There are companies that are testing AMD hardware, who might make the leap in the next couple years. But for now, the server market is still very much Intel's playground.
It is, but this is changing steadily and consistently. Nothing moves fast in the server market, so no one should expect any magic numbers on annual basis.
The problem for Intel is not just AMD, but ARM custom chips, which are already moving past 10%.
By the end of the year, Intel might drop to ~60%, AND go up towards 25-27% adn ARM solutions 13-15%.
Intel is not going to have e-core based efficiency server CPU until H1 next year (Sierra Forest) to compete against ARM and Bergamo for cloud servers.
By the time Sierra Forest with 144 e-cores releases, ARM solutions will have another generation of increased foothold and AMD will release Turin dense with 192 cores.
"Inertia" is absolutely the right word to use. A corporation has a thousand Intel-based servers in a data center. Next year's budget could be spent converting to new hardware, which might be faster but might also have hidden issues... or they can stay the course, buying Intel because it's a known ecosystem that "just works".
Both AMD and ARM server solutions are in the process of creating a strong foothold. It's a long, painstaking process in server space due to complexity of installations, long-term support, maintenance, perks and loyalty contracts.
Intel's server ecosystem has never been less perceived as "it just works" than now. There are several reasons for this. Modern companies are well-informed about what's going on in CPU performance roadmaps and AMD and ARM solutions have conquered 30% of the market in just 3-4 years, which sounds astonishing, but it has explanations. It's a dynamic situation of steady change globally and there is nothing Intel could do to stop it because they do not have enough performant and diverse portfolio of CPUs. They can only slow down this process until they finally start releasing server CPUs on 4nm and 3nm in 2025 onwards.
Below you can see diversity of CPU portfolio. AMD has three different kinds of server CPUs for different workloads and utility, plus aggressive development roadmap. Intel has had too many 'stop-gaps' in generational development and their newest CPUs are still on 10nm process since 2021. AMD started EPYC journey in 2019 when they created the first server CPU with more than 60 cores which Intel only released this year. TSMC's advanced foundary works well for everyone and brings necessary competition pushing others to try to improve, and faster.
Also, there will be a new kid on the foundary block too. Japanese semi-con is back. They swigned a contract with IBM to develop 2nm process and reach parity with TSMC and Samsung by 2027. Quite ambitious goal, but there is no doubt that they will produce server chips for other companies
I thought Raptor Lake Refresh was still considered 13th gen? Meteor lake needs a new socket and platform.
Desktop refresh is 14th. Meteor Lake mobility is also 14th, on a new socket and platform.
Metero Lake lower desktop next year will be 15th, together with Arrow Lake higher SKUs.
I wonder how the recent AMD X3D chip burning debacle, and subsequent completely idiotic actions of partners will influence this?
There is no "debacle". A few isolated cases. It's less than Nvidia melting cable issue.
You know, Asus writing in fine print of Bios they encourage you to upgrade to that you don't actually have a warranty?
Asus showed the worst slimy practices possible. They will need to come straight on this.
Intel's server market is partly captured using very low prices, hence low margins and negative operating margin this quarter.
?
