Hacker News Reader

Comment by garaetjjte | Hacker News Reader

>In this case, an EPYC 7281 in single socket mode is listed as having +63% performance (in SPECint) over a dual socket E5-2609v4 system.

So, quad-CPU is faster than dual-CPU? Not surprising.

View on HN

It's also cheaper and uses a lot less power, which is pretty surprising.

I guess the point is AMD offers significantly more bang per buck (and socket) at similar prices.

The per-socket performance might be a nice "marketing hack", since a lot of big-iron software is sold and licensed as "per-socket".

IIRC, Windows Server is sold per-core however. So lots of cores may rise the total-cost of ownership in the case of Windows Server.

Good thing most servers use Linux and other open source software. I actually read something recently that Microsoft almost got in trouble during the antitrust days for trying to sell Windows licenses "per processor." They seem to be doing exactly that. I guess they've gotten bolder since the governments stopped monitoring them closely.

The rules are different depending on whether you have a dominant market position or not. Microsoft clearly does not have a dominant market position on server hardware, so they can use "creative" licensing there.

Most enterprise database vendors now charge on a "per-core" basis. There are some OLAP / NoSQL vendors that are moving to a per XX GB Ram model.

How are these vendors getting away with this? Is their product/support so much better than open source + third party support, or are they entrenched?

Inertia and vendor lock-in are really powerful forces. Previous decision makers chose a Microsoft stack, and those who came before them did the same. Our third party services provider was a staunch Microsoft supporter. Before anyone knew it, we were completely locked in to a Microsoft ecosystem, paying tens of thousands per core for new SQL Server licenses.

Reply on news.ycombinator.com