My biggest question is... what percentage of the market is really buying the latest generation of processors upon release? Gamers, sure. I'd imagine the number of "every day use" laptops sold is a lot larger than people trying to eek out the last few frames per second. Would these budget machines also have this new advanced AMD CPU? How long does it take "the kind of laptops you pick up at Walmart" to deprecate old CPUs and implement new ones?
>My biggest question is... what percentage of the market is really buying the latest generation of processors upon release?
That really isn't relevant. Most consumers buy a new computer when their old one breaks or becomes unbearably slow. Most corporate buyers have a consistent 3-5 year procurement cycle. AMD don't need everyone to rush out and buy new computers, nor do they particularly want it - huge spikes in sales are detrimental to the supply chain, as we saw with the GPU mining fad.
AMD want and need to consistently beat Intel on value over a protracted period, so that whenever a customer does choose to upgrade their computer they choose an AMD processor. At the moment, they're looking good in terms of both architecture and fabrication; the main problem AMD is likely to face is with partner relationships, because of Intel's historical dominance and use of anti-competitive practices.
Majority of data center users replace their gear within 3-5 years. There are outliers, of course, who replace hardware with each generation or those who run their servers into the ground, but 3-5 years is far more common, so 4 years on average.
This rough math means if AMD takes half of the new CPU orders, they should have about 10% server CPU market share within a year, consistent with some analyst estimates [1]
AMD couldn’t have timed bringing a genuinely competitive product to market either on the backside of the Meltdown and MDS and Intel struggling on process.
Will be interesting to see if intel begin pulling the same kind of shenanigans they did back the last AMD was competitive (early Opteron/Athlon days).
I don’t think the market structure is the same and given they got slapped for it last time by the regulators I’m curious to see if they’ll try.
Anecdotally I know several people whose 'main' desktop PCs are now right around 3 years old. Mostly stuff from one or two intel generations prior to the introduction of the Ryzen 1xxx series several years ago. After going through all of the reviews for the Ryzen 3xxx (3700, 3800X, etc) and X570 chipset motherboards, they've upgraded and are really pleased with the performance.
It's fun seeing something that can chew through kdenlive video editing of 2160p, 60fps video and encode a project into HEVC like a hot knife going through butter.
The only problem is that such increases in speed are really only needed either for gamers or a small subset of people who work with graphics/engineering/video.
Lots of other uses for an 8-core desktop workstation... Lots of guest VMs inside virtualbox, a host operating system such as Qubes which is a xen hypervisor, etc. Not just gaming and very cpu intensive things like 2160p60 video encoding.
For a lot of the server market, I suspect Intel, being a known quantity, will remain where it is. Bringing a new design to market is expensive (tooling, supply chain, warranty, maintenance, product life cycle, etc) and I'd need to know I'd be able to sell at least a certain number of boxes over a given time to make it worth it.
If all my tooling and supply chain is geared towards Intel on the server lines, it'll take a much better value proposal, volume guarantees, and the threat my competition will do it first to make me add a new line geared towards AMD.
OTOH, for cloud powerhouses that design their own hardware, even a marginal improvement in TCO/performance will tip the scales.
Then also the market isn't only about the highest-end CPU. There's plenty of market for lower-end CPU's. Both AMD and Intel have their offerings and production capacity. The market shifts will be marginal I think.
Comments
My biggest question is... what percentage of the market is really buying the latest generation of processors upon release? Gamers, sure. I'd imagine the number of "every day use" laptops sold is a lot larger than people trying to eek out the last few frames per second. Would these budget machines also have this new advanced AMD CPU? How long does it take "the kind of laptops you pick up at Walmart" to deprecate old CPUs and implement new ones?
>My biggest question is... what percentage of the market is really buying the latest generation of processors upon release?
That really isn't relevant. Most consumers buy a new computer when their old one breaks or becomes unbearably slow. Most corporate buyers have a consistent 3-5 year procurement cycle. AMD don't need everyone to rush out and buy new computers, nor do they particularly want it - huge spikes in sales are detrimental to the supply chain, as we saw with the GPU mining fad.
AMD want and need to consistently beat Intel on value over a protracted period, so that whenever a customer does choose to upgrade their computer they choose an AMD processor. At the moment, they're looking good in terms of both architecture and fabrication; the main problem AMD is likely to face is with partner relationships, because of Intel's historical dominance and use of anti-competitive practices.
Majority of data center users replace their gear within 3-5 years. There are outliers, of course, who replace hardware with each generation or those who run their servers into the ground, but 3-5 years is far more common, so 4 years on average.
This rough math means if AMD takes half of the new CPU orders, they should have about 10% server CPU market share within a year, consistent with some analyst estimates [1]
[1] https://www.digitimes.com/news/a20190328PD200.html
AMD couldn’t have timed bringing a genuinely competitive product to market either on the backside of the Meltdown and MDS and Intel struggling on process.
Will be interesting to see if intel begin pulling the same kind of shenanigans they did back the last AMD was competitive (early Opteron/Athlon days).
I don’t think the market structure is the same and given they got slapped for it last time by the regulators I’m curious to see if they’ll try.
Anecdotally I know several people whose 'main' desktop PCs are now right around 3 years old. Mostly stuff from one or two intel generations prior to the introduction of the Ryzen 1xxx series several years ago. After going through all of the reviews for the Ryzen 3xxx (3700, 3800X, etc) and X570 chipset motherboards, they've upgraded and are really pleased with the performance.
It's fun seeing something that can chew through kdenlive video editing of 2160p, 60fps video and encode a project into HEVC like a hot knife going through butter.
The only problem is that such increases in speed are really only needed either for gamers or a small subset of people who work with graphics/engineering/video.
Lots of other uses for an 8-core desktop workstation... Lots of guest VMs inside virtualbox, a host operating system such as Qubes which is a xen hypervisor, etc. Not just gaming and very cpu intensive things like 2160p60 video encoding.
then you can get a 3200g which has 4 fast cores and a decent gpu for $99! its amazing.
For a lot of the server market, I suspect Intel, being a known quantity, will remain where it is. Bringing a new design to market is expensive (tooling, supply chain, warranty, maintenance, product life cycle, etc) and I'd need to know I'd be able to sell at least a certain number of boxes over a given time to make it worth it.
If all my tooling and supply chain is geared towards Intel on the server lines, it'll take a much better value proposal, volume guarantees, and the threat my competition will do it first to make me add a new line geared towards AMD.
OTOH, for cloud powerhouses that design their own hardware, even a marginal improvement in TCO/performance will tip the scales.
Then also the market isn't only about the highest-end CPU. There's plenty of market for lower-end CPU's. Both AMD and Intel have their offerings and production capacity. The market shifts will be marginal I think.