Today is AMD’s Financial Analyst Day, the company’s semi-annual analyst-focused gathering. While the primary purpose of the event is for AMD to reach out to investors, analysts, and others to showcase the company’s performance and why they should continue to invest in the company, FAD has also become AMD’s de-facto product roadmap event. After all, how can you invest wisely in AMD if you don’t know what’s going to happen?
As a result, the half-day series of presentations is filled with little nuggets of information about products and plans across the company. Everything here is top notch – don’t expect AMD to hand out the Zen 4 transistor map – but it’s easily our best look at AMD’s product plans for years to come.
Kicking off FAD 2022 with what is always AMD’s most interesting update is the Zen architecture roadmap. The cornerstone of AMD’s recovery and revival into a competitive and capable player in the x86 processor space, the Zen architecture is the foundation of everything from AMD’s smallest embedded CPUs to their largest enterprise chips. So what’s in the pipeline in the coming years is a really big deal for AMD and the industry as a whole.
Zen 4: Improving performance and performance per watt, available later this year
AMD dives right in and is currently ramping up their Zen 4 architecture-based products. This includes the Ryzen 7000 (Raphael) client CPUs, as well as their 4″e generation EPYC (Genoa) server CPUs. Both will be launched later this year.
We’ve seen bits and pieces of information about Zen 4 so far, most recently with the Ryzen 7000 announcement on Computex. Zen 4 brings new core CPU chiplets and a new I/O chip, with support for features such as PCI-Express 5.0 and DDR5 memory. And in terms of performance, AMD is aiming for significant improvements in performance per watt and clock speed over their current Zen 3-based products.
Meanwhile, AMD is following up on that Computex announcement by clarifying a few things. In particular, the company answers questions about Instruction per Clock (IPC) expectations, stating that they expect Zen 4 to offer an 8-10% IPC improvement over Zen 3. The first Computex announcement and demo appeared. to imply that most of AMD’s performance was gains from clock speed improvements, so AMD is working to respond to that without showing too much of their hand months after product launches.
Coupled with that, AMD also reveals that they expect an overall single-thread performance gain of over 15% – with an emphasis on “greater than”. ST performance is a mix of IPC and clock speeds, so AMD can’t get overly specific at this point as they haven’t committed to final clock speeds. But as we’ve seen with their Computex demos, there’s currently 5.5GHz (or more) for Zen 4 on the table for lightly threaded workloads.
Finally, AMD also confirms that there will be ISA extensions for AI and AVX-512 for Zen 4. At this point, the company isn’t clear on whether one (or both) of those extensions will be in all Zen 4 products or just a subset — AVX-512, for instance, is a bit of a space and power hog — but it’s reasonable at the very least. to expect these to appear in Zen 4 server components. The addition of AI instructions will help AMD keep up with Intel and other competitors in the near term, as CPU AI performance has already become a battleground for chipmakers. But exactly what this does for AMD’s competitiveness depends in large part on what instructions (and data types) are added.
AMD will produce three flavors of Zen 4 products. This includes the vanilla Zen 4 core, as well as the previously announced Zen 4c core – a compact core that is for high-density servers and goes into the 128-core EPYC Bergamo processor. AMD also confirms for the first time that there will also be V-Cache-equipped Zen 4 parts – which, while new information, shouldn’t come as a surprise given the success of AMD’s V-Cache consumer and server parts.
Interestingly, AMD plans to use both 5nm and 4nm processes for the Zen 4 family. We already know that Ryzen 7000 and Genoa will be using one of TSMC’s 5nm processes and that Zen 4c chiplets will be built on the HPC version of N5. So it’s not immediately clear where 4nm fits into AMD’s roadmap, although we can’t rule out the possibility that AMD is playing a bit fast and loose with terminology here, as TSMC’s 4nm processes are an offshoot of 5nm (rather than an entirely new node). ) and are typically classified as 5nm variants to start with.
Currently, AMD expects a > 25% increase in performance per watt with Zen 4 over Zen 3 (based on desktop 16C chips with CineBench). Meanwhile, the overall performance boost stands at >35%, no doubt benefiting from both the architecture’s better performance per thread and AMD’s previously revealed higher TDPs (which are especially useful for uncorking more performance in MT workloads). And yes, these are terrible charts.
Zen 5 Architecture: All-New Microarchitecture for 2024
Meanwhile, the Zen 5 architecture, announced today, is with AMD’s Zen architecture roadmap to 2024. Since AMD isn’t shipping Zen 4 yet, their details about Zen 5 are understandably at a very high level. Nevertheless, they also indicate that AMD is not going to rest on their laurels and have some aggressive updates planned.
The big news here is that AMD calls the Zen 5 architecture an “all new microarchitecture”. That is, it won’t just be an incremental improvement over Zen 4.
In practice, no major vendor designs a completely new CPU architecture – there will always be something good enough for reuse – but AMD’s message is clear: they are going to significantly rework their core CPU architecture to improve performance and power efficiency. to improve further.
As for what AMD will reveal at this point, Zen 5 will re-pipeline the front-end and once again increase their problem width. The devil is in the details here, but coming from Zen 3 and its 4 instruction/cycle decoding speed, it’s easy to see why AMD would want to focus on that – especially when the integer units at the back are already a 10-wide issue width.
Meanwhile, on top of Zen 4’s new AI instructions, Zen 5 integrates further optimizations of AI and machine learning. AMD isn’t saying much else here, but they have a sizable library of tools to choose from, covering everything from AI-focused instructions to adding support for even more data types.
AMD expects the Zen 5 chipset to be similar to Zen 4 – that is, they will have the same trio of designs: a vanilla Zen 5 core, a compact core (Zen 5c), and a V-Cache enabled core. For AMD’s customers, this kind of continuity is very important, as it guarantees customers that AMD’s more custom configurations (Zen 4c & V-Cache) will have successors in the 2024+ time frame. From a technical standpoint, none of this is surprising, but from a business standpoint, customers want to make sure they don’t end up using dead-end hardware.
Finally, AMD has an interesting production mix planned for Zen 5. Zen 5 CPU cores will be manufactured on a mix of 4nm and 3nm processes, which unlike the 5nm/4nm mix for Zen 4, the 4nm and 3nm nodes of TSMC are very different. 4nm is an optimized version of 5nm while 3nm is a brand new node. So if AMD’s production plans continue as they are currently framed, Zen 5 will make a big node jump. That said, it’s not unreasonable to suspect AMD is hedging their bets here and leaving 4nm on the table just in case 3nm isn’t as far along as they’d like.
To wrap things up, the Zen 5 architecture is slated for 2024. AMD isn’t giving any further information on when that could be in the year, though looking at Zen 3 and Zen 4, both were/will be released later. in 2020 and 2022 respectively. So H2/EOY 2024 is as good as guessing.