Trashbytes wrote on 2026-08-08, 03:10:

The market will self correct but it wont be in the direction we want, itll be a correction where they will simply sell the spare […]
Show full quote

ElectroSoldier wrote on 2026-08-08, 01:52:

Yeah but I see that as a good thing. […]
Show full quote

Yeah but I see that as a good thing.

I mean yeah its bad if youre in the market for a new PC or new PC parts but lets be honest most of us on here are more interested in older PC parts and they are always subject to weird pricing by their nature (being old).

On the other side this bubble we know is there and growing is going to burst big when it does burst.
And we know its going to burst because the sites that offer up these pointless AI tools, of which there are now millions cant find a way to make money from them.
Theyre all free! Or prices to use them are unsustainable, the fact google had to change its service relative to its pricing shows that.
There are other sites that are getting ready to restructure their pricing/services, so we are only at the stage where these sites are starting to see the pinch, but not feeling it yet because they can adjust the prices to fix the problems.

I shell out close to £50 a month for AI services but if it goes over that I wont bother because its just not worth it to me. And Im a fully sold up member of the AI club! And I WOULDNT pay!

So when this bubble bursts its going to dump so much hardware into the market it will be painful to watch as these guys start to wonder what the bottom of the Serpentine looks like in real terms!

This bubble cant burst until the supply catches up with demand.
Because once they get all the hardware they think they need they will start paying for it, once they do they will need to rinse money out of subscribers, but the services are not worth that much so the companies wont be able to meet their debts... Because its all borrowed money.
Then and only then can the bubble burst.

Thats exactly what happed in the 90s with all the dot coms.
Dead dot com hardware was hitting the market for years after that, and it was great buying up server hardware for buttons.
And then the hardware manufacturers will pivot to consumers to fill the gap their freed up capacity gives them. Because they will want to make money and consumers will provide the cash they need.

Its going to be crap for a few years...
2028 at least. And prices will only get higher as people get desperate, but most of us soldier on with old hardware anyway.

With any luck some of these stupid hardware assemblers that make a living on youtube die off too.

The market will self correct but it wont be in the direction we want, itll be a correction where they will simply sell the spare compute rather than sell the hardware, the Data Centers are not going anywhere even if AI dies and they will need to find a way to sell that compute. So the profits from selling hardware to the DCs will still be there, it wont be as lucrative as it is now but it will still be far more substantial than selling it to the consumer plebs. nVidia will simply pivot fully into selling compute via subscription and streaming services, Amazon and Microsoft will follow suit and the DIY space will ...well it simply wont exist as it does now and it will become a high roller luxury space where only the people with large disposable incomes will be able to afford the hardware.

People will try their best to sell me on .. "that's bullshit itll never happen" ...but we have been watching it happen the last few years like a frog in a pot of boiling water as more and more services and features have been moved to the compute cloud and chained to a monthly/annual subscription fee. So even if you own the hardware you will still be forced into their online compute system for a monthly fee of $19.99 for 100 hours of compute.

So the bubble bursting wont be the "FIX" you think it is, the fabs making the core hardware don't want lower pricing and nothing will force them to return to pre AI bubble pricing either.

The totality of it all is, THEY DONT WANT US OWNING COMPUTE, or a level of compute that can run local AI that can compete with theirs and they will 100% make sure that even if it all crashes and burns DIY PCs are going with them.

Yes that is true for the hyperscalers Meta, OpenAI, Google and Microsoft etc, they will survive the burst simply because of economies of scale will play in their favour but the smaller companies who are dumping tons of cash into hardware to "keep up" will have no choice, the receivers will sell off what they can to recoup the investment.
The hyperscalers will find a way to harness the power they have built into their datacentres. Which wont be a bad thing. I mean look at all the bad sites we had before the dot com bubble burst and how many decent ones came out of it and get better as a result of it.

