Tuesday, July 21, 2020

RadioFreeHPC is moving to RadioFreeHPC.com

Greetings to all the fans of the RadioFreeHPC blog and podcast.
Thank you for being here and thank you for giving us reason to do this blog and podcast. We hope you like it and look forward to hearing your take and suggestions.

We have big news!

We are moving, consolidating things in the radiofreehpc.com website.
The site includes other links and ways to contact us or subscribe to the podcast.

The podcast will also continue to be featured in InsideHPC.com.

We hope to see you there and to hear from you.

Thank you,
RadioFreeHPC team (DanO, Jesse, Henry, and Shahin)

Friday, July 3, 2020

Is it Time for Virtual GPUs?

VMware's Bitfusion and Dell

We start with team introductions, as usual, but drop a bombshell on listeners:  we’re going to be missing Jessi for the next couple of months as she goes off to Marine Officers Candidate School – or as she calls it “screamy summer camp.” It’s sort of like an all-inclusive Sandals Resort. You get your room, meals, drinks, plus many interesting activities and outings – plus thousands and thousands of push-ups. And to top it off, you’re nearly a Marine Corps officer when you’re done! We’re very proud of our Jessi and know she’s going to do great. In unrelated news, Henry reports in from Las Cruces, straight from his echoey bunker. He’d doing great in the new house.

AI Anywhere, Anyone?

Our main story this week concerns Dell’s integration of VMware and Bitfusion to provide “AI anywhere” type of infrastructure where any set of systems can get access to typically underutilized GPUs sitting anywhere on the network. This could mark the end of “GPU silos” in both enterprise and HPC data centers and lead to much better utilization of expensive GPU assets. The guys bat around the pros and cons of virtualization. Henry raises the point that not all applications will work right out of the box with this type of virtualization (or any virtualization, for that matter). The real key according to Henry is how much computation you need vs. how much data you need to move. So to justify the movement of data, the computational density of the job needs to be high. We kick these concepts back and forth and it’s a good and meaty discussion. Dan and Henry even violently agree a couple of times as we go through it. Amazing.

Reasons Why No One Should Ever be Online. Ever.

 Russian agents have been attacking a critical part of the Exim Email Transfer Agent – which is very bad. This could allow a ‘man in the middle’ style exploit of emails passing through these systems. Or highly targeted attacks on individual organizations. This is why I’m starting to convert back to fax.

Catch of the Week

Jessi:  Jessi’s catch of the week is, gasp, another podcast – and not ours! She heartily recommends that our listeners also listen to the Modern Figures podcast. Their motto:  A podcast elevating the voices of black women in computing. Radio Free HPC approved and certified, so you have our permission to listen.

Henry:  Several European academic supercomputers have been hijacked in order to run crypto mining. It is unknown how much crypto they mined or the value of it. The attackers were particularly sophisticated when it came to covering their tracks, making the systems look like there wasn’t anything running when they were, in fact, furiously running crypto routines. Crafty stuff.

Shahin:  Gives a shout out for Hot Chips, a virtual event this year that will be held later this August.  All of you who always wanted to attend and couldn't should take advantage of this. Either way, we'll probably do a post-show show to bring you the highlights of the event that's putting Silicon back into Silicon Valley.

Dan:  The US of A is back in space! Yay! 

Join us!

* Download the MP3 
* Sign up for the insideHPC Newsletter
* Follow us on Twitter
Subscribe on Spotify 
Subscribe on Google Play 
Subscribe on iTunes 
RSS Feed
* eMail us

Tuesday, June 30, 2020

NSF, QIS Funding, But First...

18th Listener Awards!

Run-of-the-mill start as we introduce the team. Henry has quite a bit of echo since he’s now firmly ensconced in his rammed earth survival compound.

Big news this week! We finally received our 18th email in our first ever RadioFreeHPC email-to-win contest. First, let’s see what they win…

The first item in our Grand Prize bundle is a circa 1999 suede leather IBM S/390 laptop bag. It’s roomy at a generous 16”x14”, large enough to house the biggest 17” laptop or maybe 15 or 20 tablets. If that’s not enough room, it also expands an additional 4-5” on the bottom. There are compartments galore under the flaps and you’d better fill them up with soft foam since there’s really no padding around the laptop compartment.

The suede leather is a nice touch that adds weight to the bag, it’s not a lightweight in any sense of the word. We’d advise covering the bag if you’re going to be out in the weather, because the suede is sure to stain if it is exposed to water.

This bag is in brand new condition and has never been used. Ever.

The second piece of this cornucopia of delight is a IBM Mainframe 50th anniversary medallion. It’s a very light weight metal medallion housed in a stunningly clear plastic frame. Put it on your mantle at home and impress your friends and family with how important you must be to have such a keepsake.

The final gift is a cassette tape of an IBM Talk S/390 episode covering S/390 integrated solutions. If you still have a cassette player, this will provide minutes of mainframe tech talk, marketing tips and keys to closing that big deal.

So you want to know who won this amazing collection of swag? Listen to the show and find out, or follow us on Twitter and look a couple of weeks back!

Government Bills Aim to Up the Ante for Science

In the rest of the show we discuss some recent government news. First is that the US House is looking to devote an additional $1.25 billion over the next five years. But even bigger is the proposal to expand the US National Science Foundation by $100 beeelion – which is more than 4x the size of the existing organization. We discuss the implications of this and how this might change the game in terms of base research. There is also a bill to significantly dial up funding for Quantum Information Science. All good and the team s excited to see these moves and hopes they'll proceed swimmingly!

Reasons Why No One Should Ever be Online. Ever.

Brace yourselves, GitHub has been attacked! Developers and users alike unaware that they’re downloading malware when they’re getting their open source on. Ouch.

Catch of the Week

JessieCampaign Zero and Police Scorecard aim at reducing police violence.

HenryUK broadband performance is holding up in these virus-rampant times. Yay for the UK, great job.

DanSkinny gene discovered! Eat all you want and never gain weight. Is it true?

Join us!

* Download the MP3 
* Sign up for the insideHPC Newsletter
* Follow us on Twitter
Subscribe on Spotify 
Subscribe on Google Play 
Subscribe on iTunes 
RSS Feed
* eMail us

Monday, June 22, 2020

New Top on the TOP500 – 415 PF!

Breaking News Edition

We have a new #1 on the TOP500 list of most powerful supercomputers! Big gets bigger by a factor of 2.8x as Fujitsu’s “Supercomputer Fugaku” tops the list at 415 PFlops.  There are also an additional three new entries in the top ten. We break down the top of the list in this fascinating episode of RadioFreeHPC.

Listen to us now! It will help you to amaze your friends and dismay your enemies with your newfound knowledge of the list. We have it here and first! Or at least not much later than others!

Join us!

* Download the MP3 
* Sign up for the insideHPC Newsletter
* Follow us on Twitter
Subscribe on Spotify 
Subscribe on Google Play 
Subscribe on iTunes 
RSS Feed
* eMail us

Friday, June 19, 2020

@HPC_Guru Speaketh! Kind of.

Our Guest Today is...

In this extraordinary episode of Radio Free HPC, the crew interviews the industry icon that is @HPC_Guru. This is the first time that anyone has been granted an interview with him and we’re proud to have been chosen for this honor.

We posed an even dozen questions and received very thoughtful responses, which we rendered out in a machine voice in order to fit our podcast format. In the interview, HPC_Guru tells us his top five cool things in HPC today, why he remains anonymous, where he thinks HPC hardware will be in 10 years and who he thinks will be the first to reach exascale. And that’s just four of our 12 questions!

We don’t have to tell you that @HPC_Guru is a legend in the industry, as is his Twitter account. He has more than 15,000 followers and has tweeted over 38,000 times. Just to put that in context, if his average tweet is 150 characters, then he’s tweeted 5.7 million characters. Or if you look at it as words, HPC Guru has beaten the hell out of Leon Tolstoy’s War & Peace. Tolstoy came up with a piddling 587,287 words in his novel while HPC Guru has written roughly 712,500 words – and HPC Guru has written about more difficult content. Supercomputing is much more complex than Napoleon’s invasion of Russia. While this is impressive, it’s not quite as many words/characters as Timothy Prickett Morgan writes in a typical year.

If you printed up HPC Guru’s tweets, the tome would come in at more than four pounds, and that’s single spaced. That’s as much as a high-end laptop, including the storage and maybe even the power brick. Looking for another metric? If @HPC_Guru printed out each of his tweets, two per sheet of 8.5 by 11 inch paper, and laid them end to end, it would stretch a little over 3.3 MILES.

