership here: https://lnkd.in/g28XQXTZ #AI #FinancialServices #Equinix #NVIDIA #DGXpods #AIInfrastructure #WhereTheWorldTransacts #TransactSmarter

Should InfiniBand still be the default option for network infrastructure? https://hubs.la/Q02Pj7pY0 Most enterprises are accustomed to using Nvidia solutions, including Nvidia GPUs and InfiniBand connectivity, for their excellent performance compared to standard Ethernet Clos. However, these solutions can be: - expensive - require specialized skills - lead to vendor lock-in Shai Haim explores how DriveNets Network Cloud-AI addresses the networking challenges and key requirements faced by enterprises building AI workloads.