Jump to content
Funding the Forum - Thank You ! ×

-rick-

Members
  • Posts

    1004
  • Joined

  • Last visited

  • Days Won

    3

-rick- last won the day on April 14

-rick- had the most liked content!

Personal Information

  • Location
    London

Recent Profile Visitors

The recent visitors block is disabled and is not being shown to other users.

-rick-'s Achievements

Advanced Member

Advanced Member (5/5)

346

Reputation

  1. @Pocster How desperate are you?
  2. Has been a possibility != every heatpump uses this as the solution. Reversing valves were at one point one of the more expensive components in the system so I'm guessing that in the past some manufacturers cheaped out and used alternative means for defrost.
  3. Purely a guess but I suspect this isn't entirely true. True of anything on the market today, likely not going back a few years. Can imagine some early, non-inverter types having some form of resistance electric defrost.
  4. Locally run good models unlocks so many use cases. Buying a software license for a local install would be excellent (like the good old days lol) but it won't be £20. It would break the sketched out business models of the big AI companies though so not expecting it to happen in a huge hurry. All these free models all rely on big expensive training done by researchers. The models are being given out for free as a taster with the hope it generates income in future. Can't see that continuing for long. Facebook already pulled back on open access, Deepseek, Alibaba also pulling back. AI companies also putting fear of god into governments with how powerful their 'best' models are (Mythos) and talking about how they can't let the public access them. Governments are currently on a drive to force operating systems to verify your age before you can use them, chat apps and cloud storage to scan your messages and files for anything illegal, 3d printers to verify you aren't printing something that's banned before you can print. Free and open access to things is getting harder not easier.
  5. Maybe this is true for UK specific monobloc, but heatpumps were primarly invented for cooling. The vast vast majority of heatpumps deployed worldwide do cooling first, heating second. Including air to water ones. Reversible heatpumps have been available for decades globally though only more recently in a mass market form. Once you have a working and efficient heatpump the engineering required for the reversing valve is almost trival. In any case, I think your above statement is not justifiable for any heatpump that was not designed specifically for the UK market.
  6. Agree, Apple will have to pay and therefore so will Apple's customers. They have been suggesting this so far, but the economics are not yet stacking up. OpenAI just pulled back on it's memory order and has recently changed CFO due to accounting issues. They are struggling to complete funding rounds. A lot of the 'orders' placed so far may not get executed (orders in OpenAI case not actually being orders but rather letters of intent). Datacenters that were supposed to be complete by now are delayed to late 2027 (no datacenter to host the server, no point buying the ram for the server). But even if the AI companies are able and willing to pay (and they have datacentre space), so will a lot of Apples customers. Especially those who are buying big memory Studios. These are literally the same pool of people. AI developers spending investor money. Maybe the big labs aren't buying Studios but the smaller startups trying to come up with unique/differentiated products will. This I disagree with. Apple is big enough that the makers will build whatever Apple wants. Micron can produce x amount of wafers per month. Apple commits to buy 20% of the wafers. Micron asks what patterns do you want on them? I was wrong earlier in the thread when I said the dies on the large mac's are not common. Having read more I believe its more that each package/chip has a taller stack of more standard dies (still not anything like the memory used in by Nvidia for their AI chips but likely the same as used in iPhones). So Apple sells 250 million iPhones a year. Both the iPhone and Mac Studio use LPDDR5. One 512GB Mac Studio is equivalent to 64 iPhones (8GB). If Apple is limited by how many wafers it can get, then availability of product is a purely business decision for them, where can they make the most profit. If they can make more profit selling a Mac Studio than selling 64 iPhones they will sell the Mac Studio. I suspect any limitations on Apples supply will be minimal if they are willing to pay because they are a long term customer. Lead times will always be long for Apple. The packaging of these dies onto ICs and then those ICs onto the CPU package adds many steps. So if Apple misjudges quantities it takes a lot of time to correct.
  7. Self build/custom homes are a different planet to bog standard UK housing, no? It's a situation where the customer cares about the details (to varying degrees but still much more than developers).
  8. My main argument is that Apple places bigger orders than the AI companies and is a well established long term customer. The memory manufacturers will make whatever their customers are willing to pay for. If Apple wants it and is willing to pay, it will get.
  9. It's entirely possible there will be some impact on Apple this WWDC. Their pre-orders from last year all being consumed and them having to pay higher prices from now on. My argument was always that the shortages we've been witnessing so far have been baked in since before the RAM crisis started and therefore not directly caused by it (just people buying more high memory Apple products). Apple should be able to get supply they are big enough and have high enough margins (can can always raise the price of the Mac Studio*). Price is the big question. *The people buying high ram Mac Studios almost all want them for AI and therefore will pay over the odds for them if they can get them.
  10. Seems to echo my earlier thoughts: https://www.wsj.com/tech/personal-tech/apple-mac-mini-supply-3e7a7509
  11. My feeling here would be to run the incoming to where the EV charger/solar is and then run ducting from there into the house with one mains cable for the whole house (would need additional ducting for comms). Looks like you could run almost the whole route outside the footprint of the house to allow easy access later. But there might be an electrical/building control reason why you shouldn't do that?
  12. If you think ahead and make it possible to remove the back boxes after installation (cables routed through the back rather than side holes, plasterboard cut out big enough, screws positioned with this in mind) then each socket gives access to the cableway. Might be a pain to remove and put back but less of a pain than making/patching holes. Corners are a big concern so likely need to plan for sockets on both walls fairly close to corners. Having well thought out routes between each room and a centralish location with conduit/accessible tray is another element to make this feasible.
  13. It's basically an iPhone 16 in a laptop case. Apart from the limited USB ports it appears to be a significantly better buy than similar priced windows laptops.
  14. Thought Apple tended to be quite efficient memory wise. The Macbook Neo with only 8GB is getting rave reviews and appears perfectly capable of average user usage (not specialised apps though).
×
×
  • Create New...