10-25-2010, 09:00 AM

As the title says, what company created the first processor with more than 1 core that was available on a computer? I've been looking it up online, but can't find a solid answer. Anyone know on here? I keep getting answers between intel, IBM, and even AMD.

10-25-2010, 09:11 AM

POWER4 - Wikipedia, the free encyclopedia

Regards

10-25-2010, 09:19 AM

The good folks at IBM introduced the first true multi-core processor with their Power4 systems. These were based on PowerPC instructions.

Raz0rEdge beat me too it.

Note: the IBM Power7 blade system I'm building this morning has:
4 Cores
4 Virtual processors per core
4 Logical processors per virtual processor
Totalling 64 logical CPUs worth of processing functionality all running on AIX 6.1.6 with live partition mobility. Fun stuff!

10-25-2010, 10:26 AM

Originally Posted by XJ-linux

Note: the IBM Power7 blade system I'm building this morning has:

CPUs worth of processing functionality all running on AIX 6.1.6 with live partition mobility. Fun stuff!

Huh? *Scratches head*

10-25-2010, 04:37 PM

Thanks a bunch for the links. I was talking to a friend who mentioned that the powerpc was a processor on the power mac that was a breed of processors based off of one of the first multi-core systems that were used as servers/super computers. I didn't believe it, but I guess it's true. Nice thing to know. That's one more reason for me to like IBM and Apple

10-25-2010, 04:56 PM

Originally Posted by XJ-linux

The good folks at IBM introduced the first true multi-core processor with their Power4 systems. These were based on PowerPC instructions.

Raz0rEdge beat me too it.

Note: the IBM Power7 blade system I'm building this morning has:
4 Cores
4 Virtual processors per core
4 Logical processors per virtual processor
Totalling 64 logical CPUs worth of processing functionality all running on AIX 6.1.6 with live partition mobility. Fun stuff!

What's the point of 64 processors when 58 of them sit idle?
My point is can you utilise all of them? of it is just massive over kill "just case you can"?

10-25-2010, 05:53 PM

If it's a server, yes, he is more likely to use them all, and not need a whole lot of speed per process, just the ability to run ridiculous numbers of simultaneously processes.

10-25-2010, 05:54 PM

Originally Posted by idrinorbarsaku

Thanks a bunch for the links. I was talking to a friend who mentioned that the powerpc was a processor on the power mac that was a breed of processors based off of one of the first multi-core systems that were used as servers/super computers. I didn't believe it, but I guess it's true. Nice thing to know. That's one more reason for me to like IBM and Apple

Hmmm... single core PPC was around long before the multicore variants. Also, it was a joint venture between Motorola and IBM -- although Apple was one of the earliest adopters, mostly because they relied upon Motorola's earlier CPU architecture. It essentially superseded the 68K line of 16/32-bit CPUs and was entirely RISC-based. In a perfect world, it would have also superseded x86, which it was vastly superior to upon release. Alas, the Wintel world prevailed and the PPC eventually faded mostly into the embedded market. But for a time in the mid-90's, it looked like PPC was the future - and it should have been. Even Microsoft produced variants of Windows NT that ran natively on PPC.

10-25-2010, 06:07 PM

Haven't GPUs been multicore for a looong time? Or were they multichip and some how voodooed together on the main board?

10-25-2010, 06:15 PM

It was ARM, 1983. 1986 was the date the first real ARM cores hit the market though, with the ARM2

10-25-2010, 06:19 PM

Originally Posted by the8thark

What's the point of 64 processors when 58 of them sit idle?
My point is can you utilise all of them? of it is just massive over kill "just case you can"?

The point is to put an entire SCM (Supply Chain Management) solution on a single box so hardware assets can be shared. We virtualize many different servers and run them concurrently on the same box. The system uses on average about 75% of the CPUs when I tune it correctly. It hits 90% at month end audit time. By having so many, certain components can dynamically borrow spare CPU cycles from other systems on the same box and run at 200%, 500%, 900% of their entitled capacity if needed. This one will have 4 SAP SC application servers, 2 IBM virtual I/O disk servers, 2 IBM virtual I/O network servers, a SAP central instance, and an Oracle 11g database server all virtualized and plugged into our Nexus core via an aggregated 40Gb/Sec ether-channeled connection.
Then we will build another identical system for fail over purposes.

10-25-2010, 07:55 PM

Originally Posted by XJ-linux

The point is to put an entire SCM (Supply Chain Management) solution on a single box so hardware assets can be shared. We virtualize many different servers and run them concurrently on the same box. The system uses on average about 75% of the CPUs when I tune it correctly. It hits 90% at month end audit time. By having so many, certain components can dynamically borrow spare CPU cycles from other systems on the same box and run at 200%, 500%, 900% of their entitled capacity if needed. This one will have 4 SAP SC application servers, 2 IBM virtual I/O disk servers, 2 IBM virtual I/O network servers, a SAP central instance, and an Oracle 11g database server all virtualized and plugged into our Nexus core via an aggregated 40Gb/Sec ether-channeled connection.
Then we will build another identical system for fail over purposes.

In our case, we run Citrix servers with many cores because they host a full desktop environment for hundreds and sometimes thousands of employees. The servers are grouped together in a farm. Sometimes a single server will play host to 50+ users, connecting from thin client terminals or desktop PCs running a Citrix client. As you might imagine, add a few dozen power users onto a server and they can eat up those resources pretty quickly.

Our newest iteration is virtualized, with lots of small virtual servers running on a few very large physical servers. In this way, we can scale up by adding virtual servers to the load, but that comes at a cost to the host server's cores.