Astera Labs’ Post

I am proud of our partnership with Astera Labs on this industry-first PCIe Gen6 interoperability demo, which will be showcased next week at DesignCon in Santa Clara. This marks the first public demonstration of end-to-end interoperability between the Astera Labs Scorpio PCIe Gen6 Smart Fabric Switch and Micron’s PCIe Gen6 data center SSDs. The industry transition to PCIe Gen6 is going to be a pivotal advancement for enabling the data-intensive AI workloads of the future. Thank you Michael Ocampo, Mike Hendricks, Lori Zielinski, Paroma Sen and the entire Astera Labs team for your outstanding partnership! The Micron PCIe Gen6 data center SSD used in the demo, currently available for ecosystem development, already achieves 27 GB/s, likely making this the fastest demonstration of a Gen6 storage drive to date. 😊

"Redefine Performance, Power Big Data and AI – Meet the R2520G6!“   Built to meet the demands of modern enterprises, this 2U dual-socket server supports #Intel #Xeon 6 CPUs, delivering cutting-edge performance.   It excels in #AI, cloud computing, media production, big data analytics, virtualization, hyper-converged infrastructure, and disaster recovery.   With 24 NVMe U.2 hot-swap bays, 32 DDR5 DIMMs, and PCIe 5.0 expandability, the R2520G6 ensures unmatched scalability and reliability.   Product Information👉https://pse.is/R2520G6   #ServerInnovation #DataAnalytics #EnterpriseStorage #MiTACComputing

This table is from the SMERF paper below. Our method also incurs a non-trivial training cost: in addition to training the teacher, we optimize our method for 100,000-200,000 steps on 8x V100 or 16x A100 GPUs, depending on dataset. While our method achieves higher quality than 3DGS on average, it is not universally higher in detail for all parts of all scenes. We attribute this to the voxel structure imposed on the scene by our representation. SMERF: Streamable Memory Efficient Radiance Fields for Real-Time Large-Scene Exploration https://lnkd.in/e4GkW3yT After you run this notebook: https://lnkd.in/exVP3znY You can run another script for 100,000-200,000 steps on 8x V100 or 16x A100 GPUs. SMERF bakes ZipNeRF checkpoints in jax. You get a viewer that lets you interact with what you captured. When using the notebook for NeRF in Cloud, please remember to increase your machine quota to use 8 V100 GPU AND 64 CPU.

The S8056 is a rack-optimized single-socket server motherboard, engineered for the latest #AMD #EPYC™ 9005 processor. Featuring 24 DDR5-5200 DIMM slots supporting up to 2TB memory, PCIe Gen 5 expandability, and advanced NVMe storage options, it delivers exceptional performance and scalability. Designed for #virtualization, #AI, #HPC, and cloud computing, the S8056 ensures fast processing, high memory bandwidth, and robust storage performance. Its efficient architecture addresses demanding workloads, making it an ideal choice for modern data center operations. Product Information 👉 https://pse.is/S8056MB #EPYC9005 #ServerTechnology #AIWorkloads #CloudComputing #MiTACComputing

Cloud Run fully managed serverless platform on GCP integrates Nvidia L4 GPUs, enabling organizations to run serverless inference. It has been a popular platform with developers & its ability to simplify container deployment and management. However, the greater demands of AI workloads, particularly those requiring real-time processing, have highlighted the need for more robust computational resources. The integration of GPU support opens up a wide array of use cases for Cloud Run developers.

“Exciting news! Google Cloud Run has just announced support for NVIDIA L4 GPUs, enhancing the capability to run AI inference workloads efficiently and at scale. This integration allows developers to deploy real-time AI applications without the hassle of managing infrastructure. With NVIDIA's powerful L4 GPUs, users can expect significantly improved performance for various AI tasks, including generative AI and real-time inference. This advancement not only simplifies deployment but also optimizes costs with a pay-per-use model. Experience the future of serverless AI with Cloud Run today! “ #NVIDIA #GoogleCloud #CloudRun #L4GPU

See the industry’s first #PCIe over optics technology demonstration for GPU clusters across the data center in action! Richard Ward walks through this future-proof demo showing PCIe running at full rate over single mode fiber for an aggregate bandwidth of 128GB/s. Hyperscalers will now be able to use PCIe to connect #GPU clusters in row-to-row topologies for #AI and #cloud infrastructure. Watch now: https://buff.ly/4emngm2 #PCISIGDevCon24

#AI deployment involves many considerations to ensure optimal value, including infrastructure modernization. NVIDIA #HPEProLiant Compute DL380a Gen12 delivers ultra-scalable GPU acceleration for best-in-industry fine-tuning and inference performance.

#AI deployment involves many considerations to ensure optimal value, including infrastructure modernization. NVIDIA #HPEProLiant Compute DL380a Gen12 delivers ultra-scalable GPU acceleration for best-in-industry fine-tuning and inference performance.