I love Computer Science. I made educational videos, btw.
Joined December 2023
- Tweets168
- Following49
- Followers5.9K
- Likes160
Pi Zero 2 W is finally back in stock at Micro Center after nearly six months! Limited to one per household, though, but still better than nothing.
2
448
CoreDumped retweeted
hey sam, your mclaren is only used 5% of the time. this is a terrible allocation of scarce resources. have you considered taking the bus to work
sam: local inference is a terrible allocation of scarce resources
“a lot of people want local inference, and somehow they assume that the problem is solved now because we have open-source models, we're going to be able to run them on our computers, and so we're not going to need all the stuff in the data centre
but from a global allocation perspective, actually it's quite terrible to have everybody, like, have a bunch of the world's limited supply of DRAM sitting in their own computer and being used five percent of the time
you want maximum utilisation in the cloud if you're trying to deal with this very supply-constrained environment
somehow many people's intuition is the opposite of that”
Okay, f*ck it. I’m going to be the one to say it: A LOT OF RED/VAPORWARE FLAGS WITH THIS ONE.
First: There’s no way this is profitable. I simply don’t see how they’re making money on this.
Second: Performance won’t be that great:
>They’re selling this as if it has 88 GB of unified memory, when in reality only the 24 GB of VRAM will be efficient for inference. DDR5 with that processor will offer around 70 GB/s of bandwidth. If a model doesn’t fit entirely in VRAM, you’re going to have a hard time. So be prepared to run smaller models under 35B parameters or heavily quantized larger ones.
>An RTX 3090 could outperform this thing for roughly a third of the price. It has more than twice the memory bandwidth (936 GB/s versus 432 GB/s), more CUDA cores, and better potential for scaling thanks to NVLink support.
The only area where that RTX Pro 4000 wins is power consumption. But if that’s the priority… M-series Macs do exist.
Introducing Core, The Personal Al Computer
The most powerful technology humanity has built should belong to you
Order today: ghost.ai
Being a gamer hurts so much these days. The RAM situation is so bad that we’re calling this a "deal"
3
30
1,785
BTW, if every software engineer were required to take just one year of computer science, we would live in a world where a system with 4 GB of RAM would be perfectly fine for most purposes.
I think it's because universities still do Computer Science. We don't need nowhere near that many computer scientists, and yet software engineering is taught almost nowhere.
It makes total sense for CS - CS is basically the math, proof, and formal logic land of computing - where "how" it works doesn't matter, as you basically operate on imaginary math-world computers anyway.
A computer scientist is not concerned with memory management, scheduling, and whatnot. They're working out a sort algorithm, an optimal concurrency model, or proving algorithms work. Maybe they will be concerned with scheduling if their work is strictly relating to scheduling. And even then, it's gonna be a math-land ideal scheduler, not any real situation.
That's why I kinda find CS to be a mostly useless degree nowadays. Sure, it's useful if you want to be a scientist, but a software developer is the exact opposite of a scientist, they are an engineer, and 99% of people doing CS want to be the engineer.
Tell me you have no f*cking idea what CS is about without telling me you have no f*cking idea what CS is about.
I think it's because universities still do Computer Science. We don't need nowhere near that many computer scientists, and yet software engineering is taught almost nowhere.
It makes total sense for CS - CS is basically the math, proof, and formal logic land of computing - where "how" it works doesn't matter, as you basically operate on imaginary math-world computers anyway.
A computer scientist is not concerned with memory management, scheduling, and whatnot. They're working out a sort algorithm, an optimal concurrency model, or proving algorithms work. Maybe they will be concerned with scheduling if their work is strictly relating to scheduling. And even then, it's gonna be a math-land ideal scheduler, not any real situation.
That's why I kinda find CS to be a mostly useless degree nowadays. Sure, it's useful if you want to be a scientist, but a software developer is the exact opposite of a scientist, they are an engineer, and 99% of people doing CS want to be the engineer.
Nobody is going to convince me this isn’t, at least partly, about keeping more regular people from getting access to these and realizing how capable some local AI models actually are.
🚨 OpenAI are now contributing to the Mac shortage
OpenAI have apparently bought tens of thousands of Mac minis + Mac Studios for RL and computer use training
Anthropic are mass renting Mac minis through AWS too
At this point every decent piece of compute is just going to get eaten by AI labs, looks like a good time to buy
Java aka C++
Programming languages used in popular games .
Readers added context they thought people might want to know
Minecraft Java Edition is written in Java, not C++. minecraft.wiki/w/Java_Edition en.wikipedia.org/wiki/Minecraft
2
23
2,197
The crazy part is they’re almost always on sale. Time it right and you can grab three for the price of one first-party controller.
This Post is from an account that no longer exists. Learn more
1
24
3,345
I have my own conspiracy theory that Sam Altman deliberately provoked a RAM shortage to slow down local LLM development. If that’s the case… it’s not working.
Quick spoiler: it won’t happen. Microsoft already went through this with the handheld market when the Steam Deck launched. Their response? A “handheld version” of Windows with a laggy UI, the same bloatware, and all the usual telemetry running in the background.
I hope the MacBook Neo finally humbles Microsoft.
Windows has progressively gotten worse and worse over the years, just extreme levels of bloat and no significant improvement in terms of user experience.
Somebody needs to give Microsoft a huge shock, force them to shut everything down and rebuild Windows from scratch.
CoreDumped retweeted
What is crazy is that the RTX 5070ti went up to $1,000+ and people are still purchasing it over the AMD 9070 and XT.
CoreDumped retweeted
Unironically, I don't know anyone who achieved anything meaningful with OpenClaw
NVIDIA’s CEO says OpenClaw did in 3 weeks what Linux took 30 years to achieve
"OpenClaw is probably the single most important software release ever. If you look at its adoption, Linux took some 30 years to reach this level. OpenClaw, in what is it, 3 weeks, has now surpassed Linux. It is now the single most downloaded open-source software in history, and it took just 3 weeks."
