AMD has released an investor presentation focused on the products the company will introduce later this year. The deck mostly rehashes the advances the company has already publicized regarding 7nm, Epyc versus Xeon performance comparisons, and the improvements to its Radeon Instinct family. There is, however, one slide that confirms 3rd Generation Ryzen will debut mid-year:
A mid-year launch for 3rd Generation Ryzen has always made the most sense given AMD’s communicated timelines. Mid-year likely means at the beginning of Q3 (July-September), though it’s possible that AMD could also go for a late Q2 launch in June. We should either hear about these products at Computex with a launch aligned to the show or see them announced at Computex, with the launch following shortly thereafter. It’s not clear if AMD will launch third-generation Threadripper immediately on the heels of Ryzen3(I can’t write “3rd Gen Ryzen” every single time I mention the same product or else the Journalism Gods and Mrs. Melvin from junior-year English class will punish me).
This isn’t a great layout for a roadmap slide. Aligning the “2017” and “2020” labels to the far-left and far-right sides of the graph, respectively, could throw off reader understanding of which periods those dates refer to in the graphic, rather than aligning each year to a consistent section of the relevant architecture.
All we know about Navi is that it’s coming “later” in 2019. There have been rumors of a summer launch, followed by rumors that said launch had been delayed. Midsummer to early fall still seems to be the most likely bet; AMD’s major GPU launches have all fallen into this time frame since the debut of GCN. Midsummer to fall also aligns with the availability of AMD’s 7nm Ryzen, and could be read as a sign that we’ll start seeing more silicon from TSMC’s 7nm node in general around this time frame.
The general rumor has been that Navi will debut in the mid-range first, with a higher-end version coming later. Whether the Navi launch intended for 2019 refers to the mid-range card or the entire product family isn’t currently known. We’re reading the “Next-Gen” part of the graph fairly conservatively as well: AMD has been using that phrase as stand-in for “Our next GPU launch” for quite some time, and while it’s interesting to hear that the company intends to adopt EUV for its next-generation GPU (TSMC’s 7nm+ makes use of that technology), EUV itself isn’t expected to deliver much in the way of power, performance, or area improvements its first time out the gate. It’ll likely be used for contacts and vias rather than throughout the chip, and TMSC doesn’t project major gains on that basis (up to 10 percent improved power consumption and 17 percent improved density, but no performance gains).