In this podcast, the Radio Free HPC team looks at Facebook’s testimony before Congress.
“Dan Olds and Henry Newman are the only guys on deck today, Rich and Shahin are either traveling or doing something useless. However, this episode is ground breaking. Henry and Dan agree on everything ranging from the Facebook security “scandal” to the implications of GDPR. It’s a shocking and stunning turn of events.”
We should have a full crew next week and it’s hard to believe that Dan and Henry will agree yet again, so the universe will be back in order.”
NVIDIA CEO Jensen Huang unveils the DGX-2 supercomputer.
In this podcast, the Radio Free HPC team reviews the highlights of the GPU Technology Conference.
From Rich’s perspective, the key HPC announcement centered around new NVIDIA DGX-2 supercomputer with the NVSwitch interconnect.
The rapid growth in deep learning workloads has driven the need for a faster and more scalable interconnect, as PCIe bandwidth increasingly becomes the bottleneck at the multi-GPU system level. NVLink is a great advance to enable eight GPUs in a single server, and accelerate performance beyond PCIe. But taking deep learning performance to the next level will require a GPU fabric that enables more GPUs in a single server, and full-bandwidth connectivity between them.
NVIDIA NVSwitch is the first on-node switch architecture to support 16 fully-connected GPUs in a single server node and drive simultaneous communication between all eight GPU pairs at an incredible 300 GB/s each. These 16 GPUs can be used as a single large-scale accelerator with 0.5 Terabytes of unified memory space and 2 petaFLOPS of deep learning compute power.”
Henry Newman and the Amazing Technicolor Dreamcoat
After that, we do our Catch of the Week:
Henry is very proud of his new sport coat.
Shahin notes that the data privacy the key story these days. Facebook is on the ropes, and their CEO Mark Zuckerberg will appear before Congress this week to answer questions about the Cambridge Analytics scandal.
Quantum Computing. This week Google unveiled their 72-Qbit Bristlecone Quantum Chip, which they claim has them on the road to Quantum Supremacy. At the same time, quantum seems to be going mainstream, as evidenced by a panel discussion this week at SXSW. You can listen to the full recording of the session on insideHPC.
Quantum Computing was the focus of a panel discussion at SXSW this week. From left: Bo Ewald (D-Wave Systems) Antia Lamas-Linares (TACC) Patricia Baumhart (D-Wave Systems) Jerry Chow (IBM) and Andrew Fursman 1Qbit.
Shahin revisits the Department of Deep Thinking on the topic of why Computational Trust is expensive. It’s all about Faith, Fees, and Fiat. So if you happen to have Trust in your life for free, you should cherish it.
“According to an interview in Primeur Magazine with EPI project coordinator Philippe Notton from Atos, the project involves not only a processor, but an accelerator as well. Will it be based on ARM, OpenPOWER, or something else like RISC-V? We will have to wait and see.”
After that, we do our Catch of the Week:
Rich likes the story this week about Supergeek, the new mascot at the Barcelona Supercomputing Center.
Dan likes the recent story on the origins of the HPCG benchmark, which offers an alternative way to measure the performance of the world’s top supercomputers. At the same time, Shahin reminds us that ACM Fellow David Kuck said that the minimum performance from a machine is more interesting than the maximum performance.
Paul Kocher is a Security Technology Advisor at Rambus.
In this podcast, the Radio Free HPCteam looks at the tradeoff between chip performance and security. In the aftermath of the recently disclosed Spectre and Meltdown exploits, Cryptograpy guru Paul Kocher from Rambus is calling for a divergence in processor architectures:
The direction that we need to go as an industry though is …We need to stop trying to build one processor architecture that is great for playing video games and doing wire transfers. We need to build architectures where there are cores and software stacks designed for security that can be slower, that can be simpler, and we need separate ones that are optimized for performance.”
In this video, Paul Kocher presents: Spectre – Exploiting Speculative Execution. Kocher is the lead author of the paper on the Spectre processor vulnerability.
In this podcast, the Radio Free HPC team looks at HPE’s new CEO, Antonio Neri, a longtime HPE executive who previously served as President of the company. As the number 1 server vendor in the HPC space, this change will be one to watch as we transition to the exascale era in the next five years or so.
In this podcast, the Radio Free HPC team looks at NVIDIA’s new EULA GeForce Software user license, which prohibits the use of consumer GPUs in the datacenter. We are not looking to beat up on anyone, but our focus is on what limitations this might mean for the industry:
No Datacenter Deployment. The SOFTWARE is not licensed for datacenter deployment, except that blockchain processing in a datacenter is permitted.
When we purchase hardware, aren’t we free to use it any way we please? And why do BitCoin miners get a pass during a worldwide GPU shortage?
Plus, what do we mean by the word “datacenter?” anyway? Shahin predicts the imminent proliferation of 5G networking capabilities will move computing closer to the edge, thereby changing what we mean by the term “datacenter.”
While Europe already had a number of exascale initiatives under way, this is a major step forward in that it puts up the money. Under a new legal and funding structure, the Commission’s contribution will be $486 million, or roughly half of the projected EUR 1 Billion total. Before the new structure was put in place, the Commission was effectively limited to contributing only 20 percent of HPC initiatives undertaken with member states.