News Report: AMD Chip Shortage Caused by Packaging Issues, PlayStation 5

[HC1Gunner]

Does this same packaging issue also apply to AMD "GPU" chips? I was fortunate enough to get my hands on a 5900X, but have been waiting for a 3rd party RX 6800 XT.

 

[jimmysmitty]

Sorry... "Going under" would be better read "falling behind". I did not mean to imply they were on the verge of going out of business or into bankruptcy. IT was meant as a comparison to AMD's tech improvements vs Intel's over the past few years.
Intel's woes with 10nm are because they had several issues they were unable to iron out while developing the 10nm process. The need to research this slowed them down and allowed AMD to jump ahead in performance, because they did not have to research and design a new fab technique. They were able to focus on creating a quality product and leaving the manufacturing headaches to someone else.

I understand the desire to build in-house, if for no other reason than to control production from start to end. But this comes at a cost, and was more than Intel could chew on. Now they are suffering the consequences while they look to outsource chips for a bit and narrow the focus as they should have awhile back. Specialization is an expensive luxury...One that Intel was unable to sustain indefinitely.

Its easier for a company that had a horrible product (Bulldozer) to make great progress than one that was already the top. And if current leaks are to be believed the gains in Rocket Lake look pretty impressive. Process tech is important but not the only way to improve performance.

AMD is now in the same boat as Intel though. They are waiting for node improvements that will take longer and their performance increases will become more minimal. It took three generations of Zen to finally beat Skylake. Its an impressive feat but its not as impressive as some make it out to be as Skylake is 5 years old.

Its also more than just controlling the product though. AMD is also at the mercy of these FABs. They can buy as much as they want but others can easily come in and bid higher for more capacity. This means AMD will have to compete price wise with other companies, Apple possibly moving their main system CPUs to their own adds another competitor for AMD. Intel doesn't have to worry as much. All their capacity is for their products.

 

[InvalidError]

Process tech is important but not the only way to improve performance.

Achieving performance improvements requires more transistors and packing more transistors within a chip while still meeting timing and power margins when you are already at the limits of what is possible on a given process requires process advancements in one form or another. Rocket Lake is only possible today because Intel's 14nm has been refined to the point of performing nearly as good as 10nm was supposed to be 3-4 years ago before all of the delays and other problems cropped up.

 

[jimmysmitty]

Achieving performance improvements requires more transistors and packing more transistors within a chip while still meeting timing and power margins when you are already at the limits of what is possible on a given process requires process advancements in one form or another. Rocket Lake is only possible today because Intel's 14nm has been refined to the point of performing nearly as good as 10nm was supposed to be 3-4 years ago before all of the delays and other problems cropped up.

Its still not the only way to improve performance. Enhancements to the uArch can improve performance. Before their process tech woes Intel was doing that with their Tick/Tock. After a shrink they would debut a new uArch that increased performance. Then they would shrink the new uArch.

That was my point. That shrinking the process is not the only way to achieve performance gains.

 

[InvalidError]

After a shrink they would debut a new uArch that increased performance. Then they would shrink the new uArch.

That was my point. That shrinking the process is not the only way to achieve performance gains.

I think you missed my point: in tick-tock, Intel didn't capitalize on gains from the shrink right away because the first chip was just a shrink. That's what the new architecture did. So it is still new architecture driven by new process.