As to computing as a service and hardware being only for those with disposable income.
It sounds good on the surface, and in some way I do think it will come out that way however it wont take over the computer market as we know it.
There are so many things against it happening and so few for it.
Renting compute would be like VMs in a corp enviroment right?
I mean CraftComputing on youtube has been building "servers" on that basis for years but unwittingly proven so many times over multiple systems why it wont work.
First is that the user likes to decide.
In a corperate enviroment the corp decides but in the real world the user does and 9 times out of 10 the user will choose to own rather than rent nothing. Think GeForce Now to see what I mean.
Renting compute power isnt going to be cheap, owning your own PC will not only be cheaper it will be much cheaper, even if the datacenters are already there doing nothing. AI is a thing its use will survive the bubble bursting, it will have a more defined purpose thats all.
Yes you hav watched more services move to cloud computing but you dont see them failing left right and center and theres a reason for that, and those companies are hoping AI will solve the problem for them.
Microsoft found this out with their billing structure. There is only so much people are willing to pay before they walk away. And this is one of the main problems with AI.
That companies will struggle to bill for it. Not that they wont bill for services but they will have to bill so much to make it a viable business model that people will not be willing to pay that much for it.

As to fixing the pricing of haardware, such as RAM.
Well that isnt going to be a problem the manufacturers cause. It will be the reseller who keep the prices high. They will see how much profit can be made from RAM.
Like you see now on ebay. How many people are selling old sticks of DDR5 on ebay they bought for buttons a year or so ago and now theyre selling for a 500% profit. Thats not the manuacturers causing those prices to go up. People are cashing in on the hype. Thats what will keep prices high after the bubble bursts. But the manufacturers will have no choice but to pivot to selling consumer hardware, otherwise what are they making and selling?
But no the prices will not return to the prices they were 18 months ago, but then you will have had at least 4 years of inlation by then so that idea is just stupid anyway. But they will level out from where they are now. You wont have people selling RAM on ebay for £300 when they spend £150 on it last year. It will go back to being a depreciating asset. Thats how you will know if you are right or I am.

If companies build new fabs for RAM (China sure is) they will flood the market to pay for their investment (and to hurt their competition) so prices will go down, but only after those who sat out upgrading buy what they need.

RAM is a commodity and prices swing in both directions ever so many years because of supply and demand. I do wonder if people upgrading their phones every so many years is still going to happen because that sucks up RAM and flash.

Unknown_K wrote on 2026-08-08, 07:13:

If companies build new fabs for RAM (China sure is) they will flood the market to pay for their investment (and to hurt their competition) so prices will go down, but only after those who sat out upgrading buy what they need.

RAM is a commodity and prices swing in both directions ever so many years because of supply and demand. I do wonder if people upgrading their phones every so many years is still going to happen because that sucks up RAM and flash.

The main issue with Fabs in China is . .China, they are scaling up their own AI farms and will 100% make sure that any memory they produce will be for their market first the scraps left will be fair game I guess. (Not that I would be using it, I have very distinct memories of the early 200s when Chinese memory fabs were dumping cheap DDR into the market and it was ...not great for anyone)

Their tech is also quite a bit behind the Memory Cartel boys so itll be a good number of years before they can fully compete outside of China without drawing unwanted attention from the memory Cartel and being made to tow the line on pricing.

What i would love is if Japan got back into the memory fabrication market as yet another source of memory.

cyclone3d wrote on 2026-08-08, 04:33:

So a huge last level cache (effectively) and then a single addon stick. […]
Show full quote

So a huge last level cache (effectively) and then a single addon stick.

Hahahahahahahaha. No thanks.

Yes, I would like to buy a 2TB stick of RAM.

The minimum I would accept as HBM would be 32GB (right now). And then additional RAM on slots would have to be at least dual channel, more likely quad channel.

It's amusing you think you have some say in this.
SOCs are the emerging future.
First on laptops and other mobile. Then the "convergence" debate will be had for the 593rd time in computing history.
The only issue with SOCs is capacity and upgradability. Many users/OEMs want that.
Users don't so much notice speed differences between RAM A and RAM B. Aware OS kernels will make speed differences appear moot.

Think Optane. External ram will simply replace the space Optane occupied in the existing models or awareness of "I have to talk to the next CPU to get that data" which already exists in multi-socket systems/OSs.
It won't be as fast as the on-SOC HBM, but you won't really notice if you got the "gaming tier SOC" with 16gb onboard +DIMM vs the "Office tier SOC" with only 4gb +DIMM.

myne wrote on 2026-08-08, 07:38:

It's amusing you think you have some say in this. SOCs are the emerging future. First on laptops and other mobile. Then the "con […]
Show full quote

cyclone3d wrote on 2026-08-08, 04:33:

So a huge last level cache (effectively) and then a single addon stick. […]
Show full quote

So a huge last level cache (effectively) and then a single addon stick.

Hahahahahahahaha. No thanks.

Yes, I would like to buy a 2TB stick of RAM.

The minimum I would accept as HBM would be 32GB (right now). And then additional RAM on slots would have to be at least dual channel, more likely quad channel.

It's amusing you think you have some say in this.
SOCs are the emerging future.
First on laptops and other mobile. Then the "convergence" debate will be had for the 593rd time in computing history.
The only issue with SOCs is capacity and upgradability. Many users/OEMs want that.
Users don't so much notice speed differences between RAM A and RAM B. Aware OS kernels will make speed differences appear moot.

Think Optane. External ram will simply replace the space Optane occupied in the existing models or awareness of "I have to talk to the next CPU to get that data" which already exists in multi-socket systems/OSs.
It won't be as fast as the on-SOC HBM, but you won't really notice if you got the "gaming tier SOC" with 16gb onboard +DIMM vs the "Office tier SOC" with only 4gb +DIMM.

Pretty sure 16GB will be the minimum they will stick with, HBM isn't cheap to produce and making 4GB dies would be a waste of both silicon and fabrication capacity , 16GB dies allows you to throw 2 or 4 on a SOC and scale up to 64GB, then jump to 32GB dies to scale beyond 64GB for the development/AI focused SOCs. (I could see them making 8GB dies if they were also going to use them for GPUs) having a minimum of 16GB on the SOC means they could even just not bother with DDR6 for the budget/office PCs which is a huge cost saving.

*Even with DDR6 on the motherboard, 4GB HBM dies still don't make sense for SOC scalability.

Having gone and had a look at that thread, I think I can understand your warning Shponglefan.

elon gonna take care of it
https://x.com/niccruzpatane/status/2085485425376760239
https://www.spacex.com/updates#terafab

Re:china

There was news recently about china building a bunch of immersion DUV fab machines.

These arent useful for ram or cpus, but would be great for making flash memory on.

Given 'chinese trash' is the only media for sale right now, I wont complain if the quality and availability increase.

Here's being hopeful. Not else to do, except scowl.

myne wrote on 2026-08-08, 07:38:

It's amusing you think you have some say in this. SOCs are the emerging future. First on laptops and other mobile. Then the "con […]
Show full quote

cyclone3d wrote on 2026-08-08, 04:33:

So a huge last level cache (effectively) and then a single addon stick. […]
Show full quote

So a huge last level cache (effectively) and then a single addon stick.

Hahahahahahahaha. No thanks.

Yes, I would like to buy a 2TB stick of RAM.

The minimum I would accept as HBM would be 32GB (right now). And then additional RAM on slots would have to be at least dual channel, more likely quad channel.

It's amusing you think you have some say in this.
SOCs are the emerging future.
First on laptops and other mobile. Then the "convergence" debate will be had for the 593rd time in computing history.
The only issue with SOCs is capacity and upgradability. Many users/OEMs want that.
Users don't so much notice speed differences between RAM A and RAM B. Aware OS kernels will make speed differences appear moot.

Think Optane. External ram will simply replace the space Optane occupied in the existing models or awareness of "I have to talk to the next CPU to get that data" which already exists in multi-socket systems/OSs.
It won't be as fast as the on-SOC HBM, but you won't really notice if you got the "gaming tier SOC" with 16gb onboard +DIMM vs the "Office tier SOC" with only 4gb +DIMM.

It's fine for general users...

Most general users want the thinnest lightest piece of trash they can get their hands on.

Soldered on CPU and RAM and even SSDs with 0 upgradeability. The general users eat it up like candy and the companies love it because they are assured more sales when the trash they sell dies or the user needs a faster device.

It has been the same way for years.

