forums.anandtech.com
This:
This has to be a joke right?? First of all the raptor lake degradation issues were completely on the design team for Raptor lake, there isn't a fundamental issue in intel 7 that caused that issue. Meteor lake also was not a fab issue at all, it was a design problem as being the first generation of intel's tile architecture concept. Even if intel outsourced to TSMC earlier it's not like their products would've magically been competitive. Lunar lake is essentially monolithic arrow lake and even then it barely edges out the competition at the one thing it's designed for... power efficiency. And it certainly doesn't do that through it's over two node advantage over Raptor lake or even it's Lion Cove cores. Arrow lake isn't less performance than raptor lake simply because it's a tile architectures, It also would not have gained much performance due to the fact that TSMC nodes are historically lower performance than Intel nodes even 1-2 generation behind.
What if Intel's about to do to NVidia what AMD did to them and Koduri is their Lisa Su?I'm betting on Raja Koduri as his successor
The fab team must've signed off on the design team's plan to hit 5 GHz and above. Surely, testing that the silicon does not degrade under stress should be the fab's job, no? Otherwise, they could have put their foot down and told Pat, we object to these frequencies and voltages and cannot guarantee long term stability. Let's say this actually happened and Pat couldn't see a way out (because he is old and feeble) so he decided to err on the side of chance and release Raptor Lake and its Refresh against the advice of his foundry people. That's probably why Pat is out and not the head of foundry.
Meteor Lake was a pure yields issue and they wasted wafers trying to meet demand. The process was just too immature for prime time but Pat pushed them to waste wafers because he couldn't think of anything else (old and feeble mind there again). Lunar Lake is probably the best product Pat got out and it is thanks to being on TSMC and being monolithic. Did you miss Geekerwan's chart where a Lenovo Lunar Lake laptop is beating everything else in battery life?
Is that really all due to TSMC? Not due to the on package memory, Skymont E-cores, removal of HT, cleaning up of old inefficiencies in P-core architecture preparing for future scalability? We'll see shortly with panther lake/clearwater forest whether simply manufacturing on TSMC really is the golden bullet for all of intel's problems.
Meteor Lake yields issue is separate from the discussion on the performance of the product. The product was horrible because the lp e-cores were completely useless starved of L3 cache, the regular e-cores were pretty trash too being just a relatively weak architecture compared to its successor skymont, and the tons of issues that come with a completely new 1st gen tiled architecture. With the note of a completely new tiled architecture has RDNA 3 ever entered your mind? And that was significantly less complicated than meteor lake with all the computer being located on a single die.
That's not the main reason. The underlying issue is the foundry takes years to pay off and Intel upper management is not happy with that fact. If you were to evaluate Pat solely on foundry technology he'd be the best CEO intel has had in decades.
www.semiaccurate.com
As long as they are related to Lisa Su and Jensen, Intel will have a chance.
He's calling it bad from his point of view of thinking of Pat as an engineer.
This one i wholeheartedly agree but not of his strategy being bad the strategy is fine
Ahh i don't want to put stick in this but i found
Facebook
Twitter