Updated AMD Ryzen and EPYC CPU Roadmaps March 2020: Milan, Genoa, and Vermeerby Dr. Ian Cutress on March 5, 2020 6:40 PM EST
Everyone is interested in roadmaps – they give us a sense of an idea of what is coming in the future, and for the investors, it gives a level of expectation as to where the company might be in a year to five years. Today at AMD’s Financial Analyst Day, the company gave the latest updates on the CPU side of the business, for consumer and for enterprise.
AMD stated that its CPU roadmaps for its enterprise portfolio are going to offer more vision into the future than its consumer side for a couple of reasons. First, the enterprise market is built on a longer product cycle and it helps when planning these systems to know what is in the pipe publicly, but also from an investor standpoint where the enterprise market ultimately offers the bigger financial opportunity.
To that end, AMD confirmed what we essentially knew, with Zen 3 based Milan coming in ‘late 2020’.
Zen 4 based Genoa has already been announced as the CPU to power the El Capitan supercomputer, and in this roadmap AMD has put it as coming out by 2022. We asked AMD for clarification, and they stated that in this sort of graph, we should interpret it as the full stack of Genoa should be formally launched by the end of 2022. Given AMD’s recent 12-15 month cadence with the generations of EPYC, and the expected launch of Milan late this year, we would expect to see Genoa in early 2022.
Astute users might notice that Milan / Zen 3 has been listed as ‘7nm’, where previously it was listed as ‘7nm+’. We’ve got a whole news post on why AMD has made this change, but the short of it is that AMD initially put ‘7nm+’ to mean ‘an advanced version of 7nm’. When TSMC named its EUV version of 7nm as N7+, people had assumed they were the same, and AMD wanted to clarify that Milan is on a version of 7nm, and the exact version will be disclosed at a later date. In the future the company will avoid using ‘+’ so this doesn’t happen again (!). We also have Genoa listed as a 5nm product.
Harder numbers about Milan and Genoa are expected to be unveiled closer to their respective launch times.
On the consumer side, AMD said a little less, with its roadmap only going out to Zen 3, which has the codename ‘Vermeer’ for the desktop product.
In this graph, we see that the Zen 3 product here is on the far right, but so is the date – 2021. Does this mean Zen 3 for consumers is coming 2021? We asked AMD to clarify, and were told that we should interpret this as that the range of Zen 3 consumer products, such as desktop CPUs, HEDT CPUs, mobile APUs, and consumer APUs, should all be available by the end of 2021. The company clarified that Zen 3 will hit the consumer market ‘later this year’, meaning late 2020.
So here comes a poignant question – what is going to come first in 2020? Zen 3 for enterprise is listed as ‘late 2020’, and Zen 3 for consumer is ‘later this year’. AMD makes a lot more money on its enterprise products than its consumer products, and while it enjoys a healthy performance lead in both, it really wants to push its market share in enterprise a lot more to drive home the bigger financial potential. With this in mind, I highly suspect that given AMD’s lead in the consumer market, we might see the company push more of its Zen 3 silicon into the enterprise market as a priority, with only a limited 2020 consumer release. I could be wrong, but we will find out closer to the time.
Interested in more of our AMD Financial Analyst Day 2020 Coverage? Click here.
Post Your CommentPlease log in or sign up to comment.
View All Comments
Gondalf - Friday, March 6, 2020 - linkThe bloody massacre is ending fast. Intel will beging to deliver Ice Lake SP to big OEMs at the middle of 2020 and the offical "end of 2020" launch will be only for common people. Datacenters will are satisfy a lot earlier, as usual.
But right now exist the blody massacre, AMD is in the unconfortable situation that Companies purchase only 64 cores SKUs from AMD, leaving absolute dominium to Intel up to 32(28) cores. The reason is simple, AMD Epyc is unable to clock fast enough to beat Intel in the very large sub 32 cores market. Cloud is only a portion of overall server market (15%), in many other applications a many core SKU is absolutely not the first choice, but the per core performance prevail by an huge margin.
Unfortunately TSMC 7nm is unable to clock higher than 3.3-3.4Ghz in Epyc, Intel is shipping server SKUs with 4-4.4Ghz of turbo. No competition outside cloud market, 14nm is far more efficent and great preformer.
The_Countess - Friday, March 6, 2020 - linkintel's 4.4ghz turbo for a single core, while having overal twice the power draw per core. basically almost nobody in the server market cares.
TOC is where it's at and the TOC of intel chips is ATROCIOUS compared to any AMD Rome chip.
edzieba - Friday, March 6, 2020 - linkTOC is mostly governed by software, which with per-core pricing being the norm is not favourable for AMD.
Walkeer - Friday, March 6, 2020 - linkyou live in some alternative universe, man.
schujj07 - Friday, March 6, 2020 - linkZen2 was designed to compete with Ice Lake not Cascade Lake. Right now AMD has the IPC advantage against Cascade Lake and with Ice Lake Intel will only catch up to pull a little ahead of Zen2. However, Ice Lake won't be going up against Zen2, it will be going up against Zen3.
In benchmarks when Epyc 2 was released we saw the 32 core 7502 dominating the best Intel 8280. The 8280 has a 200Mhz higher base clock and 650MHz higher boost clock. The 7502 isn't even the fastest 32 core AMD has, that would be the 7542. We also see benchmarks where the 7402p goes up against the Intel 8268 and win more benchmarks that it loses all while having a 100MHz base clock and 550MHz boost clock disadvantage.That is what the IPC advantage does, it lets the CPU compete against something that is clocked 10-20% faster. We already know that Intel cannot get past 4GHz boost with a 4c Ice Lake. What is Intel going to do when it has a CPU with 7x more cores?
We also know that TSMC can clock Epyc higher than 3.4GHz as Threadripper 3970x clocks 3.7GHz base and 4.5GHz boost. The only issue is the TDP goes from 180W on the 7502 to 280W on Threadripper.
Korguz - Friday, March 6, 2020 - linkyea ok sure gondalf, keep shilling for intel you have proof of this ?? amd doesnt need high clocks to compete with intel, intel needs the high clocks to compete with amd.
Spunjji - Friday, March 13, 2020 - linkOh, Gondalf. As reliable a source of pure gibberish as ever.
You're right that Intel will be "beging" to deliver Ice Lake. Indeed, they'll be pleading, but that won't make it happen.
I know it was a typo, but I couldn't resist. xD
prophet001 - Friday, March 6, 2020 - linkThe core clock king is still Intel.
That's all gamers really care about if they're pushing performance.
Skiddywinks - Saturday, March 7, 2020 - linkNot really. They care about single core max performance. In this case Intel are still head of the pack thanks to the significant clock lead, but AMD are ahead in the IPC race. The combination of these is what is important. If AMD can just bring up their clocks with Zen3, they would have the lead. With clocks and IPC gains, they are looking to be the convincing winner once we have seen what both teams have to offer.
Spunjji - Friday, March 13, 2020 - linkOnly a very, very specific subset of high-refresh-rate gamers. Most of the relevant performance comes from the GPU, and AMD CPUs aren't a bottleneck on that like they used to be. The rest of us care more about overall performance and value for money, and Intel suck on both of those metrics.