> > I'm gonna criticize this test for the opposite reason juanrga criticized the other one: this person
> > didn't match memory speeds. Intel was tested with DDR4-2933, AMD was tested with DDR4-3200. When
> > comparing architectures, it's best to keep variables constant as much as possible.
> >
> BTW for Matisse and Vermeer the supported speed is the same (DDR-3200) and the same was used for both so the
> Zen 3 IPC boost is not influenced by this, only Skylake's IPC is

Matisse/Vermeer only support up to DDR4-3200 because the JEDEC standard stops there. Skylake only supporting DDR4-2933 is rather arbitrary.

But both AMD and Intel are regularly run with faster memory. Insisting on sticking to 'stock' memory speeds is a misguided refusal to acknowledge the popularity of XMP. Maybe juanrga's position made sense 15 years ago when faster than officially supported memory wasn't the default. Times have changed though.

Review sites today test with DDR4-3200 and faster simply because those are common configurations. Testing with slower memory only makes sense on locked Intel setups that enforce a memory speed limit.

> (but Skylake is not so interesting, is it).

Skylake is interesting because it's still Intel's fastest desktop architecture. Sunny Cove has better IPC and is ok-ish on low power/mobile, but on 10nm it can't reach clocks high enough to replace Skylake.

SKL losing to AMD architectures is just a demonstration of Intel's dated architecture falling further behind.