I don't see why mfgs would bother putting expensive HBM in these devices when the could just use cheaper RAM.

And Optane... Ha. It was an idea.

The only laptop I had that had that, I got for free and it was because somebody was throwing it out.

I looked up the info on that part and it was well known that it was the part that died in the laptops that used it.

Never had one iota of desire to buy Optane and use it in any of my computers.

I recently experimented with a 16GB Optane NVME on a LGA 1151 Z270 motherboard and an old 10K Raptor HD. The Optane "cache" did speed up the spinning disk quite a bit.

I was hearing that Optane drives in laptops tended to die, wonder if it was just an overheating issue.

cyclone3d wrote on 2026-08-08, 21:19:

It's fine for general users... […]
Show full quote

myne wrote on 2026-08-08, 07:38:

It's amusing you think you have some say in this. SOCs are the emerging future. First on laptops and other mobile. Then the "con […]
Show full quote

cyclone3d wrote on 2026-08-08, 04:33:

So a huge last level cache (effectively) and then a single addon stick. […]
Show full quote

So a huge last level cache (effectively) and then a single addon stick.

Hahahahahahahaha. No thanks.

Yes, I would like to buy a 2TB stick of RAM.

The minimum I would accept as HBM would be 32GB (right now). And then additional RAM on slots would have to be at least dual channel, more likely quad channel.

It's amusing you think you have some say in this.
SOCs are the emerging future.
First on laptops and other mobile. Then the "convergence" debate will be had for the 593rd time in computing history.
The only issue with SOCs is capacity and upgradability. Many users/OEMs want that.
Users don't so much notice speed differences between RAM A and RAM B. Aware OS kernels will make speed differences appear moot.

Think Optane. External ram will simply replace the space Optane occupied in the existing models or awareness of "I have to talk to the next CPU to get that data" which already exists in multi-socket systems/OSs.
It won't be as fast as the on-SOC HBM, but you won't really notice if you got the "gaming tier SOC" with 16gb onboard +DIMM vs the "Office tier SOC" with only 4gb +DIMM.

It's fine for general users...

Most general users want the thinnest lightest piece of trash they can get their hands on.

Soldered on CPU and RAM and even SSDs with 0 upgradeability. The general users eat it up like candy and the companies love it because they are assured more sales when the trash they sell dies or the user needs a faster device.

It has been the same way for years.

I don't see why mfgs would bother putting expensive HBM in these devices when the could just use cheaper RAM.

And Optane... Ha. It was an idea.

The only laptop I had that had that, I got for free and it was because somebody was throwing it out.

I looked up the info on that part and it was well known that it was the part that died in the laptops that used it.

Never had one iota of desire to buy Optane and use it in any of my computers.

They will use HBM because they cannot fabricate two types of memory at the same plant at the same time .. its literally one or the other. So they either make HBM and use that or they shut the plant down and retool to make DDR ...I don't need to tell you which is currently selling for more and will be for the foreseeable future.

Makes economic sense to use what is being mass produced.

cyclone3d wrote on 2026-08-08, 21:19:

... Never had one iota of desire to buy Optane and use it in any of my computers ...

I'd kill to have one of these alien artifacts in everyone of my computers:

The Storage of the Gods

I don't even want to know what their prices must be right now ...

TELVM wrote on 2026-08-09, 05:56:

I'd kill to have one of these alien artifacts in everyone of my computers: […]
Show full quote

cyclone3d wrote on 2026-08-08, 21:19:

... Never had one iota of desire to buy Optane and use it in any of my computers ...

I'd kill to have one of these alien artifacts in everyone of my computers:

The Storage of the Gods

I don't even want to know what their prices must be right now ...

mmmmmmmmm I wanted one or two of them so bad for my X299 system back in the day, before NVME drives could do 14 gig a second it had U.2 ports for them too.

cyclone3d wrote on 2026-08-08, 21:19:

It's fine for general users... […]
Show full quote

myne wrote on 2026-08-08, 07:38:

It's amusing you think you have some say in this. SOCs are the emerging future. First on laptops and other mobile. Then the "con […]
Show full quote

cyclone3d wrote on 2026-08-08, 04:33:

So a huge last level cache (effectively) and then a single addon stick. […]
Show full quote

So a huge last level cache (effectively) and then a single addon stick.

