Print 20 comment(s) - last by aGreenAgent.. on Feb 13 at 4:02 AM

The dual core "Conroe" CPU features two cores on the same die. "Tigerton" and "Clovertown" may feature four cores independantly or on the same die.
Let he who is without multiple cores cast the first stone

Last week at ISSCC, Intel was showing off private demonstrations of its quad core 65nm Xeon processors.  The chips, dubbed Tigerton and Clovertown, are multi-core versions of Intel's upcoming Conroe and Merom architecture.  Roadmaps from Intel have already revealed Merom will feature a 4MB L2 cache, 14 stage 4-issue pipeline and operate on just under 45W for two cores.  There was no mention of clock speed, or even if all four cores were functional on the CPUs, but the demonstration did include a PC running on the processors.  As to why there are two codenames for quad core CPUs, Intel representatives hinted to us that the layouts of the cores are different between the two CPUs. 

Intel also mentioned that the architecture used on these new processors scales well past 32 cores per chip.  As of now, the architecture is still unnamed but Intel has announced the new brand-wide architecture will get a name at the Intel Developer Forum a few weeks from now. 

More details about the new chips here.  AMD has separately announced that quad core Opteron processors will be available in 2007.  AMD typically demonstrates new chipsets and processors during IDF offsite. 

Comments     Threshold

This article is over a month old, voting and posting comments is disabled

By baddog121390 on 2/11/2006 12:25:28 PM , Rating: 2
the developers are already saying its a PITA to multithread code. isnt 4 cores going to make it worse? i would think 2 faster cores would be better than 4 slower cores...

RE: bleh
By KristopherKubicki on 2/11/2006 12:30:35 PM , Rating: 2
Server stuff has been multiple thread aware for decades. I would agree with you on desktop, but this should be a slam dunk for developers on servers.


RE: bleh
By TomZ on 2/11/2006 1:10:25 PM , Rating: 2
For applications where there is a performance benefit to having multiple threads, going from 2 to 4 threads is usually no big deal, e.g., video encoding. The most work is from going from 1 to 2 threads.

RE: bleh
By oTAL on 2/12/2006 7:35:22 PM , Rating: 2
Bullshit... it's the same... your dividing a process in threads and if you have only 2 you can't use 4... if want to use four you have to further thread your program which will still be hard anyway...

RE: bleh
By zsdersw on 2/11/2006 1:30:38 PM , Rating: 2
The discussion about whether multiple cores will mean more work for developers is irrelevant because both Intel and AMD are going down the 2+ core path.

RE: bleh
By Fenixgoon on 2/11/2006 2:14:21 PM , Rating: 2
so the question is, who will get their quad cores out first? also, will the past w/ multi cores happen (AMD went server first, intel went desktop first)?

RE: bleh
By Hulk on 2/12/2006 8:15:21 AM , Rating: 2
I can see how 4 cores could be useful to quickly get an application running multithreaded. Instead of having to figure out how to balance the work load between two cores programmers could simply assign tasks to specific cores. It wouldn't be the most efficient way to do it but could be done quickly.

For example, for video encoding, one core could handle rescaling, one could handle color adjustments, another other video effects, and the final one could actually do the compression.

RE: bleh
By marcuri on 2/13/2006 1:38:58 AM , Rating: 2
Good idea except the problem is i'm sure the way the program either puts data into the cpu's or just a limitation by order of operations you can't compress a movie that hasn't been finalized yet in terms of color correction and video effects. I guess you could do it frame by frame by how would that be any faster if each processor isn't specialized and all the same. It seems like theres an exponentially decreasing preformance advantage for each core added to a CPU when the program isn't optimized for that type of setup.

Scales well beyond 32 cores per chip?
By Griswold on 2/12/2006 2:25:22 PM , Rating: 2
I wonder if all these cores would still have to communicate via FSB...

RE: Scales well beyond 32 cores per chip?
By Xenoterranos on 2/12/2006 7:04:34 PM , Rating: 1
One would asume so since Intel would rather languish in hell fire than admit AMD was right. (And since intel's 64 bit offering is called EMT64 and not x86-64, they can be happy living in their own little world where they're still the technology kings)

By JackPack on 2/12/2006 9:00:58 PM , Rating: 2
At least get the acronym correct... it's EM64T.

By JackPack on 2/12/2006 8:55:28 PM , Rating: 2
It uses a unified L2, so traffic is a lot less.

By the time they scale to >4 cores, CSI will be in place.

Microsoft :)
By johnsaw on 2/11/2006 4:46:08 PM , Rating: 2
Well, certainly MS is heavily multithreading their new Windows Vista and Office 2007.
And Vista will be out when the 4core's ship.

Also, multithreading is not an issue on servers because they get a lot of simultaneous requests.

RE: Microsoft :)
By oTAL on 2/12/2006 7:38:24 PM , Rating: 2
depends... the application can handle all the requests in a single thread... but since serves have been multi-cpu for some time now that is no longer a problem... the problem exists for desktops... and it's not even a problem... it's a development thing... programmers get their work a little more complicated...

Server Chip!!
By ncage on 2/11/2006 10:04:43 PM , Rating: 2
Guys this is a server chip for crying out loud. Its not a desktop chip. 4 Cores will be great for a server chip for lots of things like web servering and virtual pc's.

RE: Server Chip!!
By Xenoterranos on 2/12/2006 7:02:54 PM , Rating: 2
Is "servering" a word?

architecture != application
By johnsaw on 2/11/2006 6:21:29 PM , Rating: 2
"Intel also mentioned that the architecture used on these new processors scales well past 32 cores per chip."

The talk was about one application that scaled up to 16 cores and not more:

"Before the changes, the application only ran well in simulations on chips with 16 cores. After that, performance began to decline, Rattner said.

After the changes, performance continued to climb. "We got it to scale well past 32" cores, he said."

BTW, this was a software simulation of multicore chips.

By zboss on 2/11/2006 6:55:00 PM , Rating: 2
In my experience: more cores=more time spent managing=more software defects. I would rather have a balance between faster cores that require less LOE managing threads and more cores that allow distribution of dissimilar tasks.

I suspect that this is what we will see, something akin to a seperate floating point processor of old.

By aGreenAgent on 2/13/2006 4:02:08 AM , Rating: 2
I'm not sure you guys noticed, but just because a chip has more than one core, doesn't mean the apps for it will go more than 1 core.

I can run 32 single-threaded apps on a dual core, and put 16 on each, for example.

I'm not sure why you're all so focused on things being multithreaded. Some things will be, some won't. Who cares, more cores still help (especially with the Windows Scheduler).

ya ya
By tjr508 on 2/12/2006 2:30:24 AM , Rating: 1
Funny how Intel and AMD are all gung ho about putting all these cores on the same chip and saving the need for extra heat sinks when they purposely remove all their oem chips from the retail market causing so many of their stupid HSFs to go to waste anyways.

I know theres more reasons for multi core (more so with AMDs ht bus than intel's fsb) i just thought I'd throw this out.

"We basically took a look at this situation and said, this is bullshit." -- Newegg Chief Legal Officer Lee Cheng's take on patent troll Soverain

Copyright 2016 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki