Jerrill Johnson ๐บ๐ธ๐ฝ ๋ฆฌํธ์ํจ
@RapidResponse47 Mr. President, we are just getting started ๐บ๐ธ
English
Jerrill Johnson ๐บ๐ธ๐ฝ
811 posts
Jerrill Johnson ๐บ๐ธ๐ฝ ๋ฆฌํธ์ํจ
Jerrill Johnson ๐บ๐ธ๐ฝ ๋ฆฌํธ์ํจ
๐ฅ Peter Bence turning Queenโs Bohemian Rhapsody into absolute classical piano FIRE!
When a world-class concert pianist gets asked for the ultimate rock operaโฆ this is what happens. Insane speed, perfect dynamics, and pure genius on the keys ๐น๐
English
Jerrill Johnson ๐บ๐ธ๐ฝ ๋ฆฌํธ์ํจ
Iโm not asking you, Iโm telling you.
Take a moment and listen to this 81 second response from Victor Glover after being asked if he had any thoughts leading up to Easter.
I donโt quite think it can be overstated how perfect this crew is for the job.
English
Jerrill Johnson ๐บ๐ธ๐ฝ ๋ฆฌํธ์ํจ
Just 1,500 feet (457 m) from the Artemis II launch, we deployed our special high-resolution, slow-motion camera to record the blast-off in jaw-dropping detail at 2,000 frames per second. Find out how we did it: on.natgeo.com/0BXGmi
English
Jerrill Johnson ๐บ๐ธ๐ฝ ๋ฆฌํธ์ํจ
50 years ago @Apple was founded, today astronauts are using iPhones on their way to the moon
English
Jerrill Johnson ๐บ๐ธ๐ฝ ๋ฆฌํธ์ํจ
Liftoff.
The Artemis II mission launched from @NASAKennedy at 6:35pm ET (2235 UTC), propelling four astronauts on a journey around the Moon.
Artemis II will pave the way for future Moon landings, as well as the next giant leap โ astronauts on Mars.
English
Jerrill Johnson ๐บ๐ธ๐ฝ ๋ฆฌํธ์ํจ
Jerrill Johnson ๐บ๐ธ๐ฝ ๋ฆฌํธ์ํจ
I'm honestly SHOCKED at how the general public has NO IDEA Artemis II is taking humans out to the moon and will be the furthest humans have ever flown. Every non-space nerd I've talked to has no idea. WE GOTTA GET PEOPLE STOKED!!!! THESE FOUR HUMANS ARE FLYING TO THE MOON!!!
English
@digitalix I am curious how per-session tokens/s scales with concurrent sessions on the quad DGX Spark cluster. Is the cluster a better agent swarm host than a Mac Studio?
English
what do you want to know about the DGX Spark that might still be unknown?
Like, are there features that you thought it has, but not sure. What questions do you want answered?
Iโm especially looking for questions about real performance, clustering, model support, software, and whether this makes sense vs a traditional GPU setup.
Drop them below.
English
@claudeai I just installed Claude for Windows and there is continuous high CPU usage when idle. Quitting the app via the system tray doesnโt terminate the offending cowork-svc.exe either. Please be better neighbors on my machine in an update soon. Love your work otherwise. Thank you.
English
@DaveShapi Itโs honestly just because it landed in Trumpโs second term. If itโs anywhere close to Trump it must be bad. Thatโs about all the rigor the left seems to apply to anything nowadays. They have abdicated their role in shaping the future.
English
Here's what I sincerely don't get.
Star Trek has always been coded as extremely left. Liberal. Woke. Whatever you want to call it.
And now that we're creating Star Trek like technology, which is explicitly required to create a Star Trek like post-capitalist society, all the leftists are up in arms about AI.
Unless I'm missing something.
English
@digitalix I hope you do a video showing concurrency benefits! It would be nice to know that there are some before I take out a second mortgage to get my agent swarm game on.
English
It started with 4, but it escalated from there lol
Alex Ziskind@digitalix
New video. My attempts are clustering DGX Sparks (and giveaway)
English
@digitalix Please run some concurrency tests for your YT video! Thank you! x.com/MrCatid/statusโฆ
catid@MrCatid
Cost: $35k for 12 tokens/second on a 32B model at FP8 with 1120W of power. Tons of cables, setup, networking. Alternatively: One $9k GPU (RTX Pro 6000 MaxQ): 22 tokens/second for one request at ~500W of power. Fits in the desktop you already have.
English
@MrCatid What about concurrent requests? What does the RTX 6000 Max Q max out at in tokens/s for this model?
English
Cost: $35k for 12 tokens/second on a 32B model at FP8 with 1120W of power. Tons of cables, setup, networking.
Alternatively: One $9k GPU (RTX Pro 6000 MaxQ): 22 tokens/second for one request at ~500W of power. Fits in the desktop you already have.
Alex Ziskind@digitalix
8 sparks cluster running
English
@Ananyaedits @MrVladimirX @ziglang @rustlang Are you running one model across 6 nodes clustered? Or are the nodes just in parallel for 6x performance for models that would fit on one? Or something in between? I'm considering a quad DGX Spark cluster using QSFP switches, but I'm not seeing a lot of people trying this?
English
@MrVladimirX @ziglang @rustlang Currently I am working on a build to run open source llms locally
Going with 6 DGX spark
I will be using it as a Linux server which I can run remotely
After that.
I need to build a custom RAG
I have seen one of my friends using the same setup & it boosted his performance
GIF
English
Jerrill Johnson ๐บ๐ธ๐ฝ ๋ฆฌํธ์ํจ
Jerrill Johnson ๐บ๐ธ๐ฝ ๋ฆฌํธ์ํจ
Jerrill Johnson ๐บ๐ธ๐ฝ ๋ฆฌํธ์ํจ
Remember that @awscloud + @Cloudflare outage from earlier?
Yeahโฆ this meme explains it perfectly.
Turns out the entire internet is basically a Jenga tower built on unpaid open-source devs, left-pad, DNS, and 3 C programmers who still refuse to use dynamic arrays.
One tiny wobble and poof half the web forgot how to internet ๐๐คฆโโ๏ธ
English
I recall a time long, long ago where I could read a 600-page microcontroller datasheet and that be all I needed for nice, steady 18-month project. I didn't appreciate that as much as I should have at the time. I probably need to appreciate the now a lot more- day-by-day vesperance.
English