Andrei F (andrei.delete@this.anandtech.com) on November 22, 2020 4:08 am wrote:
> Chester (lamchester.delete@this.gmail.com) on November 21, 2020 2:49 pm wrote:
> > Some interesting things:
> > - Zen 3 beats Intel on full random mem access latency despite using
> > a separate IO die. Maybe 6 page walkers are playing a role.
>
> Page walkers have nothing to do with it, you can't page-walk ahead of an actual proper
> working pointer-chase pattern. The Sandra test is just broken and showing nonsense.
>
> https://www.anandtech.com/show/16214/amd-zen-3-ryzen-deep-dive-review-5950x-5900x-5800x-and-5700x-tested/5
>
> The real data shows that there's essentially no change between Zen2 and Zen3 in cycles,
> which is expected given it's the exact same IOD, and Intel is still ahead in latency.

How does AT's methodology differ from Sandra's, and why is Sandra's latency lower? I'm not saying AT's wrong (you're probably right, - I highly doubt latency changed like that with the same IOD). I'm just curious what's actually being measured.

> > Try scrolling down a bit. The little bar on the right of your browser window can do that - try dragging
> > it down or pushing the down arrow at the bottom, and you'll see a whole lot more than rendering.
> >
> > There's compression, gaming, video editing, image editing, as well as a couple
> > of synthetics. Not the most complete coverage, but miles better than SPEC.
>
> Yea, if you mean miles better means 1/4th coverage of the subtests.

Since most/all of the subtests are irrelevant and meaningless for consumers, yeah I consider it miles better. There's a reason SPEC isn't popular among review sites. No one cares about it.