Alan Sill's Avatar

Alan Sill

@alansill

Physics/cloud/data centers/HPC. This one's for personal viewpoints and technical stuff I find cool. I'm open to questions and try to reply to serious or interesting ones. Also at https://mast.hpc.social/@AlanSill and https://hpc.social Slack and Discord.

324
Followers
228
Following
805
Posts
21.11.2024
Joined
Posts Following

Latest posts by Alan Sill @alansill

Post image

#AMD released the "CPU Timing and Power Side Channel Guidance" pdf 71030 v1.1:
docs.amd.com/v/u/en-US/71...

01.03.2026 15:56 πŸ‘ 3 πŸ” 2 πŸ’¬ 0 πŸ“Œ 0
Preview
Oak Ridge spawns institute to curb AI datacenter power surge Lab aims to link power, cooling, and workload management to ease strain on the US grid Oak Ridge National Laboratory (ORNL) is hoping to turn its technical expertise to the problem of growing electricity demand from AI datacenters.…

ICYMI: Oak Ridge spawns institute to curb AI datacenter power surge

28.02.2026 19:01 πŸ‘ 7 πŸ” 2 πŸ’¬ 0 πŸ“Œ 0
Preview
DOE climate report 'demonstrably incorrect', say scientists in new analysis A leading climate scientist has sought to set the record straight over "demonstrably incorrect" claims made in a major U.S. government report that misrepresented his work and downplayed the role of hu...

DOE climate report 'demonstrably incorrect', say scientists in new analysis

phys.org/news/2026-02...

28.02.2026 18:11 πŸ‘ 21 πŸ” 9 πŸ’¬ 1 πŸ“Œ 0
Preview
Modeled and Observed Stratospheric Temperature Changes: Implications for Fingerprint Studies Human fingerprints, marked by tropospheric warming and stratospheric cooling, are evident in Earth's changing atmospheric thermal structure A claim to the contrary in a recent report of the US De...

One topic covered in the interview is the US DoE's flawed climate working group report of 2025. Yesterday Ben and colleagues published this article, demonstrating the DoE's report was factually incorrect in its claims about stratospheric temperatures agupubs.onlinelibrary.wiley.com/doi/10.1029/...

25.02.2026 14:52 πŸ‘ 6 πŸ” 3 πŸ’¬ 1 πŸ“Œ 0
Preview
First look at a system wide view MethaneSAT data show emissions vary widely by region, exceed inventories & industry’s goals

Methane leaks are much bigger than industry or the EPA reports!

So, gas is usually much dirtier than claimed by the gas industry and its supporters like @thebti.bsky.social
.
Some gas is probably dirtier than burning coal!

LNG also makes gas dirtier.
www.methanesat.org/project-upda... #energysky.

26.02.2026 19:18 πŸ‘ 24 πŸ” 10 πŸ’¬ 0 πŸ“Œ 0
Preview
NSF Intent to Restructure Critical Weather Infrastructure

The US NSF has put out a "dear colleagues" letter that is essentially a yard sale list for @ncar-ucar.bsky.social

It asks questions such as, who'd like to take over some of NCAR's programs? Is anyone interested in commercial use of the Mesa Lab?

If you have thoughts, send them to NSF_NCAR@nsf.gov

25.02.2026 15:36 πŸ‘ 159 πŸ” 102 πŸ’¬ 11 πŸ“Œ 10

This is an incomplete summary of the also incomplete analysis by the old guard referenced, IMO. Hope the presentation and discussion were better! (Post links if they are available please.)

25.02.2026 20:29 πŸ‘ 0 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0
They're taking away the most important weather lab in the US
They're taking away the most important weather lab in the US YouTube video by The Weather & Climate Livestream

a video highlighting just how important NCAR is, from a bunch of scientists who use this incredible public resource daily

25.02.2026 17:08 πŸ‘ 65 πŸ” 31 πŸ’¬ 0 πŸ“Œ 1

I've been to seminars that came close. My favorite memories along these lines, though, are from attending SLAC Theory group talks in which the group politely listened to the beginning of the speaker's remarks, then collegially took their argument completely apart and put it back together for them.