Hahahahahahahaha. No thanks.

Yes, I would like to buy a 2TB stick of RAM.

The minimum I would accept as HBM would be 32GB (right now). And then additional RAM on slots would have to be at least dual channel, more likely quad channel.

It's amusing you think you have some say in this.
SOCs are the emerging future.
First on laptops and other mobile. Then the "convergence" debate will be had for the 593rd time in computing history.
The only issue with SOCs is capacity and upgradability. Many users/OEMs want that.
Users don't so much notice speed differences between RAM A and RAM B. Aware OS kernels will make speed differences appear moot.

Think Optane. External ram will simply replace the space Optane occupied in the existing models or awareness of "I have to talk to the next CPU to get that data" which already exists in multi-socket systems/OSs.
It won't be as fast as the on-SOC HBM, but you won't really notice if you got the "gaming tier SOC" with 16gb onboard +DIMM vs the "Office tier SOC" with only 4gb +DIMM.

It's fine for general users...

Most general users want the thinnest lightest piece of trash they can get their hands on.

Soldered on CPU and RAM and even SSDs with 0 upgradeability. The general users eat it up like candy and the companies love it because they are assured more sales when the trash they sell dies or the user needs a faster device.

It has been the same way for years.

I don't see why mfgs would bother putting expensive HBM in these devices when the could just use cheaper RAM.

And Optane... Ha. It was an idea.

The only laptop I had that had that, I got for free and it was because somebody was throwing it out.

I looked up the info on that part and it was well known that it was the part that died in the laptops that used it.

Never had one iota of desire to buy Optane and use it in any of my computers.

They're already using HBM for laptop SOCs.
First, battery life.
That'll bleed into desktop at the ultra low end non-upgradable.
Then server... possibly low end, possibly high. Not sure.
Server will be the first with HBM+DDR. HBM will be the first level ram, and have DDRx on the side.
Which will move into HEDT as some supermegaultracore.
And finally midrange desktops.

Why?
Cache is mad expensive.
Ram isn't fast enough.
DDR isn't close enough or low power enough.
HBM will allow 90% of the effective performance and power of an L3, and relatively slower DDR will compliment it for pure size.

myne wrote on 2026-08-09, 06:13:

They're already using HBM for laptop SOCs. First, battery life. That'll bleed into desktop at the ultra low end non-upgradable. […]
Show full quote

cyclone3d wrote on 2026-08-08, 21:19:

It's fine for general users... […]
Show full quote

myne wrote on 2026-08-08, 07:38:

It's amusing you think you have some say in this. SOCs are the emerging future. First on laptops and other mobile. Then the "con […]
Show full quote

It's amusing you think you have some say in this.
SOCs are the emerging future.
First on laptops and other mobile. Then the "convergence" debate will be had for the 593rd time in computing history.
The only issue with SOCs is capacity and upgradability. Many users/OEMs want that.
Users don't so much notice speed differences between RAM A and RAM B. Aware OS kernels will make speed differences appear moot.

Think Optane. External ram will simply replace the space Optane occupied in the existing models or awareness of "I have to talk to the next CPU to get that data" which already exists in multi-socket systems/OSs.
It won't be as fast as the on-SOC HBM, but you won't really notice if you got the "gaming tier SOC" with 16gb onboard +DIMM vs the "Office tier SOC" with only 4gb +DIMM.

It's fine for general users...

Most general users want the thinnest lightest piece of trash they can get their hands on.

Soldered on CPU and RAM and even SSDs with 0 upgradeability. The general users eat it up like candy and the companies love it because they are assured more sales when the trash they sell dies or the user needs a faster device.

It has been the same way for years.

I don't see why mfgs would bother putting expensive HBM in these devices when the could just use cheaper RAM.

And Optane... Ha. It was an idea.

The only laptop I had that had that, I got for free and it was because somebody was throwing it out.

I looked up the info on that part and it was well known that it was the part that died in the laptops that used it.

