Quantum computing: from cloud to on-prem infrastructure

Meet Matt Pryor, Senior Cloud Native Platform Engineer at Nscale. Together with Nick Jones, Matt will present “The Fast and the Flexible: bare-metal performance for HPC and AI with the flexibility of cloud.” High-performance computing and AI workloads demand bare-metal speed—but also the flexibility and scalability of the cloud. Traditionally, virtualization created a performance trade-off. In this talk, Matt and Nick will show how Nscale bridges that gap with open-source technologies: ⚡ High-performance virtual Kubernetes clusters provisioned in minutes ⚡ Direct hardware access for maximum speed ⚡ A familiar Slurm interface layered with Slinky from SchedMD ⚡ Full-stack benchmarking built into the hardware lifecycle This is a must-see for anyone working with HPC, AI, or performance-sensitive workloads who wants the best of both worlds: bare-metal power with cloud-native flexibility.

TechRadar writes "Google's most powerful supercomputer ever has a combined memory of 1.77PB - apparently a new world record for shared memory multi-CPU setups - Ironwood TPU is already deployed in Google Cloud data centers" https://lnkd.in/ecxFNX2J. #google #tpu #ironwoodtpu #googledatacenters #artificialintelligence #aiaccelerators #techradar #worldrecord

AWS has launched its 8th-generation memory-optimized EC2 instances—R8i and R8i‑Flex—powered by custom Intel Xeon 6 processors, now generally available on the AWS cloud platform. These new instances deliver up to 20% better price-performance than their predecessors and 2.5× more memory throughput, along with significantly larger cache capacity. They support DDR5 7200 MT/s memory and can scale from 2 to 384 vCPUs, achieving up to 3.9 GHz all-core turbo frequency. Ideal for data- and memory-intensive workloads—such as analytics, in-memory caching, and enterprise databases—these enhancements reflect Intel’s continued relevance and AWS’s focus on hardware-optimized cloud infrastructure. #AWS #Intel #CloudComputing #AI #TechNews #TemokTechnologies

AWS Chooses Intel Again: Powering the Next Generation of Cloud Infrastructure Great news for the cloud computing world! AWS has announced its new eighth-generation memory-optimized EC2 instances, the R8i and R8i-flex, and they are powered by custom-built Intel Xeon 6 processors. This is a big win for Intel, demonstrating their continued relevance and strong partnership with the world's largest cloud provider. What does this mean for developers and businesses? These new instances are a game-changer for memory-intensive workloads. They offer: * Up to 20% more computing power than the previous generation. * 2.5 times higher memory throughput. * 15% better price-performance. This makes the R8i instances an ideal solution for critical applications like SAP HANA, SQL and NoSQL databases, and big data analytics with Apache Hadoop and Spark. This collaboration highlights a continued commitment to innovation, providing customers with cutting-edge performance for their most demanding workloads. It's exciting to see how this partnership will continue to shape the future of the cloud. #AWS #Intel #CloudComputing #Infrastructure #TechNews #DataAnalytics #AI

VCF 9 Gets MCP Support, GPU Optimization at No Extra Cost. Broadcom expands its cloud platform with AI capabilities, MCP support and enhanced security features — all at no additional cost to existing customers.

As big data, cloud, and AI evolve, reliable network performance is now a mission-critical factor for business and data center success. The FS PicOS® data center switch, validated through Ixia RFC 2544 testing, delivers low latency, zero packet loss, and 98% line-rate forwarding, ensuring exceptional performance and reliability in data center networks under high-demand scenarios. 🔗 See how FS verifies the reliability of the data center network: https://lnkd.in/gNPuGhPH #LowLatency #HighThroughput #DataCenter #IxiaRFC2544 #switch #TechSolutions #Innovation #QSFP28

The race to build a distributed GPU runtime is heating up. 🚀 As Mike Beaumont highlights in our recent blog post, the bottleneck at datacenter scale isn’t FLOPS, it’s data movement. That’s why we built Theseus with a data-movement first design, overlapping compute with I/O and prefetching exact byte ranges. In head-to-head, cost-matched cloud tests, Theseus outperforms Photon by up to 4X and completes 100TB TPC-H/DS with just two DGX A100 640 GB nodes. And it now runs on AMD via ROCm-DS/hipDF, bringing cross-platform performance into the conversation. Read more in our latest blog: https://lnkd.in/gRPGb-U7 #DistributedSystems #GPUs #DataEngineering #ApacheArrow #SQL #AI

Moving data at scale is the toughest challenge in accelerated computing. Nvidia is trying to solve. AMD (w/ROCm) is also. Voltron Data's Theseus... ... is architected for data movement. Theseus workers run four specialized, asynchronous executors: Compute, Memory, Pre-Load, and Network. This architecture allows I/O, spill/prefetch, and shuffle to happen in parallel with GPU compute. We prioritize data movement: - we guarantee places for data to live - we prefetch the exact bytes - we make host memory a fast transit lane - our network executor supports TCP and UCX/GPUDirect RDMA, with optional compression Voltron Data's Theseus has solved the data movement problem. [Link to Blog in Comments]

Mike Beaumont drops big time knowledge on the race to build a distributed GPU runtime. It's well worth the read. Voltron Data.

I had an interesting exchange with Gemini recently. It told me I was looking at the value of Dell or HPE’s “AI Factory” all wrong — that it wasn’t about cost control, but speed to velocity. But here’s the problem: speed to velocity doesn’t come from infrastructure alone. You can have racks full of GPUs ready to go, but if you don’t have the abstractions — the platform layer — you're still moving slow. This is the same mistake we made with private cloud. Just because you’ve virtualized compute doesn’t mean you’ve built a cloud. And just because you’ve installed GPUs doesn’t mean you’ve built an AI platform. If I need a prototype in 3 weeks, I’m not building pipelines from scratch or dealing with YAML and ticket queues. I’m going to Vertex AI, Bedrock, or even Hugging Face — because that’s where the abstraction exists to move fast. Infrastructure is necessary. But it's not sufficient. Speed lives at the service layer.