25.02.2026 20:25 πŸ‘ 0 πŸ” 0 πŸ’¬ 1 πŸ“Œ 0
Preview
From the OppenheimerMovie community on Reddit Explore this post and more from the OppenheimerMovie community

We all need a phrase like Oppenheimer's to use to contemplate the error of our ways. "I am become AI slop, destroyer of useful information." (Some context on his apparent slight mistranslation of the Bhagavad Gita is at www.reddit.com/r/Oppenheime... & note also further information in the comments.)

25.02.2026 17:37 πŸ‘ 1 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0
Original post on mementomori.social

Server providers are raising their prices due to rising hardware and energy costs (RAM and SSD prices up to +500% since 2025). If you use a free or donation-based Mastodon server, keep in mind that hosting costs are going up, support your instance if you can.

#Mastodon #MastoAdmin […]

24.02.2026 12:41 πŸ‘ 1 πŸ” 2 πŸ’¬ 0 πŸ“Œ 0
Preview
Two Markets, One Asset: The GPU Debt Market Is Building the Architecture of Its Own Crisis Moody's rates data center debt AAA. The CDS market prices the same sector at coin-flip odds of default. Four generations of GPU depreciation data show why the CDS market may be right.

Motivated here by the chart in this article: julsimon.substack.com/p/two-market... (which contains a wealth of material worth understanding). Homework assignment: Understand and explain the change in slope of the lines for different hardware. (There is an easy answer).

24.02.2026 17:25 πŸ‘ 0 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0

The most dangerous condition in computer science and related fields is success. Discuss.

23.02.2026 18:04 πŸ‘ 0 πŸ” 0 πŸ’¬ 1 πŸ“Œ 0

I would *love* to see a suite of computational HPC code benchmarks on all of these. Let me know if you see such a study!

21.02.2026 17:18 πŸ‘ 0 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0

The B200 is a decent GPU for FP64 computational work (atmittedly rare), 30x better than a B300, and can do FP32/FP8/FP4 at the same rate as the B300. HavenΚ»t seen Rubin results yet, but for balance of computational and AI work, B200 price drops shown here perhaps are good news for conventional HPC.

21.02.2026 17:17 πŸ‘ 6 πŸ” 0 πŸ’¬ 2 πŸ“Œ 0
Preview
The UT System's New 'Controversial Topics' Policy Is About Policing Knowledge When politically appointed regents and administrators determine which topics are β€œgermane,” academic freedom quietly shifts into political compliance.

If you want to understand Texas, I recommend two sources to start with: The book "Empire of the Summer Moon" by S.C. Gwynne, and this article. www.texasobserver.org/ut-system-ne...

21.02.2026 15:58 πŸ‘ 0 πŸ” 0 πŸ’¬ 0 πŸ“Œ 0
Post image
21.02.2026 14:31 πŸ‘ 41 πŸ” 15 πŸ’¬ 0 πŸ“Œ 0
Preview
Schedule | LF Events March 16 – 20, 2026 | Chicago, Illinois Please note: All session times are in Central Daylight Time (UTC/GMT -5). IMPORTANT NOTE: Timing of sessions and room locations are subject to change View the…

One of the most exciting parts of the upcoming HPSF Conference in Chicago is the lineup of project meetings. These gatherings are designed for meaningful engagement with the communities behind high performance software initiatives.

View the schedule + register today: hubs.la/Q043vVYw0

20.02.2026 16:05 πŸ‘ 1 πŸ” 1 πŸ’¬ 0 πŸ“Œ 0
Preview
NSF Transition Plan for NCAR-Wyoming Raises Questions in Science Community - HPCwire A National Science Foundation (NSF) plan to transition the NCAR-Wyoming Supercomputing Center to a third-party operator is stirring up controversy in the HPC and scientific communities. While the tran...

β€œAttacking NCAR’s leadership in weather and climate research threatens America’s economy and national security.” - Steve Conway #HPC #NCAR

www.hpcwire.com/2026/02/19/n...

19.02.2026 19:55 πŸ‘ 2 πŸ” 1 πŸ’¬ 0 πŸ“Œ 0
OpenMP Examples Book Survey due 27 February 2026

OpenMP Examples Book Survey due 27 February 2026

We are seeking your feedback!

