Microsoft Won't Charge More for Multicore Licenses

esimp writes "According to technewsworld: 'As servers with dual-core processors come closer to hitting the market, Microsoft announced today it will not base its per-processor software licensing charges on the number of cores in a chip, sticking to the traditional price per processor, regardless of its number of cores." Update: 10/20 00:37 GMT by T : One of the identical links to TechNewsWorld's story has now been deleted.

Comments

Re:Stupid question

[Anonymous Coward]

In a quick nutshell.

Essentially a chip with more than 1 CPU on it.

Instead of having a dual CPU with 2 fully seperate Xeons (for example), you now (in 1-2 years) will have a single Xeon that looks to the OS like 2 seperate CPUs.

The part of the CPU that contains the real logic is called the core, and the cache and interface stuff is well the non-core. So, they put the heart of 2 CPUs on a single chip and wrap 1 non-core cache & bus interconnect around it, and call it a dual-core CPU, or multi-core to be generic.

They make some changes in the bus interconnect to support this of course.

You'll see it in high end server chips at first and then it'll work its way down to the desktop. Business care about the per CPU licensing because that is usually how they pay for software.

Re:What's today's date?

[Anonymous Coward]

Is today April 1?

No, it's October 19th. It's clearly marked right below the subject.

{and if the mods mark this informative i'll kill my self}

Re:Stupid question

[rborek]

A multicore processor is a processor with more than one processor core in a single die. So, an example of a multicore processor would be putting two Pentium 4 processor cores in the same processor die, thereby giving the operating system two processors to work with, instead of just one. This is roughly the same as having a dual-processor system, except that because the processor cores are side-by-side they can interact at processor speeds, rather than bus interconnect speeds. It's one method of creating faster computers without having to create new processors, or continually speed up the processor by increasing the frequency.

The Xeon processors will be the first Intel chips to use multicore processors, and will eventually make its way into mainstream chips.

Re:Obviously

[slash-tard]

Redhat ES vs AS server are based on CPU.

Re:Microsoft did the same with Hyperthreading

[Zeever]

There's one difference: HT (Intel's SMT) is a way to use the processor resources more efficiently. Dual Core is like having two real CPU's (not exactly, because they share lots of stuff). The performance difference between HT and Dual Core is abysmal.

Re:Obviously

[Martin Blank]

Fewer pissed-off home users who don't have to pay an extra fee to use the second 'CPU,' I would imagine.

But why is this news? Microsoft confirmed this back when Hyperthreading first came out. They were charging on the basis of sockets, not cores.

Re:Obviously

[Vaevictis666]

As mentioned on The Register, they're not doing this so much for the OS side of things, but for the server software. Oracle charges per core, but is MS SQL Server charges per processor, that's half the cost right there. If Oracle or whoever switched models, that's more or less half gross revenue from that product gone.

Suggested reasoning for this was that we didn't charge more when processors increased in speed by upping the clock rate, so why do it when processors increase speed by adding more cores on the die?

Re:Both links are the same!

[metlin]

For subscriber preview, you have an e-mail address you can mail to if you notice any errors/mistakes that you may see in the story.

That e-mail address happens to be daddypants @ slashdot.org, hence the parent poster's comment.

Re:Multi core processors, or multi processor cores

[Carnildo]

With multicore, the CPUs are sharing a single memory bus. At the two-core level, this isn't too much of a performance hit, but by the time you hit four cores, you lose most of the benefit of that fourth core to the lack of memory bandwidth.

Intel's Xeon chips are running into this problem already. A single Xeon CPU has better memory performance than a single Opteron, but a four-way Opteron system, with a separate memory controller and RAM bank for each chip, blows away a four-way Xeon system, since the four Xeons have to share the memory controller and memory.