You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!
That is not a reliable source. They have been banned from SemiWiki due to threat of legal action when I called them out on it. No retraction was ever made even though the previous story was patently false. Please do not post anything from TrendF*rce on SemiWiki.
That is not a reliable source. They have been banned from SemiWiki due to threat of legal action when I called them out on it. No retraction was ever made even though the previous story was patently false. Please do not post anything from TrendF*rce on SemiWiki.
Well it’s certainly delayed as noted on the previous Q2 call where Pat mentioned issues with numerous (unexpected) steppings needed to get certain SKUs production ready. Now, are the steppings bugs in the design, or manufacturing, or both?
"TrendForce said the production yield of Sapphire Rapids is estimated to be between 50 and 60 percent. As a result mass production is being put back from 4Q22 to 1H23."
Mass production has already started, parts are in the supply chain, this story is false. This information came from the supply chain Last week.
It is right that all such details in their new technologies might not be correct/visible, but nevertheless Intel, who previously/initially lost their leadership in top new technologies given their lack of (tick-tock...) transitions in 10nm, has also now completely lost their top1 revenue in semiconductor market, with now this year some decrease YoY while most other companies where booming close to or strongly above +30%...
Intel just covering its R&D cost on Xeon Ice in q3 sold at average marginal cost. On Intel depreciated process the R&D cost savings is not FCF maintenance debited CapEx accrual. This exposes a reality on Intel's plant expansion plan budgeted up to $234 billion which at an R&D contribution per unit of Xeon production of $300 requires 781,666,667 units to offset the expansion cost.
In my last 18 months on the 10-Q assessment, Intel has produced 41,671,262 Xeon (not including hidden bundle deal sales sacrifice) of which less than one half are thought Ice Lake. The objective then must be to return Xeon to 100 M units per year production to pay the capital cost of expansion.
Production assessment shows back through Scalable Intel quite capable of 400 mm2ish HCC 14 to 16 core dies. My inquiry would address package final test. Next the issue of whole product capable of filling the general OEM finished systems sales distribution chain. mb
Apples and Oranges. Alder Lake is client part. Last DC part was Ice Lake at 505 mm^2. SR tile is 419 mm^2, so should yield higher than Ice Lake, but uses 4 of these tiles for each part, so higher cost.
"TrendForce said the production yield of Sapphire Rapids is estimated to be between 50 and 60 percent. As a result mass production is being put back from 4Q22 to 1H23."
Mass production has already started, parts are in the supply chain, this story is false. This information came from the supply chain Last week.
T-F article sounds like BS, but even if 50-60% yield were true, it would correspond to defect density of 0.12-0.17 cm^-2, which is not too far from full yield entitlement of best-in-class mature foundry node at 0.09 cm^-2. At 0.09, SR only yields 69%, so 50-60% likely well within intel's expectations, and it would not stop them from ramping.
