Last Day Highlights from SC14

While SC14 is perhaps one of the best computer hardware shows on the planet, it isn’t just all about hardware. Software, especially GPU accelerated software, increasingly plays an important role in scientific discovery. One of the most broadly used HPC software packages has long been MathLab, and it is great to see them highlighting GPU acceleration on at least two sides of their booth.

GPU powered systems large and small were highlighted throughout the show floor, and with Google’s Nexus 9 tablet now shipping with a full 192 Kepler core powered Tegra TK1 SOC, it was only a matter of time before folks started building mini-clusters out of NVIDIA’s Jetson TK1 developer kit. French phone company Orange commissioned this system, built out of several dozen Jetson TK1 dev kits, for a data analytics project. If you wonder what all the excitement is about, you can order your own 192 core Jetson TK1 dev kit for only $199

One booth I missed earlier was SGI. Keeping the legendary SGI name and HPC legacy alive when cloud hardware vendor Rackable acquired the name and various assets in 2009, it is no surprise that SGI prominently displayed their 8-way high density (2RU) NVIDIA Tesla GPU platform in their booth. As GPUs become increasingly popular for machine learning and other big data analytics, SGI has a great opportunity to sell their 8-way box to both HPC and big data customers. And they are giving away a great looking tee-shirt too!

If 8 GPUs in 2RU isn’t enough for you, OneStop systems sells a chassis with 16 NVIDIA GPUs in 3RU. The OneStop chassis doesn’t include the server, and is intended for customers who want to cable up more GPUs to a server than it can physically hold. Several OEMs on the show floor also showed off OneStop chassis connected to their servers in other booths on the show floor.

As customers increasingly look to optimize the energy efficiency of their computing solution, various liquid cooling solutions continue to proliferate at SC14. And while they don’t build servers, the 3M booth had an interesting display of their fluid immersion technology, just one of the many liquid cooling solutions displayed out on the show floor.

Posted in Uncategorized | Leave a comment

Marc’s Best GPU Servers of SC14

This afternoon I spent a bit of time walking around the SC14 show floor and here is a list of my favorite NVIDIA GPU-powered servers. While very unofficial, I did follow a few guidelines. First, the NVIDIA partner had to have the server with NVIDIA GPUs displayed on the show floor. Second, there had to be someone in the booth who offered to talk to me intelligently about their GPU powered solutions. Based on those simple guidelines, here are my favorites. Its great to see so many new GPU powered solutions out on the show floor. If I missed one of your favorites, let me know, I’ll be out on the show floor again tomorrow and happy to take a look and listen.

Best Water-Cooled GPU Solution

HP’s Apollo 8000 also wins extra bonus points as the tallest GPU server. This beast is not for the casual user. Not only does this rack pack 72 server nodes with 144 GPUs into a single rack, it also manages to include all the Mellanox InfiniBand leaf switches you need for a full fat-tree topology. Besides the efficiency of HP’s unique liquid cooling solution, the Apollo 8000 also saves power with its 480V power supply and HVDC internal power distribution. While some of the other solutions may physically fit more than 144 GPUs in a rack, this is likely the densest GPU solution you can actually operate, especially when you consider it integrates in all the InfiniBand leaf switches. Downside? Only two GPUs per node are offered.

Best 8-GPU Solution Proven in Top500

Cray’s CS-Storm hits the other end of the GPUs per node range, supporting 8 GPUs in a compact 2RU form factor. As so many new GPU powered servers are now available, many of the systems out on the show floor have yet to be proven out in large Top500 configurations. Not the CS-Storm, that managed to be the only new server to break into the Top 10 of the Top500. While the CS-Storm is a standalone rack-mount server, it really is intended to be sold in complete rack configurations, with Cray integrating not only the power and optional rear door water cooling likely to be required by most full-rack configurations, Cray also does one of the best jobs of integrating an entire software stack including OS and management tools. Downside? The CS-Storm requires a non-standard width rack. Penalty points for only being displayed behind a plastic cover. SC14 is the last great hardware show on the planet, we want to leave fingerprints on your servers.

