Pocster

They dropped the 512gb m3 because they couldn’t supply the bin in a sensible time frame due to ram shortages .They up’d price of ram increases also + of course Apple ram tax . With m5 I guess there might not be a 512gb option initially . Clearly Apple can weather it better than some suppliers but they are experiencing issues . Ram shortages are set to continue into 2027/28 . So Apple are suffering ram increase costs , ram shortage, high demand for high bin and a new version on route . So bit of a perfect storm tbh . https://www.macrumors.com/2026/04/06/mac-mini-and-mac-studio-long-shipping-delays/ or as chat would say …. You may be right that Apple sold through the amount of high-memory M3 Ultras it planned to build. But that still doesn’t prove the wider memory shortage is irrelevant, and it definitely doesn’t prove Apple will offer M5 substitutions to people waiting on M3 orders. So at the moment the solid bit is just this: high-memory Apple SKUs are hard to get. The rest is theory. Honestly though, he sounds like one of those people who wants to win the mechanism, not discuss the outcome.

Source ? lol ! Apple website - try ordering one . Try getting one m3 / m4 from anywhere - high bin model . It’s true Apple uses a different packaging approach to normal server DIMMs, but that doesn’t make it magically immune to the wider memory supply problem. High-RAM M3 Ultra lead times are already very long, so clearly Apple is not unaffected. My point isn’t that Mac Studio uses identical RAM sticks to AI servers; it’s that the same broader supply chain pressure on high-capacity memory is still likely to hit Apple’s larger bins too. So yes, Apple’s implementation is different. No, that doesn’t prove the current delays are unrelated to the wider memory shortage.

Current delivery on high ram bins is 5 months + The ram shortage is effecting Apple now let alone before new hardware appears .

Global ram demand for hi bins ; nvidia etc . Not easy for Apple that’s sure . 96gb I’ll be ok with . Ram shortage set to continue until 2027 maybe 2028 until new production comes online . Can you all stop using chat and Claude please !

Gave up . M5 ultra who knows when ! Probably wwdc . But try ordering a 256gb m3 ultra now … delivery august! Ram delays crippling everything . So decided to buy a stop gap after a few failed attempts on eBay I.e paid ; then eBay refund as scam / issue . Found stock at believe it or not curry’s - but only the m3 ultra 96gb - so that arrives tomorrow.

Datable ?

Oops yes ! . wtf was I on . Anyway if anyone wants to build a machine to impregnate with an AI hybrid life form I’m more than up for that .

lol yeah I’ve watch his videos . But ! I’m not just using local llm to write code ; the house will be controlled via realtime local llm ! . Ever watched the film with Julie Andrews ? Not the sound of (expletive deleted)ing music !! Demon Seed …

I know ! I know ! But I’m planning a 5 yr project . The thing is I could go less say 128gb or 256gb but …. If that becomes a limitation I’m going to be real pissed . So I decided get something that in theory can’t be limited except by my ability rather than hardware .

Global ram shortages biting badly due to AI server rush . Decided go big or go home ( a man behind back of a pub said that to me once ) . 512gb got to be! ; if available . Roll on April 1st ( Apple 50th ) or wwdc ( early June )

Deployed ! code not an issue with ChatGPT but the supplied analog to digital converter that came with the probe is shite . Completely non stable readings even when not in water ! Quick chat with my best mate and we ordered a different one . With current code worked instantly. The percentage water remaining seems incorrect ( though the calcs were based off the tank height ) . Rather than bring in and reflash ( didn’t add OTT updates ) - might just frig the number by a multiplier to scale it better in home assistant .Doesn’t need to be especially accurate but reasonable .

Finally deployed and wired up . It would seem some speaker connectors are wired wrong way round ( though I checked a zillion times ! ) . But no worries . Not told SWMBO - so that’ll be the ultimate test !

Even I got an SE at the start . Foundation stuff is critical. Even if I posted on here still want an SE to spec and check . So op post details for sure . But get an SE aswell . This isn’t going to be a few hundred quid if you (expletive deleted) up . Once foundations are done you can just wing it like me and not give a (expletive deleted) .

All testing completed! No issues . Full deployment tomorrow!

Got my new probe . Cant wait to stick it in .

The fitting at the wall end is fixed to the spring base - so it is just 1 hole .

This spring gate closer was a tough (expletive deleted) to install . After under a year it’s pulling out the quoin . Assume quoin is quite soft . What fixing do I need so this mofo doesn’t pull out ?

Yeah I meant about the bar keepers friend 😉

Tell me more!

Tested uma8 . Really good . Sensitive enough and direction of speech can be retrieved .

Already ahead of you . Got a usb powered fan to install

Don’t know ; “ Andrews a nonce “ ???

First proper test … A few issues ( software and hardware ) but my ugly newborn lives !

Dr. Frankenstein would be proud of my beast . Work in progress but getting there .

(expletive deleted)ing depth sensor ! It’s shite ! Or at least the beam is so wide it’s useless in a water barrel