3 ms·

Compare https://images.anandtech.com/doci/16823/AnandTechRoadmaps3.png https://images.anandtech.com/doci/16823/AnandTechRoadmaps3.p... against what actually hap

Compare https://images.anandtech.com/doci/16823/AnandTechRoadmaps3.png https://images.anandtech.com/doci/16823/AnandTechRoadmaps3.p... against what actually happened: Intel 7 was reasonably successful, except for the desktop chips that were burning themselves out because Intel couldn't get their voltage regulation working right. Intel 4 resulted in slight performance regressions for laptops, but battery life improvements because they moved everything except the CPU cores over to TSMC. Intel 3 was late. Intel 20A was cancelled. Intel 18A isn't here, and neither is the CEO in charge of the above strategy.

tester756 2y ago
>except for the desktop chips that were burning themselves out because Intel couldn't get their voltage regulation working right. that was firmware issue, not the node, yup?
- wtallis 2y ago
  It was a combination of issues. Intel needed to be able to supply quite high voltages (at the right moments) to reach high enough clock speeds to win benchmarks with their desktop CPUs, so they couldn't implement a simple conservative, safe cap on the maximum voltage that would be requested. Their attempts to build an on-die voltage regulator (DLVR) repeatedly failed, leaving Intel relying exclusively on a different voltage regulator mechanism than they'd originally planned for. And then they seemingly did a sloppy rush job of characterizing how the chip behaved with their fallback voltage regulator strategy, and messed up the code controlling that system. There were many steps at which Intel could have avoided this outcome, but it started from the fact that Intel had to push voltages and clocks beyond reasonable limits to make up for having a microarchitecture with poor enough performance per clock that Intel needed to hit 6+ GHz to be competitive.

3 ms·

Compare https://images.anandtech.com/doci/16823/AnandTechRoadmaps3.png https://images.anandtech.com/doci/16823/AnandTechRoadmaps3.p... against what actually hap

by wtallis 2y ago

tester756 2y ago
>except for the desktop chips that were burning themselves out because Intel couldn't get their voltage regulation working right. that was firmware issue, not the node, yup?
- wtallis 2y ago
  It was a combination of issues. Intel needed to be able to supply quite high voltages (at the right moments) to reach high enough clock speeds to win benchmarks with their desktop CPUs, so they couldn't implement a simple conservative, safe cap on the maximum voltage that would be requested. Their attempts to build an on-die voltage regulator (DLVR) repeatedly failed, leaving Intel relying exclusively on a different voltage regulator mechanism than they'd originally planned for. And then they seemingly did a sloppy rush job of characterizing how the chip behaved with their fallback voltage regulator strategy, and messed up the code controlling that system. There were many steps at which Intel could have avoided this outcome, but it started from the fact that Intel had to push voltages and clocks beyond reasonable limits to make up for having a microarchitecture with poor enough performance per clock that Intel needed to hit 6+ GHz to be competitive.