Most Improved 4-GPU Solution

While Dell ships a lot of NVIDIA GPUs, they haven’t historically had category leading products. Well that all changed on Monday with the new C4130. Moving away from earlier multi-node GPU designs and their complications, the C4130 is a new single-node, 1RU, 4GPU server which is even “EDR-ready” for the new Mellanox 100G InfiniBand, thanks to careful PCI slot layout. Dell also figured out how to support all 4 GPUs with a single x86 CPU, so customers who’s applications don’t need the extra serial performance can skip paying for the extra CPU. Especially with the new NVIDIA K80 GPU module sporting 2 Kepler GK210 GPU chips in each module (8 GPUs total), the C4130 promises to quickly become a workhorse GPU solution.

Best Non-x86 GPU Server

The IBM booth was happily displaying this unnamed future OpenPower based server. Supporting two NVIDIA K80 GPUs in 2RU, with up to 1TB of RAM, this promises to be an interesting server for customers wanting to get started with with Power + GPUs today before the next-generation NVLink connected Pascal + Power8+ systems start shipping.

Best 8-way PCI Design

Cirrascale has an interesting 8-way design that allows up to 8 NVIDIA GPUs to configured on a single PCIe root complex which is optimal for some applications with heavy GPU peer to peer communications. Most other 8-way designs split the GPUs between the separate PCIe root complex’s of the two host CPUs. The same server also supports a more traditional split PCI design. This isn’t the densest solution at 5RU, but as denser solutions typically require water cooling, the 5RU design isn’t likely to be an issue and in fact makes air cooling a lot easier than some of the denser designs.

Best Dense 8-way Standard Rack Mount Server

Penguin wins this one by managing to fit 8 K80’s into a standard width 2RU rack mount server. More than a quarter rack of these and you had better start shopping for water cooled rear doors. But if you are looking for a super-dense 8-way server, you should take a look at Penguin.

Posted in Uncategorized

NVIDIA SC14 Highlights

No doubt an incomplete list, I will update throughout the week, but some of my favorite highlights so far:

  • NVIDIA SC14 landing page
  • NVIDIA announces K80, the world’s fastest accelerator for data analytics and scientific computing, K80 Specs
  • In-situ visualization is one of the hottest new trends that users are demanding, NVIDIA partner Kitware Works with NVIDIA to Drive Massive Acceleration in HPC Visualization, Kitware announcement
  • Only 1 new system makes it into Top 10 of the Top500 list, an NVIDIA-powered Cray CS-Storm system, Top500 List. Of all the recently announced GPU-enabled servers, the Cray CS-Storm supporting up to eight NVIDIA GPU cards is one of my favorites.
  • San Diego Supercomputer Center’s new Comet supercomputer to feature NVIDIA K80 powered GPU nodes. Given the chilly weather this week in New Orleans, I’m looking forward to visiting Comet in San Diego.
  • K80 specs look great, but real world application performance from customers is the ultimate test, Amber Molecular Dynamics benchmarks on K80.
  • Posted in Uncategorized

    Introducing Summit and Sierra

    The U.S. Department of Energy today unveiled plans to build two GPU-accelerated supercomputers — expected to deliver at least three-times greater performance than today’s most powerful system — which will move the world closer to the long-held goal of exascale computing.

    On NVIDIA’s Parallel ForAll blog, Mark Harris explains how NVIDIA’s NVLink technology will enable faster, easier multi-GPU computing on these systems.

    The video below describes some of the ways the new systems will be used in scientific discovery.

    Lots of details in the NVIDIA White Paper on Coral

    Everyone at NVIDIA is super-excited to have our accelerated computing Tesla GPU platform used in these systems. Come by the NVIDIA booth at SC14 to learn more about the technology in these systems and how they will help empower a new generation of scientific discovery.

    Additional articles on the new systems from:


    insideHPC SlideCast

    Posted in Uncategorized

    NVIDIA SC14 Preview

    While SC14 doesn’t officially kick off until Sunday, I thought I would share a quick preview of what I’m looking forward to in New Orleans.

    Before you go, you definitely should download InsideHPC’s Print and Fly guide to New Orleans. Then head over to the NVIDIA SC14 Event Highlights page and get your must-do list scheduled.

