Intel CEO Bracing For EPYC Impact, Aims to Keep AMD Under 20% of Server Market Share

[Tarkin77]

https://epycly.com/ another Epyc Cloud

 

[Cerb]

*Cough* whew, that's dusty. Been off in shinier, more distracting, RGB pastures.

But, anyway, my two cents: can AMD get 20% of the market, *even if there is enough demand*? I doubt it. Sounds very much like CYA from Intel, to me. "We're going to make sure that AMD doesn't do the impossible, you just watch." That's supposed to be impressive, intimidating, or a confidence booster? Yeah...

Even with the added latency, many workloads should do fine with the MCM, and AMD doesn't need to take over the market. With improved performance regarding Meltdown and Spectre, along with some models offering better density on a per-core/thread-per-dollar-per-U/per-sqft basis, surely there is a hefty amount of potential demand with distributed computing (especially business intelligence, and big data analysis), and generic shared servers. Consider that most users of hosted VMs don't care about performance, even if the VM on an EPYC MCM runs slower than a comparable Intel chip. If it serves your requests as well as the hardware from several generations ago did, most users will not even care what's going on behind the curtain. The hoster may notice the ability to improve density, and in some cases significantly reduce supporting infrastructure, and that matters much more, most of the time, even if it's a 5% or 10% improvement. The heavy users can go into whatever set of servers is known to be more stable and/or performant, for now. With the added IO resources available, and no Meltdown, I could see network switching and local storage bound loads, or servers solely handing those loads, doing especially well.

But, it would be quite a feat if they could even get enough production ramped up to meet 10%, at least in the next couple of years. OTOH, if they can get high margins for those parts, 5% of the server market could well mean close to a billion in added revenue, and possibly a quarter or more of a billion profit, which they could reinvest quite well. VM hosters can easily buy enough to gill many racks, and make sure everything works well, while also making money on them, to get the ball rolling, and see about any hypervisor quirks that will need ironing out.