• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Question 5 Nodes in 4 years discussion

Page 5 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
I'm surprised that Nvidia aren't looking at putting lower end parts on an Intel process. They used Samsung 14nm for the 1050 cards back in the day, they could put the 5050/5060 on Intel. Frees up more of their TSMC capacity for top end GPUs and AI parts, gives them more leverage next time they negotiate with TSMC, and helps cultivate a plan B in case they can't get TSMC capacity for whatever reason.
Now that 50 series looks to be on 4N which is not cutting edge, alternative processes can be viable performance wise. But ironically, 4N is probably not that expensive either compared to 3N which most people would have guessed would be used 1-2 years ago, so why multi source? I suspect they won't until there's heavy incentives. NV went for maximizing profit, 40 series seems to have been the exception, 30 series was on a cheap node, and 50 series is on semi cheap one.
 
I'm surprised that Nvidia aren't looking at putting lower end parts on an Intel process. They used Samsung 14nm for the 1050 cards back in the day, they could put the 5050/5060 on Intel. Frees up more of their TSMC capacity for top end GPUs and AI parts, gives them more leverage next time they negotiate with TSMC, and helps cultivate a plan B in case they can't get TSMC capacity for whatever reason.

Intel nodes aren't cheaper than TSMC, have lower yields and worse performance. Intel would have to subsidize their foundry prices to attract customers, which they cant do because they are closer to bankruptcy than having Samsung's deep pockets.
 
I'm surprised that Nvidia aren't looking at putting lower end parts on an Intel process. They used Samsung 14nm for the 1050 cards back in the day, they could put the 5050/5060 on Intel. Frees up more of their TSMC capacity for top end GPUs and AI parts, gives them more leverage next time they negotiate with TSMC, and helps cultivate a plan B in case they can't get TSMC capacity for whatever reason.
Cause Samsung is much better than Intel and its not even close. Intel literally has nothing for external foundries and anyone signing up would be doing alpha testing with massive delays. TSMC is just way better than either choice and you can still use lagging edge nodes on TSMC without spending that much money.
 
Cause Samsung is much better than Intel and its not even close. Intel literally has nothing for external foundries and anyone signing up would be doing alpha testing with massive delays. TSMC is just way better than either choice and you can still use lagging edge nodes on TSMC without spending that much money.
Intel 3 is available for costumers and Intel 18A seems to be close to HVM. Why they would be signing to massive delays? Not everyone needs leading edge, MediaTek has already a deal to produce on Intel 16 and later I guess on UMC 12. Mediatek is not a small customer.
 
Intel 3 is available for costumers and Intel 18A seems to be close to HVM. Why they would be signing to massive delays? Not everyone needs leading edge, MediaTek has already a deal to produce on Intel 16 and later I guess on UMC 12. Mediatek is not a small customer.

He may be citing this report:


That's not directly saying there will be delays, but if Intel needs to revise 18a before anyone would actually want to use it, then effectively there would be delays.
 

Intel to Produce Custom AI Fabric Chip on Intel 18A and Custom Xeon 6 Chip on Intel 3 for AWS

Multi-Year, Multi-Billion-Dollar Collaboration Accelerates Development of Chip Manufacturing in Ohio
 

Intel to Produce Custom AI Fabric Chip on Intel 18A and Custom Xeon 6 Chip on Intel 3 for AWS

Multi-Year, Multi-Billion-Dollar Collaboration Accelerates Development of Chip Manufacturing in Ohio

The Xeon announcement isn’t really anything, Intel has long provided “custom” Xeons for big customers every generation. The AI fabric deal might be interesting though, wish there were more details. If it’s similar to the Xeon deal, it’s not very interesting, but if Amazon is actually opening a new product portfolio for Intel to manufacture, that would be big news. I’m guessing it’s the former though.
 
Intel 3 is available for costumers and Intel 18A seems to be close to HVM. Why they would be signing to massive delays? Not everyone needs leading edge, MediaTek has already a deal to produce on Intel 16 and later I guess on UMC 12. Mediatek is not a small customer.
Mediatek has not signed a deal for cutting edge nodes, their deal is for IOT devices probably on old nodes offered the acquisition of Tower, which continues for Intel. It just allows Mediatek to have another supplier for chips using old processes without buying them all from China. A large company could still make small deals and if you read between the lines, theres nothing of substance there for IDF 2.0.
 
He may be citing this report:


That's not directly saying there will be delays, but if Intel needs to revise 18a before anyone would actually want to use it, then effectively there would be delays.
Considering Broadcom didn't even have PDK 1.0 (Which was released two months ago) I am not sure why people put up much stock in that report.
18A seems to be progressing well judging by the fact that Amazon just signed a deal to use it.
 
Considering Broadcom didn't even have PDK 1.0 (Which was released two months ago) I am not sure why people put up much stock in that report.
18A seems to be progressing well judging by the fact that Amazon just signed a deal to use it.

Amazon signed a deal for Intel to use it to make a product for Amazon.
 
Back
Top