Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

DRAM pricing is killing the hobbyist SBC market

jeffgeerling.com

Post reply on HN
Print view View original

121–130 of 592 posts

Re: DRAM pricing is killing the hobbyist SBC market

#121

Post by Forgeties79 »

JollySharp0 wrote: post #51
zozbot234 wrote: post #47

Earlier quoted context omitted.

> People are quick to forget that during COVID prices were high for SBCs because of supply chain issues. inb4 AI has the same supply chain effects as a worldwide pandemic. I guess those AI doomers that talked about it being the end of the world had it right!

Doomers IMO are just click baiting. There is a saying that is often trotted out my economists "That the cure for high prices, is high prices". There is a consumer market and business need for DRAM outside of AI. Someone will fulfil the need as there is a high incentive to. It just going to take a bit of time for this to happen. My equipment is going to be fine for another few years. So I am going to just hang tight a…

So are AI evangelists to be fair.

Re: DRAM pricing is killing the hobbyist SBC market

#122

Post by Quothling »

tempest_ wrote: post #36
rat9988 wrote: post #8

Earlier quoted context omitted.

The art is not lost, just not funded. Feel free to fund the programmers for your own software projects.

It isnt lost but it also isnt a common skill set in programmers any more. Most programmers are JS web devs writing client side code or server side CRUD. I would guess < 10% of programmers writing code today get perf / valgrind out on the regular. I know I dont.

You don't have to go that deep. 99% of the time our analytics or risk management teams have some really memory inefficient Python and they want me to write them one of our "magic C things" it turns out to be fixable by replacing their in-memory iterations with a generator.

Re: DRAM pricing is killing the hobbyist SBC market

#123

Post by BizarroLand »

ticulatedspline wrote: post #111
Aurornis wrote: post #73

The extreme DRAM market has had an unexpected side effect of triggering a lot of panic buying. I know several people who delayed PC upgrades for years but then panic bought new systems in this market. The trigger was seeing all of the "It's only going to get worse" and "This is the end of personal computing" headlines. They're already regretting spending so much now that prices have started to tick downward. I keep t…

wasn't "panic" buy but I built a new comp early 2025, cuz at worst case would be complete supply crash and at best case it was going to be more expensive. Def don't regret doing that, though I regret not springing for the extra RAM.

Same. I got 64gb for my new build the day this whole thing started but I kind of wish I had gotten 128 just for bragging rights.

Re: DRAM pricing is killing the hobbyist SBC market

#124

Post by icedchai »

dangus wrote: post #10
jonathantf2 wrote: post #7

DRAM pricing is killing the everything market. We just had a vendor uplift our quote 50% per unit for some machines because of a mix of memory + supply chain issues.

That’s strange, there aren’t wider market supply chain issues outside of DRAM. Maybe your vendor is just throwing excuses around.

I assume this is sarcasm.

Re: DRAM pricing is killing the hobbyist SBC market

#126

Post by megous »

throwaway85825 wrote: post #82
IncreasePosts wrote: post #49

Earlier quoted context omitted.

Why would we need rust, if the AI can just write really good code in C that doesn't exhibit any of the issues that rust protects you from?

The Rust ecosystem and build tools are much easier to use than C. The value of a language isn't just syntax.

LLMs are great at C, probably because C is historically the most popular language in the world, by far. It only declined slightly very recently. But there's insane amount of code written in it.

Re: DRAM pricing is killing the hobbyist SBC market

#127

Post by josephg »

jokoon wrote: post #6

it's probably time to call those old retired programmers to ask them how to reduce software memory footprint or to teach that again

Taking a big, complex, already well optimised program like Chrome or the linux kernel and optimising the memory footprint is hard. But 90% of programs are just crappy web apps that nobody has even bothered to optimise at all. (Sometimes wrapped in electron or something.)

If you go look, you often discover that 90% of the requests are useless, or at least could be combined. That 60% of bandwidth is used up by 3 high res images which get displayed at 30x30 pixels. That CPU performance is dominated by some rubbish code that populates an array of a million items every call, then looks up 1 element then throws the whole thing away, only to regenerate the exact same list again a few microseconds later.

We have plenty of RAM. In absolute terms, 8gb of ram in the macbook neo is 8 billion bytes. 64 billion ones and zeros. You don't need rocket science to make a CRUD app that runs well with that much ram.

Computers don't get slower over time. If we were merely as lazy with computing resources as programmers 10 years ago, most programs would scream on modern hardware.

Re: DRAM pricing is killing the hobbyist SBC market

#128

Post by pixl97 »

throwaway85825 wrote: post #91
Permit wrote: post #16

Earlier quoted context omitted.

Can you elaborate on the collusion aspect? Is the implication that OpenAI and Anthropic are coordinating their purchases in such a way that they target the hobbyist market? What’s the collusion angle here?

OpenAI signed letters of intent for 40% of the DRAM supply because they have no moat and want to starve their competition.

Only works so long as you eventually pay up... well unless the manufacturers make too much this way. That said are there some Chinese manufacturers that aren't part of the cabal and could undercut them?

Re: DRAM pricing is killing the hobbyist SBC market

#130

Post by bluescrn »

throwaway85825 wrote: post #77
esskay wrote: post #61

The SBC markets been on life support for a long time. Youtubers making videos about them don't seem to grasp that and keep pumping out reviews and projects like its still 2019. The pi specifically has plummeted in popularity and for most use cases they just aren't a cost effective option when second hand micro pcs are dirt cheap and vastly more capable.

PCs don't have GPIO. They're different markets and the desktop replacement never materialized.

For most projects using GPIO, a <$10 ESP32 board or Arduino clone will suffice
Post reply on HN