    In the NVDIA booth we will once again have our GPU Technology Theatre with an amazing lineup of speakers from around the world. They are sure to inspire even more developers to learn CUDA, so you better sign up before the show for our Developer Zone Live, where right in the NVIDIA booth you can take a hands-on CUDA programming tutorial powered by QwikLABs and mentored by NVIDIA CUDA gurus. Who knows, next year your work may be selected to highlight in the theatre. Rumor has it that our CEO Jen-Hsun Huang will even be making a few appearances in the booth.

    If you don’t have time during the show, or can’t make it to the show, you can still take the same hands-on CUDA tutorials from the QwikLABs site. At the show or not, you will want to follow all the latest NVIDIA SC14 news online, follow @NVIDIA and @NVIDIATesla on Twitter and watch for hashtag #NVSC14.

    I’ll be in New Orleans early to attend HP-CAST, our partner HP’s HPC user group. There is still time to register and you won’t want to miss my HP-CAST Friday afternoon session titled On The Road to Exascale, Powering New Discovery with GPUs. On Saturday I’m presenting another session on GPUs for Machine Learning and Big Data.

    Check back during the week for more updates and analysis from the conference, it promises to be an exciting week!

    Posted in Uncategorized

    Visual Computing For Cloud Mobile – Part 2

    Yesterday I hadn’t planned on a part 2 for this topic, but thanks to Google buying Zync I have another great example to share. Prior to Google’s purchase, Zync was a company that helped you render in the cloud using many popular applications like Maya and V-Ray. It just so happened that Zync used NVIDIA GPUs running on Amazon Web Services as the visual computing resource to powering the rendering software. Google, no doubt, will likely move Zync off of AWS and onto their own cloud resources and perhaps incorporate the software into new Google services. As I said yesterday, it is becoming increasingly clear that mobile cloud architectures must support not just the consumption of visually rich content but also the creation.

    At SIGGRAPH a few weeks back, one of the hits of the show was a preview of CONSTRUCT, a sci-fi short film advancing the art of filmaking, visual special effects, and virtual production. The trailer below gives you an idea.

    CONSTRUCT’s filmmaker Kevin Margo was lucky enough to have access to several pre-release NVIDIA Quadro 5200 GPUs to use in his short, and those weren’t in any public cloud prior to being announced. But with their Zync acquisition, I’m sure it won’t be too long before some filmmaker is rendering their work on the Google Cloud.

    Posted in Uncategorized

    Visual Computing For Mobile Cloud

    Two pivotal announcements this week indicate that mobile cloud architectures are moving into the realm of visual computing. Either as consumers, or a employees, today’s mobile users are no more likely to be content with simple text box interfaces to their corporate data than they are to play mobile games with clunky yellow ducks. With mainstream phones supporting displays  of 1440 x 2560 pixels or greater, we want retina quality visual computing no matter what task we are doing. As I spend much of my time away from NVIDIA offices but still on corporate tasks, I’m super-excited by the newly announced VMware, NVIDIA, and Google partnership to deliver high-performance virtual desktops and workstation-class graphics to Google Chromebooks. The video below makes it pretty clear why.

    The second is Amazon’s announcement earlier this week of their plans to purchase Twitch for nearly $1B. Now while a lot of corporate folks may never have heard of Twitch, not so at NVIDIA. Twitch, which has quickly become the fifth largest source of Internet traffic, lets its nearly 45 million unique viewers per month stream their gaming sessions for a few friends or a few million friends to watch. And you don’t need a GeForce powered PC or laptop to stream to Twitch, the new Shield Tablet from NVIDIA lets you stream directly from a tablet.

    With both of these announcements, it is becoming increasingly clear that mobile cloud architectures must support not just the consumption of visually rich content but also the creation. While perhaps several hundred million PCs can connect to one or more clouds, billions of mobile devices will soon be connected and capable of visual computing. And when you get that new mobile phone, Chromebook, or Shield Tablet that has a display resolution several times greater than your old workplace VGA monitor, and runs all your high performance desktop apps, the power of visual computing for mobile cloud will become clear.

    Posted in Uncategorized