Topics including (but not limited to)....
Radio Free HPC (RFHPC) started in 2012 and, after a short break of 193 months, has returned to the airwaves (cyber waves?) with co-hosts Dan Olds, Henry Newman, and Shahin Khan.
The podcast covers a wide range of topics but is particularly focused on data center technology and challenges. These include high performance computing, power and cooling, new technology, and, really, whatever the hell they want to talk about.
Quick FAQ
"Why are you just covering HPC? There's this new thing called AI, have you heard of it?"
Yeah, we've heard of it and we do cover it. To us, HPC is a broad term used to refer to a very broad group of apps/workloads that have common characteristics.
- They are computationally intense
- Time to solution and quality of solution are critical
- The output is important enough to justify investment in high-end clusters and associated hardware/software
AI meets all of the criteria above just like bioscience, aerospace, weather prediction, and several other domains. In fact, AI is or will soon be folded into the workflows of nearly every 'traditional' HPC and enterprise application over time. This is going to put a lot of pressure on data centers and we'll be talking about how to deal with it.
"Was the old version of Radio Free HPC popular? I've never heard of it..."
Yes, it rose from obscurity to being mildly popular and, at times, entertaining. At our peak, we had as many as 16 sustained listeners and as many as 19 theoretical max listeners.
"I listened to the old show a couple of times and liked the arguing. Will there be more of that in this reboot?"
Yes, in fact we expect as much as 29% more arguing and a significant increase in the number of personal insults per episode.
"How often will this thing come out and how do I find out about it?"
We used to try to do weekly shows but failed due to other commitments, travel, etc. And there were just weeks when nothing important happened. But to answer your question, could be every couple of weeks, could be once a month, or we could record a marathon 8-10 hour single episode. Depends on what's happening and our mood.
We're getting together our RSS feeds and such but in the short term we'll be publicizing the episodes through the RFHPC X account and our individual social media accounts.
Ok, enough questions. If you want to learn more about the show and hosts, check out the "About" page.
Here are our most recent episodes along with the 345 or so previous editions, enjoy!
Latest Episode
-
We’re finally starting to cut into our product backlog and, as a first, we’re presenting our special road trip double episode where Jessi and Dan traveled to the House of Henry, buried deeply in Las Cruces, New Mexico. We visit his rammed earth compound and recorded two RFHPC episodes there – and one of them is a video episode!
In our first installment, Jessi is back! She passed her Marine Officer Candidate course – even though she fractured her femur in the process. So in addition to the rigors of taking the course, she also had to contend with a huge amount of pain. She’s earned even more respect from us – as if we weren’t in awe of her already.
Getting down to the HPC, Jessi, Dan and Henry discuss the recently released Power10 processor, kicking around the pros and cons in our usual way. To us, this is another instance of IBM leapfrogging CPU competitors by bringing out a processor that has great specifications and has all the earmarks of a game changer. It’s a 15 core, AI focused chip that can handle up to 410 GB/s peak memory bandwidth out of the box, with 1TB/s in the near future using their PowerAXON interface.
Multiple systems can be attached together and share memory, which yields a very powerful SMP system with up to 2 PB shared memory. In the intro slides, IBM showed several different memory sharing configurations including point-to-point connections where every system has a direct connection to every other system or hub-and-spoke configurations where all systems connect to a set of central servers. The system is memory composable, meaning that in a shared memory cluster, each workload can be assigned memory based on need, not based on the physical memory installed in its home node.
There wasn’t a lot of HPC content in the initial introductory slide set. One slide did show a 5x LINPACK gain and another showed a 3x floating point increase vs. Power 9. Solid results, but most of the presentation seemed to discuss enterprise features and performance rather than HPC. All in all, it’s a pretty sporty chip. But will IBM be able to ride this processor into HPC relevance again? That’s the question.
On one of our many tangents, Henry and Dan get into an argument as to whether the Sun E10000 was a SMP system or not. Henry says it was a NUMA system while Dan strongly asserts that it was a SMP system. One of these two guys actually worked at the company and worked with the engineers who built the system; thus he absolutely knows the true answer. Who was that guy? Dan. Case closed.
Catch of the Week
Jessi: she’s back! And brings to our attention that all of her family who were trying to watch her Marine Corps graduation ceremony were inflicted with heinous malware that infected many of their devices.
Henry: Has nothing, empty net due to having house guests (Jessi and Dan). But we did have a discussion of the pig that he’s roasting for dinner tonight.
Dan: Teams are signing up to the Winter Classic Invitational Cluster Competition, but we still want more, so check out www.winterclassicinternational.com
Second Feature: Exploring Henry’s House live and in color!
Join us as we interview Henry about his rammed earth home in sunny Las Cruces, New Mexico. While sitting in his kitchen we talk about the origin of his rammed earth design, what he has hidden under the house (the bunker), and the innovative features contained therein. One of the most interesting parts of the discussion is how Henry, who is always bringing up examples of people being hacked by their online devices, has somewhere around 500 online devices in his new home. Everything from the Kevlar/titanium window shades to his washer/dryer to his music system, etc. etc. Watch the episode to hear his explanation for why he has so many potential hacker access points – it’s a pretty good one.
* Follow us on Twitter
* Subscribe on Spotify
* Subscribe on Google Play
* Subscribe on iTunes
* RSS Feed
* eMail us
Recent Episodes
-
Episode 346: SC22 Preview
We’ve got the whole crew here as we look at the upcoming SC22 event. How many will be there? What will be happening? We answer only some of these questions.…
-
Episode 345: Dow Gets Cloudy
Dan rides alone on this one as he interviews the HPC folks from Dow who are behind their move to cloud bursting. Interesting and informative. *Download the MP3 * Follow…
-
Episode 344: Finally back! And with a new co-host!
We introduce new co-host Ben Carr and talk some NIST and Intel trevails. RadioFreeHPC is distributed in partnership with HPCwire. *Download the MP3 * Follow us on Twitter * Subscribe on Spotify * Subscribe…
-
Episode 343: It’s GTC, Baby
In this episode of RadioFreeHPC, Dan and Jessi talk GTC along with special guest Todd Weiss. RadioFreeHPC is distributed in partnership with HPCwire. *Download the MP3
-
Episode 342: Jazz, Melting Diamonds, and Schrödinger’s Cat
In this episode of RadioFreeHPC, the crew round up the most intriguing batch of recent HPC topics including Graphcore, HPC efforts to better understand melting diamonds and Jazz, as well…
-
Episode 341: Ponte Vecchio
In this episode of RadioFreeHPC, the crew talk Intel’s Ponte Vecchio. RadioFreeHPC is distributed in partnership with HPCwire. *Download the MP3 * Follow us on Twitter * Subscribe on Spotify * Subscribe on…
-
Episode 340: HPC with Trish Damkroger and Brent Gorda
In this episode of RadioFreeHPC, Dan and Jessi are joined by Trish Damkroger and Brent Gorda to talk HPC topics, thoughts, and intriguing trends. RadioFreeHPC is distributed in partnership with HPCwire.…
-
Episode 339: 12 Cores Per Soul
In this episode of RadioFreeHPC, the crew discusses Intel’s investment into RISC-V, the 2022 Winter Classic student cluster competition, and quantum computing in Canada. RadioFreeHPC is distributed in partnership with HPCwire.…
-
Episode 338: NVIDIA Slinks Away, Meta Bulks Up, And We Zap Bugs
In this episode of RadioFreeHPC, the crew talks Meta’s new AI supercomputer, NVIDIA abandoning ARM, and our latest competition winners. RadioFreeHPC is brought to you by the Surfs Up Hippie and…
The High Performance Computing Podcast



