Hacker News new | past | comments | ask | show | jobs | submit login

The new 18-core i9-10980XE isn't out yet, and review embargoes haven't been lifted. We know it's a slightly clock bumped version of the i9-9980XE, which was included in the AnandTech review and not the Linus review.

Linus tested the 8-core i9-9900KS, which came out a few weeks ago, and those results are also in the link which is the title to this thread.




Ian,

What are your thoughts on the question of whether Intel may be strategically timing releases of some mitigations to favor their performance in review benchmarks?


This decision was likely made from a different department, and just lucked out that it was the day before an AMD launch. If Intel wanted to delay it specifically to appear for better benchmarks, they would have waited until after Cascade Lake-X benchmarks were out. It's meant to be being launched later this month.


Yet the main question remain if the benchmark were done with the latest mitigation (and version of the microcode) or without...


No, I didn't have time to run 30hr+ of tests on 7-10 Intel CPUs with only 24 hours notice from the announcement yesterday


That would be responsible to mention in the review, with a promise to publish new comparisons whenever new mitigations are released.

Intel's security flaws pushed our LCRs up at least two years and cost us extra in over-provisioning to account for real and potential performance loss. If Intel isn't actually hardware mitigating (as in, making no effort to redesign their chips with security in mind and instead playing whack a mole with vulnerabilities as they're discovered) and reviewers are not highlighting and disclosing this, they're lying by omission.

I'm not being rude, but don't give Intel a pass here because of any prior fondness you might have held for them. Every benchmark of an Intel CPU you've ever published was made false by their negligence. That's a big, confusing problem for consumers, and they rely on your clarity and diligence to bring it to their attention and correct.


How much does performance degrade an average year for Intel and AMD? To a consumer a desktop CPU is an investment lasting for 3-8 years. Attempting to factor in predicted degradation of performance results based on history would be interesting and I think the right thing to do.


Those results are not valid anymore.

Better phrased: won't be valid for long, if the security patches are released ;)




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: