££££ prices shocked me today

LongLensPhotography

Th..th..that's all folks!
Messages
18,809
Name
LongLensPhotography
Edit My Images
No
Wow. It looks like you have to pay just over £4000 one way or another if you want a fairly high end truly modern machine. Macbook pro 16 with sensible specs. The new mac studio. AMD rzyn Halo. The nvdia Asus G10 which is only for LLMs. All of them land at 4k+.

Luckily my PC is still OK for another year or two but it does limit what I can do with it. Davinci resolves struggles a little, local LLMs are restricted to really dumb models. 32GB RAM is a little limiting and I wish I had specced 64 but really now this is all about VRAM or unified memory. The latter is much preferable. I could drop a 16GB GPU at exorbitant cost, 24 or 32 is multiple 4 figures... that would fix the resolve. 32gb is not that much for LLMs. 128GB is really the low end here.
I will just wait for now. But what would you do if things broke and you had to buy?
 
RAM, SSDs and HDDs are all massively inflated at the moment and sadly likely to be for another couple of years. It's a bad time to be after a new machine.
 
At the moment it's looking like it's going to be a long time if ever that prices will come back down to the levels they were before now companies are seeing how much more people will pay for the parts, There's also considerable damage being down to other hardware like motherboards because there's been a drop in demand due to people not building as many PCs due to the high prices of components.

I'm just hoping no parts fail and I'll need to weigh up paying well over the odds to get a machine up and running again if something does go.
 
I'm still using my base spec M1 Mac Mini; I don't do anything with video really but it does everything I need for music and photo editing. If it died I'd probably look at an M4 model, or maybe even a Macbook Air.
 
You are in for a long wait as long as this AI bull is around, prices will not come down unless AI falls flat on its face which I hope happens.

Not to long I fear, the massive investment in data centers and consumption of memory and "chips" is horrific for those businesses, We all have learned that technology advances massively every two years or so, to invest a few trillion in a technology that will have to be replaced in a couple of years is mind bendingly irrational IMO, The big corporations will have to have "the latest and greatest" to stay competitive, or so they will believe. I will sit and wait until the second hand stuff floods the market, they will be giving it away.
 
Not to long I fear, the massive investment in data centers and consumption of memory and "chips" is horrific for those businesses, We all have learned that technology advances massively every two years or so, to invest a few trillion in a technology that will have to be replaced in a couple of years is mind bendingly irrational IMO, The big corporations will have to have "the latest and greatest" to stay competitive, or so they will believe. I will sit and wait until the second hand stuff floods the market, they will be giving it away.
Just in time for Starmers Data Centres to have cost the taxpayers maximum £'s and the need to disappear
 
Luckily my PC is still OK for another year or two but it does limit what I can do with it. Davinci resolves struggles a little, local LLMs are restricted to really dumb models. 32GB RAM is a little limiting and I wish I had specced 64 but really now this is all about VRAM or unified memory. The latter is much preferable. I could drop a 16GB GPU at exorbitant cost, 24 or 32 is multiple 4 figures... that would fix the resolve. 32gb is not that much for LLMs. 128GB is really the low end here.
I will just wait for now. But what would you do if things broke and you had to buy?

From your own post, not using LLMs would reduce your memory requirements and costs enormously! :D

If everyone using AI adopted the the same approach (including industrial users) then maybe we could get back to some sane memory and storage prices
 
Last edited:
I'm looking at replacing my M1Mac Mini now because it has 256Gb storage and it's nearly full - I had to delete some apps to be able to run Time Machine. So I was seeing these prices, and Apple are now showing the M5 and M6 chips/machines, starting at £1800 or £5500. I'll probably go for a Mac Studio or a high spec M5, to future proof myself.

However, LLM's: why are you running LLM's on a client side machine? I thought the models churned data on big servers and returned results to a client that could be fairly modest? That's the whole point behind subscription services such as Adobe. Although I guess if you need your input data to be kept private, that is the price you pay for not letting the commercial models include your data.
 
What on earth are you doing with it - my Macbook is 5 years old (M1) and works fine with LR and imovie for video. Why need 128 RAM?
 
However, LLM's: why are you running LLM's on a client side machine? I thought the models churned data on big servers and returned results to a client that could be fairly modest? That's the whole point behind subscription services such as Adobe. Although I guess if you need your input data to be kept private, that is the price you pay for not letting the commercial models include your data.
Data privacy is certainly one concern, and cost of access is potentially another. The big 4 are pushing everyone on $200 monthly plans already; the basic £20 claude just keeps getting out of allowance so darn quickly like I want to drop it entirely. Maybe a combination of last chatgpt and latest china models can replace it.

I feel like 2027-8 will be when the local AI really starts to take off. qwen 3.8 is supposedly very good, almost the top tier level and you could run in on such AI desktop. If you want to run an AI adjacent or AI first business you may want one. If photography work is thin thats all many of us can do unless you want to do like garbage collections with a PhD.
 
Last edited:
Your looking at least until 2029 maybe 2030 before prices come down when the new Chinese fabrication lines come online.
 
Back
Top