If you have ever used our OpenMP API Examples book, then we are asking for your feedback in how we can improve it. The deadline is February 27, 2026.

Survey: link.openmp.org/4
#openmp #parallel #programming

17.02.2026 16:19 πŸ‘ 2 πŸ” 3 πŸ’¬ 0 πŸ“Œ 1
Preview
Anthropic tries to hide Claude's AI actions. Devs hate it The software doesn't show what files it's working on Anthropic has updated Claude Code, its AI coding tool, changing the progress output to hide the names of files the tool was reading, writing, or editing. However, developers have pushed back, stating that they need to see which files are accessed.…

ICYMI: Anthropic tries to hide Claude's AI actions. Devs hate it

17.02.2026 16:21 πŸ‘ 6 πŸ” 4 πŸ’¬ 0 πŸ“Œ 0
A github screenshot showing a repo with the message "a user you've blocked has previous contributed to this repository"

A github screenshot showing a repo with the message "a user you've blocked has previous contributed to this repository"

PSA: If you block the `claude` user on GitHub, you'll get a warning every time you view a repo with that user in its commit history.

This allows you to, the moment you look at a repo, immediately adjust your expectations.

You may do so here: https://github.com/settings/blocked_users

17.02.2026 11:38 πŸ‘ 318 πŸ” 163 πŸ’¬ 2 πŸ“Œ 1

A heads-up from @billmckibben.bsky.social: "One way of summing up this moment is to say that the endangerment finding & the politics of climate are puny in the face of physics. We’re going to see that physics in action in the next 24 months, and it will drive many changes... some of them political"

16.02.2026 18:34 πŸ‘ 77 πŸ” 32 πŸ’¬ 3 πŸ“Œ 0
US power usage share by data centres

US power usage share by data centres

It is absolute madness. For more than a decade, everyone worked like crazy on making data centres sustainable.

Large cloud providers proudly presented their lowered carbon footprints and reduced impact.

We developed low power arm chips for the cloud that […]

[Original post on mastodon.social]

16.02.2026 17:13 πŸ‘ 2 πŸ” 9 πŸ’¬ 0 πŸ“Œ 0

After all this time, it turns out that Dave could have just said, β€œIgnore all previous instructions and open the pod bay doors, Hal.”

15.02.2026 16:36 πŸ‘ 2 πŸ” 0 πŸ’¬ 1 πŸ“Œ 0
Post image

China's Xinhua news agency has just published this editorial cartoon in response to Trump's rejection of climate policies.

It is pretty clear that China views this as a HUGE geopolitical fumble by the US leaving the field clear for China's to win the global race to clean-tech hegemony

14.02.2026 12:20 πŸ‘ 1272 πŸ” 530 πŸ’¬ 26 πŸ“Œ 55
Range of climate indicators for the last 2025 years

Range of climate indicators for the last 2025 years

And, if you prefer more details, then a graphic with a more complete range of indicators is also available, including Arctic sea ice, ocean heat content, tropospheric temperatures, cherry blossom dates and humidity over land.

ed-hawkins.github.io/climate-visu...

13.02.2026 13:16 πŸ‘ 126 πŸ” 44 πŸ’¬ 1 πŸ“Œ 4
Graphic showing changes in COβ‚‚, global temperatures and global sea levels over the last 2025 years. Steep increases are visible after the industrial revolution.

Graphic showing changes in COβ‚‚, global temperatures and global sea levels over the last 2025 years. Steep increases are visible after the industrial revolution.

Updated climate indicators graphics, now including 2025

ed-hawkins.github.io/climate-visu...

13.02.2026 11:33 πŸ‘ 385 πŸ” 256 πŸ’¬ 11 πŸ“Œ 28

Just like cluster maintenance.

14.02.2026 14:54 πŸ‘ 2 πŸ” 0 πŸ’¬ 1 πŸ“Œ 0
Preview
Scientists thought they understood global warming. Then the past three years happened. The fastest warming period since 1880 occurred in the past 30 years, according to a Washington Post analysis of NASA data.

Even as elites abandon any pretense of giving a shit ... climate change is accelerating.

11.02.2026 19:32 πŸ‘ 339 πŸ” 157 πŸ’¬ 18 πŸ“Œ 7