News 7nm Supply Showdown: AMD, Nvidia May Fight for Scraps as Apple Reportedly Ups A13 Production

hannibal

Distinguished
Apr 1, 2004
2,267
26
19,840
14
Boosting production takes time. It takes two year to build fab and make it running properly. So we will have 7nm shortage to 2022... if the demand does not get bigger in the meanwhile...
Heh... maybe amd and nvidianhave to go back to 12nm in some of their products like Intel is stuck at 14nm+++++ though Intel problems Are a Little different. It seems than 7nm from TSMC is desent and because of that demand is higher.
Big Navi moves to 2021 because of this? Nvidia would like that!
 
What I rather don't get is, why are people still thrilled about iPhones to the point where they "beat expected demands"? Sure they have unrivalled performance but isn't a second hand iPhone X or heck even the iPhone 8 a much wiser choice considering there's no revolutionary update since then?
 
Reactions: bit_user

NightHawkRMX

Illustrious
Truely new iphones are just so overpriced i cant understand why people still buy them.

I long for the iPhone 5s days. Revolutionary features like a fingherprint sensor and incredably solid build.

Metal and glass and feels great in the had.
 

InvalidError

Titan
Moderator
This is unfortunate, but at least if production is boosted significantly and they are able to meet the demand, prices shouldn't spike too much
I'd expect AMD's contract for baseline wafer volume to stretch quite far into the future to avoid having an outright shortage at any point in time, I wouldn't worry about it unless TSMC has a power outage or other event that messes up production for weeks or months.
 
Reactions: bit_user
May 31, 2019
71
20
35
0
Apple and Huawei probably still 1 and 2 in tsmc 7nm. They combined for about 85% of 7nm volume there in 2018.

I suspect they have the leverage to demand priority processing.

Maybe AMD gets some love, since they are pushing development of high performance cpus. The AMD partnership gives tsmc an experiment with higher power process corners vs the low power mobile phone stuff.
 

InvalidError

Titan
Moderator
I suspect they have the leverage to demand priority processing.
If AMD's agreement with TSMC is for a hypothetical 50k wafers starts per month baseline commitment for the next two years, AMD gets 50k wafers/month regardless of anybody else's orders for the next two years. If someone wants to jump queue and displace TSMC's commitment to AMD, they will have to negotiate compensation with AMD first to get AMD to relinquish those wafer starts.
 

martel80

Distinguished
Dec 8, 2006
368
0
18,780
0
Aren't mobile chips much smaller compared to Radeon and Ryzen dies, so there are more of them per wafer?
Apple manufacturing the most CPUs does not mean they take the most production capacity.
 

jimmysmitty

Champion
Moderator
If AMD's agreement with TSMC is for a hypothetical 50k wafers starts per month baseline commitment for the next two years, AMD gets 50k wafers/month regardless of anybody else's orders for the next two years. If someone wants to jump queue and displace TSMC's commitment to AMD, they will have to negotiate compensation with AMD first to get AMD to relinquish those wafer starts.
I don't think they would have to relinquish but if they are selling more than they are producing they will have a harder time increasing that amount if TSMC is pushing towards capacity. Its the only downside to not running your own FAB, you have no control over how much you can get out of it.
 

InvalidError

Titan
Moderator
I don't think they would have to relinquish but if they are selling more than they are producing they will have a harder time increasing that amount if TSMC is pushing towards capacity. Its the only downside to not running your own FAB, you have no control over how much you can get out of it.
You do have SOME control, it just comes with the disadvantage of extra lag from the supply chain. If you need 50k extra wafers to catch up with demand, you can still get them, you just have to wait six months for your extra order to reach the front of queue if you didn't already have an option for additional wafers or a sufficiently large one in the original agreement to get them sooner.
 

jimmysmitty

Champion
Moderator
You do have SOME control, it just comes with the disadvantage of extra lag from the supply chain. If you need 50k extra wafers to catch up with demand, you can still get them, you just have to wait six months for your extra order to reach the front of queue if you didn't already have an option for additional wafers or a sufficiently large one in the original agreement to get them sooner.
And hopefully the others are not putting in more to keep ahead of the curve and pushing yours out further. Yes order extra but it will be a bit of a struggle.

Its not the end of the world but if AMD really wants to compete with Intel they need to be able to fill vendor channels at the drop of a hat and they have already had some supply issues with the higher end chips. The more popular they get in mainstream, HEDT and server markets the more this will become an issue especially as more and more companies move to 7nm.

This also means they would make sure certain channels are filled more than others. Mainstream is not as much of a money maker as server/HPC is so that's where I imagine they would focus most of their available supply.

I am honestly surprised Intel didn't run into supply issues long ago considering they held 99% of the server market and 80% of the desktop market for many years.
 

InvalidError

Titan
Moderator
I am honestly surprised Intel didn't run into supply issues long ago considering they held 99% of the server market and 80% of the desktop market for many years.
I can think of a simple reason why: pricing. At Intel's high premiums (at least until the recent ~50% cuts) at the higher-end, most people don't upgrade unless they absolutely have to. With no competition to speak of, Intel could regulate demand by setting prices accordingly, similar to what AMD is doing with its own higher-end parts where the retail price is way ahead of the incremental cost of adding chiplets.
 

ASK THE COMMUNITY