Mini mal

$INTC can increase as much SRAM density. Groq can build inference chips with $NVDA. But can they force undo the enshittification of the software stack ? How far can hardware efficiency go to cover for the the inefficiencies of the software bloat?

@alexocheema You are comparing a garage built go kart with a le mans Ferrari. Not a fair comparison

My M4 Max MacBook gets 3,756,165 tok/sec in pure C, compared to ~50,000 tok/sec with the FPGA. Try it yourself: github.com/AlexCheema/tal…

Have you wondered that they don't spend enough time on verification and testing as they used to? They have transitioned from a 3 product per year company to dozens per year, many of them custom builds for large clients. I have seen this happen many times in the semiconductor industry. Shorter design-to-product end up in all design stages getting compressed, including verification and testing.

I used to love Nvidia... that is, until I started to use more of their high-end, more expensive products. The more I bought them and tried to use them, the more I saw Nvidia shipped buggy, faulty products that simply don't live up to the hype (or even the claimed specifications). And when I tried to have a $9000+ GPU replaced under warranty, well within the claimed warranty period, Nvidia refused. They stopped responding and took zero responsibility for their faulty product. Then I checked the pricing for their newest "Spark Station" products (based on the GB 300 chipset) shipping now. One systems integrator quoted me $180,000 for a single system. And I knew there would be virtually zero driver support and no warranty support either, based on my previous experience. So I said no. I used to be a fan of Jensen Huang (he's from Taiwan, where I lived for a few years), but now I increasingly see him as a sales hypester pushing product fantasies that don't ship on time, that don't perform to spec, and that simply aren't worth the money. It all seems like vaporware gaslighting to pump up Nvidia's stock price now. AMD's Strix Halo platform, I found, runs LLM inference models (even very large ones) at very close to the same speed as NVIDIA hardware, but for typically less than half the price. And they use a lot less power, too. Apple Mac hardware has much faster unified RAM and is becoming increasingly popular for AI projects, with long wait times due to the surge in orders from customers. Meanwhile, Nvidia keeps promising revolutionary chipsets that never ship on time and that utterly lack driver support or warranty support when they do actually ship. Just a couple of months ago, they DOUBLED the prices of their 5090 cards, raising them from around $2500 to $5000 with no real justification that made any sense. They were just PRICE GOUGING their customers because they could get away with it. If you want a $5,000 brick on your desk, buy an Nvidia DGX Spark. And hope it doesn't set your desk on fire, because it has horrifically bad thermal management and has been known to literally melt many materials. The AMD Strix Halo platform (I own some mini-PC boxes running it) run cool and quiet, with a fraction of the energy usage. Like I said, I used to be a fan of Nvidia. But something changed with them. It seems they abandoned quality control. They became obsessed with higher stock prices instead of higher quality products. They abandoned their customers and price gouged them wherever they could. Using Nvidia products feels like you're fighting against a corporation that hates you. That's why I'm done with Nvidia. My company typically spends several hundred thousand dollars a year on hardware, and I run a mini data center with 48 workstations and a few proper rackmount servers, serving millions of users with high-quality AI content creation, AI-enhanced research engines, infographic image generation and more. From now on, I'm ditching Nvidia and looking elsewhere. Intel, AMD, Apple, TensTorrent... I'll take my AI business to a company that actually cares about its quality, its warranties and its customers.

Does anyone know of a lighter alternative of APB to access control and status registers on a low speed peripheral ? APB is very light, but I need something even lighter.

This is definitely a low hanging fruit. All it requires is a pissed off superdev with dozen LLM agents. Take the source code, point those agents at it. Start refactoring. These big corporate gaming studios will be decimated in 8-12 months tops.

@oshillex Inflation.

Can someone give me a legitimate reason as to why game compression no longer exists

@yacineMTB An AGI that predicts the next ragebait token from this account.

@yacineMTB At this point someone should make a yacine ragebait tracker

Everything should be statically linked. Dynamic linking is idiotic

@KyleMorgenstein You put three emojis in your title card. You don't get to preach huh

LinkedIn gorilla marketing tactics for the first equation they teach you in undergrad robotics courses. /tau = K * I What are we doing man.

@1a1n1d1y It ain't paying my bills it is a time worthy occupation while my agents fire up those tokens.

you wanna run a whole god damn tech startup? be funny on twitter

@weswinder The "perception of value" is different from "value" itself. Your perception is a small weight. The market value will be the weighted sum of all perceived values.

i’m sorry, but anthropic and openai are not $1T companies

There is an incentive to add slop wherever possible, whichever market allows it. > It happened to food (processed sugar replaced real fruits, palm oil replaced natural fats). > It happened to cars (premium leather replaced by cheap substitutes, rubber quality became poor, solid metal parts replaced with cheap plastic) > happened to fuel You think they won't do it to mere AI tokens? What is so shocking about this

It's incredible how I was basically 5 months early on this. Which is like an entire decade in AI time

@splinedrive MMU doing the heavy lifting. Nice. Have you measured a slowdown factor (how many real world seconds it takes to run a CPU second) ?

Fuck FPGA, nobody needs it...Verilator on x86 rules. KianV Stealth next generation performance SoC

@yacineMTB Not dissing on the founders. But on the system that makes everything boring.

Funded startups follow scripts and playbooks because they are funded by culturally similar sources. This guy got his Series B. The agency allocated a marketing consultancy budget. The consultant had a script and a schedule and a camera. He just acted it out. Most startups are pretending to be different while being all the same. Look at everything they do.

I cannot stand this uninspired bullshit ad shooting style. You are a startup founder. Take out a phone and record yourself talking like a normal human being. You are literally just a person. You are ctrl cing the "personal" directing why don't you actually be personal??

You can never tell whether this guy is serious or ragebaiting.

@yacineMTB Skill issue

Man. For some reason codex started performing really poorly on programming my goblin roguelite indie game

@devadam88 Boy you are in luck

@gatelevelanon only real estate web developers

London based developers, reveal yourselves! 🇬🇧

If ISA ESOP NDA documents mean anything to you, GitHub is irrelevant. This is also unfortunate because it creates a gap between what you can show to potential future employers/partners as proof of work. They have to take your word for it.

@nilscmr Sent in less than 24 hours. Few minutes remaining.

@nilscmr Check your DM here.

CPUs suck. We're building a new general-purpose chip that scales to thousands of cores while being more energy-efficient. We're hiring hardware design engineers, consider joining us tendrils.co/jobs What we do differently ...