Again and again
https://www.igorslab.de/en/intel-18a-delay-manufacturing-problems-foundry-offensive/
5 replies (most recent on top)
OP forget about x86 50 years old addictive legacy architect implemented in the brain with hundreds of surgery micro software patches every year.
Intel's 18A process has faced significant yield challenges, with reported yields around 10% that are far below what is needed for cost-effective production. Intel acknowledges ongoing yield improvements but notes that adjustments to Panther Lake CPUs impacted 18A yields, delaying stabilization.
Overview of Intel's 18A Process Issues
Intel's 18A manufacturing process has encountered serious challenges that threaten its production capabilities and timelines. The primary concern is the low yield rate, which is currently around 10%. This yield is significantly below the levels required for cost-effective mass production.
Key Challenges
Yield Problems: The 10% yield rate indicates that only a small fraction of produced chips meet quality standards, making it economically unfeasible to continue at this level.
Complex Technology: The 18A process incorporates advanced technologies such as gate-all-around transistors (RibbonFET) and a backside power supply (PowerVia). This complexity increases the likelihood of production errors.
Impact on Panther Lake CPUs: Adjustments made to the Panther Lake CPUs have negatively affected the yield rates of the 18A process. Intel is working to stabilize these yields, but the timeline remains uncertain.
Implications
Aspect Current Status Potential Impact
Yield Rate Approximately 10% Economically unviable production
Production Timeline Officially on track for late 2025 Risk of delays and limited product availability
Strategic Ambitions Competing in high-end manufacturing Possible withdrawal from high-end segment
Intel acknowledges the need for significant improvements in yield rates to ensure the economic viability of the 18A process. Without these improvements, the launch of the Panther Lake CPUs and Intel's broader strategic goals may be jeopardized.
How long is Intel going to need to prove that it can’t run a fcking fab before it admits it?
Probably never. They’ll just keep Tin Cupping it until they can position themselves for another bailout from POTUS.
AMD and NVIDIA went fabless decade ago. They are still rocking and rolling without any fab :-)
Intel refused to learn but wanting to lead. More struggles to come because Moore Law is dead.
LBT Law "more work less people" will create misery life for employees
It almost seems braindead that Intel continues to try to fab parts, something it should have reconsidered 20 years ago. But then it also struggles to design competitive parts.