Intel and AMD are currently in a core count stalemate thanks to the launch of Intel’s Coffee Lake refresh and the I9-9900K. While it is considerably more expensive than that of the AMD Ryzen 7 2700X, it also performs better (particularly when single thread performance is concerned) – but times are changing.
Details concerning Intel’s 10nm delays are well infamous, so I won’t use words to describe them. But suffice to say, Intel didn’t still expect to be stuck on a 14nm process anywhere near as long as they have been.
The company recently detailed the “Sunny Cove” microarchitecture, and processors based on Sunny Cove will have a wider execution pipeline, more efficient, higher clock speeds and it will be a noticeable evolution to the currently rehashed Skylake architecture that Intel’s released back in the summer of 2015. How the Sunny Cove architecture will evolve over the desktop and future processor evolutions isn’t yet clear though [edited for clarity].
I was told by a very reliable source (who told me information about the AMD 500 series before AMD’s announcement and other things) that Intel will indeed be launching Comet Lake for desktop and not a form of Ice Lake [corrected typo here where I incorrectly said Ice Lake SP – sorry about that!], but Comet Lake is 10 cores for the mainstream parts, being produced on Intel’s 14nm Process [EDIT – correction from my source] Intel’s goals appear to be to shoot for the highest possible clock / IPC for those cores as possible.
Intel will then follow up Comet Lake with Tiger Lake, which the company are targeting for launch in 2021 [EDIT since article was live as source provided extra information].
The alarming thing is that System Integrators (OEM’s) and manufacturers are concerned with Intel’s ability to compete with this new architecture with AMD’s new Ryzen 3000 (Matisse). While Intel will do well in a pure single thread metric (as always) AMD’s core count advantage and pricing strategy could prove a huge blow against Intel.
This will make for a very interesting situation for end users, and could also affect the pricing we see AMD choose for the various Matisse parts, as without competition from Intel, AMD have less incentive to be ultra aggressive.
I was also told that the launch date of Ryzen 3000, X570 and Navi will be July 7th (get it, 7nm?) with an announcement at Computex for all three products. My source has said that he is less certain about Navi because AMD is under pressure from TSMC’s manufacturing capabilities. I had heard from another source the company were aiming for a July launch, but then a third source said that this might be delayed until October because of TSMC.
The latest news is that AMD is trying to pull every trick in the book to get the cards on shelves by July (likely concerned by the GTX 16 series). I wouldn’t be surprised if we see some shortage of supply though (as a warning) but that’s just speculation on my part.
July for the Matisse / Ryzen 3000 / X570 makes sense given Lisa Su herself pointing to the company targeting late Q2 for the product launch. I was told that Reynoir (the next generation APU) was going to release in the second half of this year (2019) [EDIT – originally my source said this was ‘Rhino’ but it was an English mixup on his part and I’ve edited to correct]
I was also told that SI / OEMs are expecting at LEAST 12 cores from AMD (not surprising given what we’ve already covered regarding the engineering sample leaks of the 12 core Matisse part), but manufacturers are almost certain there’ll be 16 cores in the AM4 lineup. There are no engineering sample processors in the hands of vendors (OEM’s or motherboard) yet, and they’ll not have them until at least April (that information is according to two different sources).
I do wonder (as a bit of speculation) if AMD will opt to hold back the 16 core processors initially, both because of yields and also to sandbag and then counter Comet Lake with a 16 core Matisse part. So essentially 12 core Ryzen 3000 will fight the 8 core I9-9900K, with the 10 core Comet Lake having the displeasuring of facing the might of a 16-core Zen 2 Matisse. This is just a possible scenario though, and nothing my source has said, but would be about the worst case for customers (I don’t think AMD will choose to do this though).
I was also told by two separate sources AMD is targeting higher clock frequencies for Ryzen 3000 (unsurprising given what we know) but also that the motherboard platforms will need to be built (on the higher end motherboard SKUs) with support for more overclocking features. This likely means that robust VRM design and other usual considerations will be very important – and could also have interesting consequences for B550 overclocking (though I wasn’t given any details of that).
All of this means that for at least 9 ish months AMD will be fighting Intel with a Zen 2 powered Matisse lineup, while team blue only can put up the I9-9900K for a fight… which isn’t a very good sign at all. While it’s great AMD are able to compete with Intel, it will make a real turn the tables, where as back 3 years ago it was almost a “which Intel CPU do you recommend” instead of the current “which manufacturer and processor is best for your usage scenario?”
For at least 9 months (three quarters) AMD could simply dominate Intel in all areas – and given the IPC isn’t that much different between a Ryzen 7 2700X and a I9-9900K (as we discovered in our IPC testing late last year), a Zen 2 processor with 10 – 15 percent IPC gains, and a 4.5 – 5GHZ core clock would also take Intel’s gaming advantage away from them to.
A cheap 8 core, 16 thread Matisse processor that costs 100 – 200 USD (especially as AMD motherboards are cheaper) would essentially mean that no one would purchase an Intel processor. I was told though by the same source that Intel’s future processor roadmap (after Comet Lake, moving into Tiger Lake) is very impressive, but that doesn’t help right now.
AMD have managed to execute rather well since the original Zen launch, and after 7nm Zen 2 we have the company launching Zen 3, which is a 7nm+ version, with very modest IPC tweaks but with the aim to reduce power consumption further still from Matisse and also improve clocks (from a refined 7nm process).