This has been an astoundingly quiet Winter Classic. And it’s all my stupid fault! I had a packed international travel schedule towards the end of WC26, plus big projects coming to a head. My body took that opportunity to attack me with a combination of pneumonia and sepsis, which took me off my wheels for a couple of months at the worst time possible.
Anyway, I’m back up and running (not quite 100% yet but spinning up) and need to let everyone know what happened in WC26 – the kids worked hard, did well, and they deserve to have the closure.
Here’s how it went:
Module One: The HPC Crash Course
This was a first time event in 2025. A mentor dropped out at the last minute, creating a hole that had to be filled. No way they’re getting a week off, not on my watch! I had found the PSC Bootcamp series of HPC tutorial videos in 2024 and posted them as a training resource. Why not make them watch the 16 hours of video and then test them on it? Yeah, that’s the ticket!
The 2025 students told me that this material was so helpful in explaining HPC concepts and how to use them that I had to keep it in the competition and have it as the first module. So I did…with some changes…
I think the exam was too easy in 2025, so I went hardcore on it. No more multiple-choice selections, no more true/false questions. This year? It was all essay questions, but with a twist to make it even more difficult – word limits. With those, they can’t just write and write until they’ve covered every possible answer.
Even with this, the student teams did well with an average score of 86%. Damn, I have to find a way to make it even harder next year. The Texas Tech Matadors took first place with 95.4% with the FSU House of Coder team only three tenths behind at 95.1%. In third were the Texas Tech Red Raiders with 92.6 and the FSU Codaholics grabbed an honorable mention in fourth with their 88.7.
Module Two: LINPACK & HPCG the HPE Way
Time to jump into their first supercomputer and take it for a spin. As usual, the first hands-on task is conquering the “Dongarra Suite” of benchmarks. We had some real-life trouble during the competition week with technical problems on the clusters. It was fixed quickly and the deadlines were eased a little bit, but it hurt some teams because they waited until the last minute to start. Normal stuff, but it did bring the average scores down.
On LINPACK, the overall normalized average was 65%, lower than usual, but this was because some teams missed the turn in deadline or didn’t turn in valid results. We saw the same thing on HPCG, where the average scores were just barely over 40%.
Santa Cruz Team One had the highest score, receiving the full 100 points. Less than 1% behind them was cross campus rival Santa Cruz Team A. The Channel Island Wave Runners got on the scoreboard with a 95.78 while the TTU Matadors finished in fourth with 91.91%
Santa Cruz Team One also took the top score on HPCG but this time followed by TTU Team Red Raider a mere 3% behind. FSU’s House of Coders missed second place by only 2%, earning third place by a wide margin over the rest of the field.
Module Three: Pittsburgh Supercomputing Quantum Tour
In this task, students had to perform a number of tasks on quantum simulators using GPUs at various precisions and explain their findings by answering a number of highly detailed questions.
Nine out of eleven teams turned in valid results – which was great. The normalized scores on this one spanned quite a range which served to separate the teams a bit. Santa Cruz Team One was the top team on this task, fractionally ahead of second-place FSU House of Coders (who is having their best competition ever so far.) The TTU Red Raiders was barely behind FSU, landing in third. The other TTU team, the Matadors, get a head nod for a fairly close fourth.
Module Four: Diagnose & Fix Oak Ridge Clusters
This was a really cool challenge. Oak Ridge presented the teams with an eight-node cluster. Do they all work or are there problems? (Hint: there are some problems.) They had to test each for functionality, then for performance (are they all in the same range, if not, why not?) and finally, verify the output. Not easy, but it gives them valuable skills in hardware testing, and they’ll learn how to automate testing to make life easier.
Eight teams successfully completed this challenge with an average score of 55%. Once again, Santa Cruz Team One was once again number one on this module. The Matadors from TTU put up a fight, taking second with 90%, just ahead of their cross-campus Red Raider rivals by less than 2%. FSU House of Coders completed the top four on this one.
Module Five: The Mystery Mentor
The final boss monster this year was, drum roll, Oak Ridge with a quantum machine learning challenge (QML.) Like last year, the kids got to run code on a quantum simulator and then on a real quantum system, which is certainly a resume builder, right?
The first part of the task was to evaluate a benchmark code using one to eight GPUs on a quantum simulator for accuracy and performance. After answering a bunch of questions, they move on to the next phase where they load the pretrained QML routine on a IQM system and compare their results with the simulator. And, of course, show their work and explain their results.
Yet one more time, Santa Cruz Team One nabbed the top slot – no, they didn’t. The FSU Codaholics just barely edged out Santa Cruz by fractions of a point! We haven’t seen the Codaholics in the top ranks on previous challenges, but they did it, great job. Santa Cruz Team One placed second and another unheard from team, the Channel Islands Wave Runners came in third, behind the leaders by just a couple of points. The TTU Red Raiders finished fourth on what was a very difficult set of tasks.
And in the end….
Santa Cruz Team One almost pitched a perfect game by winning every task. Only the TTU Red Raiders had done this before, so it would have been a big deal. But even so, Santa Cruz had an amazing performance with their overall score of 97% of the possible points. The Red Raiders did well with a score of 92%, but did miss out on their Winter Classic three-peat dream.
The FSU House of Coders achieved the highest finish in FSU Winter Classic history, coming in third place. After their disappointing LINPACK outing, the Channel Islands Wave Runners lurked under the radar but steadily piled up points – enough points to give them fourth place and an honorable mention on the final scoreboard.
Great job by the kids in what was a difficult year.