End2End: From Silicon to System Delivery
summary
Marvell delivers ThunderX2 end-to-end: silicon with 8 memory channels per SoC and out‑of‑order cores supporting up to four threads per core, a CRB reference platform with CRB reference platform with Gigabyte and AMI firmware/BMC, plus upstream kernel work 9–12 months before production. HPE’s Apollo 70 and Sandia’s Astra reach Top500, Cadence ports EDA tools with a Flexera licensing wrapper, and customers report improved performance per dollar and memory‑bandwidth wins for HPC and cloud workloads.
AI-written summary · about
chapters (10)
- 0:00Introduction to end-to-end delivery
- 4:20Marvell’s infrastructure focus
- 10:31Enabling platform and standards
- 12:00HPC use cases with Sandia and Los Alamos
- 15:51Cray and compiler performance
- 17:41Microsoft Azure and OCP integration
- 19:06University of Michigan Big Data cluster
- 21:13Cadence EDA tools on ThunderX2
- 23:45Performance metrics and optimization
- 27:38Call to action for Linaro and ecosystem
related talks
05mFrom Momentum to Mainstream
15mKeynote: Co-Designing Software and Hardware: Pillars of Advancing RISC-V for Application Success
12mAccelerating Software Development for High Performance Chiplet...
19mInnovation at the Intersection of Hardware and Software
14mRISC-V Spotlight: RISC-V: Everyone Wins
15mKeynote: From Hardware Innovation to a Thriving RISC-V Ecosystem
01h 01mPanel: Building a Scalable RISC-V Software Ecosystem
39mSoftware Defined Hardware on Heterogeneous SoCs Using OpenAMP
