Kalray’s newest acceleration boards to provide big leap forward for cloud-computing and networking and storage applications
Uniquely combines the necessary performance, power consumption and programmability for new requirements such as virtualization protocols and deep learning
Montbonnot, France, June 22, 2015 — Kalray, the inventor of supercomputing on a chip™, today announced two new cloud acceleration boards: TurboCard3™ for computing acceleration; and K-ONIC™ for networking and storage applications, which are programmable in C/C++. The boards are based on Bostan, Kalray’s second generation MPPA2® (massive parallel processing array) 256 cores processor, which brings a huge boost in FPU performance and computational power efficiency over the previous generation, and integrates 128 crypto accelerator cores and a smart load balancer between the Ethernet interface and internal cores further strengthening its position and offerings for the cloud.
TurboCard3, built on four MPPA2® 256 processors, is Kalray’s newest PCIe card. It accelerates workload and kernels that do not typically perform well on traditional host processors such as x86 or ARM. It also improves the 1 TFLOPs SP performance of its predecessor, TurboCard2 by a factor 3. For critical power saving, TurboCard3 consumes three times less power than a GPU board. All this gives TurboCard3 a competitive advantage over other solutions for applications such as oil and gas seismic algorithms, Monte Carlo algorithms for finance risk option calculation, OpenSSL security protocol, and deep learning. The large internal memory of the Kalray processor makes it an extremely powerful solution for Convolutional Neural Networks (CNN) deep learning algorithms.
K-ONIC, a smartNIC, solves a critical problem facing networking today. Existing ASIC-based solutions cannot adapt to the quick pace of change imposed by new virtualization protocols, forcing protocols to run on the host using up to 50% of its workload. Offloading this function, the K-ONIC enables the host processor to spend its time on clients’ services to generate revenue. Because of the deterministic architecture of Kalray’s Bostan-N processor, its configurable dispatcher, its crypto accelerators and the availability of 256 cores in parallel, K-ONIC is an ideal solution for implementing networking protocols such as OVS, VxLAN, NVGRE, and TRILL. On the storage side, K-ONIC will be able to pass through all data stored in a remote server using NVME and iSCSI protocols. There will be two versions of these smartNIC boards: K-ONIC 80 (half length, full profile); and K-ONIC40 (half length, low profile) offering 80GBE (up to 240 million of packets) and 40 GBE respectively.
TurboCard3 and K-ONIC80 will available in the fourth quarter of 2015.
About Kalray
Kalray is a fabless semiconductor company and pioneer in developing many-core processor solutions. Its innovative MPPA® architecture uniquely delivers supercomputing on a chip™, enabling real-time acceleration for cloud applications in video, networking, telecom and big data, and embedded high performance computing for embedded applications in aerospace, defense and automotive.
|
Related News
- Kalray and Pliops enters into exclusive negotiations to create a global leader in data accelerators for AI and storage acceleration
- Xilinx Launches Industry's First SmartNIC Platform Bringing Turnkey Network, Storage and Compute Acceleration to Cloud Data Centers
- Erwan Menard, Director of Infrastructure and Applications Modernization Solutions at Google Cloud, Joins Kalray Board of Directors
- Mobiveil to Exhibit at DAC in Avery Design Systems' Booth, Showcasing Portfolio of IP, Platforms, Solutions for Storage, IoT, Networking, Enterprise Markets
- NVM Express Releases NVM Express Specifications to Unify AI, Cloud, Client, and Enterprise Storage
Breaking News
- Jury is out in the Arm vs Qualcomm trial
- Ceva Seeks To Exploit Synergies in Portfolio with Nano NPU
- Synopsys Responds to U.K. Competition and Markets Authority's Phase 1 Announcement Regarding Ansys Acquisition
- Alphawave Semi Scales UCIe™ to 64 Gbps Enabling >20 Tbps/mm Bandwidth Density for Die-to-Die Chiplet Connectivity
- RaiderChip Hardware NPU adds Falcon-3 LLM to its supported AI models
Most Popular
E-mail This Article | Printer-Friendly Page |