Give this groundbreaking episode a listen, in fact, listen to it twice to get the full impact.

Join us!

* Download the MP3 
* Sign up for the insideHPC Newsletter
* Follow us on Twitter
Subscribe on Spotify 
Subscribe on Google Play 
Subscribe on iTunes 
RSS Feed
* eMail us

Friday, June 12, 2020

HPC on the Edge

Edge Computing Book Makes Sense of It All

After our typical inane opening, which includes an update on Henry’s Las Cruces bunker with  21” rammed earth walls, we get quickly down to business and begin our interview with very special guest, Dr. Cody Bumgardner, Assistant Professor of Pathology & Laboratory Medicine at the University of Kentucky.

Our topic? Edge computing, what it is, what is isn’t, and why it’s important. Cody has just authored a book “Making Sense of Edge Computing.”

Dan starts us off with a question about the definition of edge computing, which as it turns out, covers a lot of ground. Henry comes in with a question about how 5G might impact edge computing and Shahin follows up by asking if 5G really has much to do with edge computing at all. There is some overlap, but 5G is a transmission mechanism rather than a computing mechanism, meaning that it’s the 5G device and what it is tasked to do that can make it either an edge device or not an edge device.

As the show goes on, we talk about the edge capacity in place and what still needs to be built, along with current and potential use cases. A good time is had by all.

Reasons Why No One Should Ever Be Online. Ever.

Henry shocks us yet again with news that people are being scammed by work-at-home deals that are actually ‘money mule’ schemes to help criminals launder their ill-gotten gains. Dan seems a little too curious about the mechanics and potential pay-offs of these deals. 

Catch of the Week

Jessi: Jessi is pointing out that she has now moved for the third time in one academic semester and that she has more books than anything else. Ouch.

Henry:  Henry is off yelling at his contractors when his turn comes up, so no catch from him this week.

Shahin:  Article about how to program Aurora, highly recommended by Shahin.

Dan:   Discusses how the ISC2020 Student Cluster Competition will be virtual this year. He’ll be covering it as usual, but with a LOT of Zoom. Every team is going to be using the same cluster, but sharing it. Stay tuned for more news.

Join us!

* Download the MP3 
* Sign up for the insideHPC Newsletter
* Follow us on Twitter
Subscribe on Spotify 
Subscribe on Google Play 
Subscribe on iTunes 
RSS Feed
* eMail us

Monday, June 1, 2020

A is for Ampere, Nvidia A100's Public Debut

NOTE:  The publication of this episode was delayed due to the untimely passing of our partner and pal Rich Brueckner. So what we’re announcing as "breaking news" isn’t so fresh today, but our takes on what NVIDIA’s new A100 processor brings to the table are still valid. 


Breaking News! This special edition of RadioFreeHPC takes a deep dive into NVIDIA’s spanking new A100 GPU – which is an impressive achievement in processor-dom. The new chip is built with a 7nm process and weighs in at a hefty 54 billion transistors and capped at 400 Watts. It sports 6,912 FP32 CUDA cores, 3,456 FP64 CUDA cores and 422 Tensor cores.
This 8th generation GPU, using what the company calls its Ampere technology, is a replacement for both their V100 GPU and Turing T4 processors, giving the company a single platform for both AI training and inferencing.
We talk about the specs of the A100, breaking down its game both in terms of typical HPC FP64 processing and FP32 (and lower precision) computing for AI workloads. On the HPC side, the new GPU seems to offer an across the board 25% speedup, which is substantial. But the A100 really shines when it comes to tensor core performance which the company reports at an average speed up of 10x on Tensor Core 32 bit vs. V100 FP32.
New features of the A100 include Sparsity (a mechanism that doubles sparse matrix performance), a much speedier NVLink (2x), and a hardware feature that allows the A100 to be partitioned into as many as 7 GPU instances to support individual workloads.
All in all, this is an amazing new processor, a behemoth large and hot, but so fast, chip that is heavily tilted towards new AI and Tensor workloads with a passing but welcome nod to 64-bit HPC apps.

Join us!

* Download the MP3 
* Sign up for the insideHPC Newsletter
* Follow us on Twitter
Subscribe on Spotify 
Subscribe on Google Play 
Subscribe on iTunes 
RSS Feed
* eMail us