Kind of? It's inherently a slow process that's also kinda expensive.Is the problem with SoIC that it is God level tech that can not be replicated because all knowledge how it was done is now totally lost
It's cheap enough to be in 9800X3D, what's the premium there: 100 bucks max?Kind of? It's inherently a slow process that's also kinda expensive.
Which will shift the deadline for the non cache products if you want to have sufficient stock.No, it does NOT have to launch later - they can first accumulate non-3D chips to actually have stock for launch, and some 3D ones that will be super popular and out of stock quickly, we are talking months here max, not years, so it's totally doable and should have been done.
Not really it is a from of hybrid bonding and Sony uses Hybrid Bonding for their sensors.Is the problem with SoIC that it is God level tech that can not be replicated because all knowledge how it was done is now totally lost?
N2 Wafer is ridiculously expensive compared to what they are paying now for N4P at minimum 50% increase is guaranteed in Wafer cost.If AMD makes commitment, and it can't be too expensive, this ain't N2 fabs, then everything will be done with bells on. That's the problem - AMD.
You'll produce them and stock up, small number (those made first) goes for premium 3D models that take longer to finish than you launch the whole thing. It's not like launches are every quarter - they are once every 20-22 months, plenty of time available to plan it out and execute without usual paper launching.Which will shift the deadline for the non cache products if you want to have sufficient stock.
It's not idiotic when in the very near future they will be forced to use zero L3 cache on very expensive sub N2 processes, it will become totally essential, so making sure there is capacity for it NOW is smart thing to do.X3D doesn't add value for majority of application, making whole portfolio base on such process is idiotic.
You're almost getting there. On the one hand you say that they don't have to launch later than after you say that you acknowledge that it would launch later by months. The latter part is right, if you assume there would be enough capacity to do so, which there isn't.No, it does NOT have to launch later - they can first accumulate non-3D chips to actually have stock for launch, and some 3D ones that will be super popular and out of stock quickly, we are talking months here max, not years, so it's totally doable and should have been done.
They need to launch all this new stuff at the same time, or at least announce later (but not much) availability for desired parts.On the one hand you say that they don't have to launch later than after you say that you acknowledge that it would launch later by months.
Thank youThey need to launch all this new stuff at the same time, or at least announce later (but not much) availability for desired parts.
What is better:
1) wait 20 months till launch, then X months while they make unannounced 3D parts and who knows if they do quick launch like Zen 5 - bloody pain knowing already new cool things about new CPUs
2) wait 23 months till launch - everything is available!
It's still a very long wait but the most painful wait is after they announce! 3 months before and after are not the same
MI300X already did thatI think before AMD Produces such thing if Intel launches Clearwater Forest on Q1 26 it would be comical they would become the first one to put cache and IMC in hybrid bonded die.
No!So we've arrived at the conclusion that they can either launch earlier with everything or delay launch and have X3D ready for launch.
Ah right, your whole argument this time has been about the word delay. Fair enough.No!
A delayed launch is a launch that was announced (with date) and then delayed - like Zen 5 was.
Getting all your ducks in a row is not a delayed launch - Apple could have launched in August if they did not have to do more complex newer Pro versions of phones, is their launch delayed? No - it's planned to launch together in perfect synchrony.
If we were to accept your logic then not launching product as soon as first items are finished is somehow delayed launch - not at all, enough should be made for launch and it's totally normal that it might take few months - like Apple does, which is why they sell a lot when they launch stuff when the price is at the highest point. There should be bloody stock at MSRP for longer than 5 seconds, that's now normalized but it's not acceptable!
What should be good stock for launch? Assuming even production it's obvious that new stuff should have a lot more than 1 month of production, I'd say 3-4 months minimum - with that in mind there is plenty of time to do 3D versions for the units made in month 1, that's the way to do it.
Downside? Beancounters at AMD will have having unsold inventory on balance sheet for 3-4 months.
Is it a launch delay if they wait until 1000 CPUs ready? 10000? 1 mln?Ah right, your whole argument this time has been about the word delay. Fair enough.
Yeah fair enough. He's a tough nut to crack. Difficult. Numbers and concepts have no meaning hereGuys, can we stop this nonsense?
Adding 3D stacking takes time. Thats the only concept you need to understand, which you already acknowledged. Let's knock this on the head.Is it a launch delay if they wait until 1000 CPUs ready? 10000? 1 mln?
How many days of production before launch means delay of launch?
Any big launch should be at least 3-4 months of production, not 1 week and not 1 month, this should be normal especially given that waiting cycle is almost 2 years anyway.
You do understand that DIY CPUs have obscene margins, don't you?It's cheap enough to be in 9800X3D, what's the premium there: 100 bucks max?
N2 Wafer is ridiculously expensive compared to what they are paying now for N4P at minimum 50% increase is guaranteed in Wafer cost.
You are forgetting the design cost as well which also 1.6X also N4P wafer is 15.5K and N2 wafer is around 23-25K(this is my estimate based on N3 and from what i have heard).How big of a problem is that, really?
Let's say N4P is $20K and N2 is $30K, and our chips are 100mm^2, to give us nice round numbers to work with. That gives you 600 chips per wafer, costing about $33 on N4P and $50 on N2. Taking a yield hit (larger on N2 since N4P is mature) lets call that $38 for N4P and $60 for N2. Is it really going to be a major blow to AMD's profitability if they have to pay an extra $22 per chip on N2? It isn't like they have to use N2 across their whole product portfolio.
It already doesn't make sense to produce their lowest end stuff in N4P, and the crossover in ASP between where they have to stick with older nodes and where they can afford to pay more for newer nodes will continue to increase over time.
Those are fake numbers for design costs.You are forgetting the design cost as well which also 1.6X also N4P wafer is 15.5K and N2 wafer is around 23-25K(this is my estimate based on N3 and from what i have heard).
View attachment 122972
You are forgetting the design cost as well which also 1.6X also N4P wafer is 15.5K and N2 wafer is around 23-25K(this is my estimate based on N3 and from what i have heard).
Yes, and what does obscene margins mean on 450 bucks product? This means cost of making is very cheap - including 3D stuff.You do understand that DIY CPUs have obscene margins, don't you?
Means the margins are obscene because the volume is tiny.Yes, and what does obscene margins mean on 450 bucks product?
No it means the volume is tiny.This means cost of making is very cheap - including 3D stuff.
You can't just scale hybrid bonding capacity 10 times, that's a ton of equipment.Scale it 10 times and it will be a lot cheaper and bigger volume.