Never had one iota of desire to buy Optane and use it in any of my computers.

They're already using HBM for laptop SOCs.
First, battery life.
That'll bleed into desktop at the ultra low end non-upgradable.
Then server... possibly low end, possibly high. Not sure.
Server will be the first with HBM+DDR. HBM will be the first level ram, and have DDRx on the side.
Which will move into HEDT as some supermegaultracore.
And finally midrange desktops.

Why?
Cache is mad expensive.
Ram isn't fast enough.
DDR isn't close enough or low power enough.
HBM will allow 90% of the effective performance and power of an L3, and relatively slower DDR will compliment it for pure size.

Agree mostly with the exception of HBM on SOC for servers . .that wont happen, itll be HBM on CAMM2 mostly for scalability and maintenance reasons.

Trashbytes wrote on 2026-08-09, 06:06:

... before NVME drives could do 14 gig a second ...

^ At sequential, yes.

But at random 4K low queue depth, and latency, even the fastest NVMe @ 2026 is still trampled over and left in the dust by those years old alien artifacts ...

TELVM wrote on 2026-08-09, 06:55:

^ At sequential, yes. […]
Show full quote

Trashbytes wrote on 2026-08-09, 06:06:

... before NVME drives could do 14 gig a second ...

^ At sequential, yes.

But at random 4K low queue depth, and latency, even the fastest NVMe @ 2026 is still trampled over and left in the dust by those years old alien artifacts ...

That less to do with the drive and more to do with how Optane worked, Intel had some black magic voodoo built into Optane.

Them U.2 Optane drives still fetch a high price too, they never really depreciated.

Trashbytes wrote on 2026-08-09, 06:59:

... Intel had some black magic voodoo built into Optane ...

Some alien tech from Area 51, I suspect 😀

TELVM wrote on 2026-08-09, 07:18:

Some alien tech from Area 51, I suspect :-) […]
Show full quote

Trashbytes wrote on 2026-08-09, 06:59:

... Intel had some black magic voodoo built into Optane ...

Some alien tech from Area 51, I suspect 😀

I wish they had kept on developing it, Intel seems to develop some insane tech and then just .. abandons it or just stops development but doesn't do much with it.

Trashbytes wrote on 2026-08-09, 06:18:

myne wrote on 2026-08-09, 06:13:

They're already using HBM for laptop SOCs. First, battery life. That'll bleed into desktop at the ultra low end non-upgradable. […]
Show full quote

cyclone3d wrote on 2026-08-08, 21:19:

It's fine for general users... […]
Show full quote

It's fine for general users...

Most general users want the thinnest lightest piece of trash they can get their hands on.

Soldered on CPU and RAM and even SSDs with 0 upgradeability. The general users eat it up like candy and the companies love it because they are assured more sales when the trash they sell dies or the user needs a faster device.

It has been the same way for years.

I don't see why mfgs would bother putting expensive HBM in these devices when the could just use cheaper RAM.

And Optane... Ha. It was an idea.

The only laptop I had that had that, I got for free and it was because somebody was throwing it out.

I looked up the info on that part and it was well known that it was the part that died in the laptops that used it.

Never had one iota of desire to buy Optane and use it in any of my computers.

They're already using HBM for laptop SOCs.
First, battery life.
That'll bleed into desktop at the ultra low end non-upgradable.
Then server... possibly low end, possibly high. Not sure.
Server will be the first with HBM+DDR. HBM will be the first level ram, and have DDRx on the side.
Which will move into HEDT as some supermegaultracore.
And finally midrange desktops.

Why?
Cache is mad expensive.
Ram isn't fast enough.
DDR isn't close enough or low power enough.
HBM will allow 90% of the effective performance and power of an L3, and relatively slower DDR will compliment it for pure size.

Agree mostly with the exception of HBM on SOC for servers . .that wont happen, itll be HBM on CAMM2 mostly for scalability and maintenance reasons.

Eh. I just can't see 8 sticks of CAMM2 per socket. Which is what the server guys always want/expect regardless of whether they need it.
Plus, DDR is likely to remain cheaper in bulk forever since you can validate each chip individually. HBM has to be assembled and then tested.