Let’s get started with AMD. We’re almost a year into the release of their super successful Zen 2 desktop cpu lineup. And of course, as with any products… we’re just waiting to hear about the next one: Zen 3. So far, all we’ve known is really just that AMD has a minimum target for IPC improvements of about 8% for every 12 to 18 months.
Well, according to a post by Ice universe on the Chinese platform Weibo. AMD is going way beyond that. Now, while ice universe is a known leaker, you still need to take that info with a grain of salt. The original post was translated by retiredengineer on twitter. Essentially its some gossip between some engineers at 2 different OEM’s on the upcoming intel vs AMD fight, and zen 3.
Apparently the engineering samples have risen to an alarming level of performance. And judging from current information available on intel’s upcoming products, Ryzen 4000 CPU’s for the desktop will beat out intel in power efficiency and performance. It kind of goes without saying for power efficiency since intel will still be using 14nm’s for rocket lake S, which is the generation that is supposed to follow comet lake S. The only difference is that rocket lake S will be a backport of tiger lake from 10 to 14nm.
And lastly, in that gossip, it says that AMD calculated a theoretical improvement of 15 to 17% in IPC for zen 3. That is already a ton of performance uplift, especially since the node doesn’t change, But in testing, these engineers say that the performance seemed even better than that. This goes in line with what Forrest Norrod said in the past. He said that zen 2, had an abnormally high IPC uplift, but that Zen 3’s IPC gain would be in line with an entirely new architecture. Anyways, once again, grain of salt, but I mean it does seem achievable from AMD. What do you think?