AMD CEO, Dr. Lisa Su, states that Moore's Law is not dead and that innovations such as chiplets & 3D packaging will help overcome the challenges.
Moore's Law Is Not Dead, Says AMD's CEO: Working On 3nm, 2nm & Beyond With Latest Innovations
In an interview with Barron's, AMD CEO, Dr. Lisa Su, points out that Moore's Law is not dead but has slowed down and things need to be done differently to overcome the performance, efficiency, and cost challenges. AMD has been the pioneer of advancing 3D packaging and chiplet technology with its first HBM designs back in 2015, chiplet processors in 2017, and also the first 3D packaging on a chip with its 3D V-Cache design in 2022.
Moore's Law was proposed by the co-founder of Intel, Gordon Moore, back in 1965 (~1975) and said that the number of transistors would double every year. Gordon Moore passed away on the 24th of March but his legacy still lives in the tech world with Intel and AMD still standing by his law. Meanwhile, NVIDIA thinks that Moore's Law has run out of gas and no longer applies to its business strategy.
AMD's CEO states that chiplets and 3D Packaging are the solutions that their company has invested in, today, and there's a lot more on the way. The chipmaker will be launching its MI300, a true exa-scale APU that combines various CPU, GPU, and Memory IPs in several chiplets and 3D dies on one singular package. It's a gargantuan chip and will be leading AMD's charge into the AI segment later this year.
I would certainly say I don’t think Moore’s Law is dead. I think Moore’s Law has slowed down. We have to do different things to continue to get that performance and that energy efficiency.
We’ve done chiplets—that’s been one big step. We’ve now done 3-D packaging. We think there are a number of other innovations, as well.
Software and algorithms are also quite important. I think you need all of these pieces for us to continue this performance trajectory that we’ve all been on.
AMD CEO, Dr. Lisa Su
Lisa also mentioned that despite the increasing cost & lowering gen-over-gen performance benefit coming from each transitioning process node, they will still continue to move forward. AMD says its already working with 3nm as of right now and that they are also looking at 2nm and beyond that. In a similar strategy, this would lead to the Angstrom era or sub-nm which on the Intel side begins at 20A & 18A.
Yes. The transistor costs and the amount of improvement you’re getting from density and overall energy reduction is less from each generation. But we’re still moving [forward] generation to generation. We’re doing plenty of work in 3 nanometer today, and we’re looking beyond that to 2 nm as well. But we’ll continue to use chiplets and these type of constructions to try to get around some of the Moore’s Law challenges.
AMD CEO, Dr. Lisa Su
Currently, AMD has products utilizing the 5nm and 4nm process nodes with a few 6nm and 7nm products in between. Starting next year, the company is expected to introduce its first 3nm chips which are likely going to be targeted at the server segment first before the company intros it to the client side.
Many PC companies have pushed back orders of products based on TSMC's 3nm node due to the current market conditions and AMD was quoted to be among those. It will be interesting to see what choice of nodes AMD uses for its next-gen CPU/GPU/APU architectures since TSMC has a broad range of solutions as unveiled in its recent roadmap.
AMD Zen CPU / APU Roadmap:
| Zen Architecture | Zen 6 | Zen 5 (C) | Zen 4 (C) | Zen 3+ | Zen 3 | Zen 2 | Zen+ | Zen 1 |
|---|---|---|---|---|---|---|---|---|
| Core Codename | Morpheus | Nirvana (Zen 5) Prometheus (Zen 5C) | Persphone (Zen 4) Dionysus (Zen 4C) | Warhol | Cerebrus | Valhalla | Zen+ | Zen |
| CCD Codename | TBA | Eldora | Durango | TBC | Brekenridge | Aspen Highlands | N/A | N/A |
| Process Node | 3nm/2nm? | 4nm/3nm | 5nm/4nm | 6nm | 7nm | 7nm | 12nm | 14nm |
| Server | EPYC Venice (6th Gen) | EPYC Turin (5th Gen) | EPYC Genoa (4th Gen) EPYC Siena (4th Gen) EPYC Bergamo (4th Gen) | N/A | EPYC Milan (3rd Gen) | EPYC Rome (2nd Gen) | N/A | EPYC Naples (1st Gen) |
| High-End Desktop | TBA | Ryzen Threadripper 8000 (Shamida Peak) | Ryzen Threadripper 7000 (Storm Peak) | N/A | Ryzen Threadripper 5000 (Chagal) | Ryzen Threadripper 3000 (Castle Peak) | Ryzen Threadripper 2000 (Coflax) | Ryzen Threadripper 1000 (White Haven) |
| Mainstream Desktop CPUs | Ryzen **** (Medusa) | Ryzen 8000 (Granite Ridge) | Ryzen 7000 (Raphael) | Ryzen 6000 (Warhol / Cancelled) | Ryzen 5000 (Vermeer) | Ryzen 3000 (Matisse) | Ryzen 2000 (Pinnacle Ridge) | Ryzen 1000 (Summit Ridge) |
| Mainstream Desktop . Notebook APU | TBA | Ryzen 8000 (Strix Point) Ryzen **** (Krackan Point) | Ryzen 7000 (Phoenix) | Ryzen 6000 (Rembrandt) | Ryzen 5000 (Cezanne) Ryzen 6000 (Barcelo) | Ryzen 4000 (Renoir) Ryzen 5000 (Lucienne) | Ryzen 3000 (Picasso) | Ryzen 2000 (Raven Ridge) |
| Low-Power Mobile | TBA | Ryzen 8000 (Escher) | Ryzen 7000 (Mendocino) | TBA | TBA | Ryzen 5000 (Van Gogh) Ryzen 6000 (Dragon Crest) | N/A | N/A |
