• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

How did AMD do dualcore?

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Jakalwarrior

Member
Joined
Apr 5, 2004
Location
Louisiana
I got to thinkin about this yesterday. How were they able to use the same chipsets and socket thats been around forever and have it magically work with dualcore? How does a motherboad use two chips at once through the same pins it was using for just one with the help of a bios update even though noone was thinking of dualcore when they were designing Nforce 3.
 
because one core goes "through" the other :) and they are using the same cores that were already being used on that socket, so they didnt need to increase the pin count i guess.
 
Because its much like a dual cpu environment just with 1 cpu 2 cores sharing all the same busses. So theoreticaly could you hit some kind of bottleneck? I think yes but in real world we don't hit that sence todays bus width and speeds outpase the data transfer of a cpu. A program needs to be multithreaded to be able to utilize a dual core.
Theoretically if I am running Adobe Photoshop and Doom 3 at the same time one core would handle Photoshop and one would handle Doom. But all data passes through the same busand same pins on the cpu. I am fairly uneducated on the way it works and when it works but this is what I think happenes. So making it compatible with the hardware was probably the least of AMD's worries as long as there is something there to assign each core to a process.
 
think about it this way though, if both cores are working at 100% then the mobo is recieving in my case 5.4 gigaherts (since speed is rated by how many pulses the processor makes per second, or so my comp teacher said). So both signals are coming in through the same pins even though they are seperate signals? How is a mobo designed to read one signal reading two at once? or does the chip have some extra stuff added in to sort all that out? Or has the motherboard always fed the chip more than it can handle, the chip takes the info, processes it in the specified priority and sends it back only now it gets it done faster because it can send it to two dif places. Bah im lost.
 
Gunner, I get what you mean by a bottleneck. We got 2 cores on 1 die sharing the same single memory controller. That should be a bottleneck right there but I think it actually works to the CPU's advantage. I noticed with my 3700+ San Deigo @ 300x9, I was pulling roughly 7,450mb/s memory bandwidth. With my Opteron 165 @ 300x9, It pulls significantly more, 7,810mb/s. Its almost like pressure cutters at dams, forcing a huge mass of water though an extremely small pathway causes it to go faster? I don't know...I'm just guessing :p
 
rseven said:
It's all made possible by magic pixie dust. It becomes crystal clear on the fifth drink! :beer:

The only time AMD and or myself hits a bottleneck is when we take a swig from our beers! :beer:

Anyway, Dual Core tech was under develpment back in like 2001... So yeah I forgot where I was going with that.... :bang head
 
Remember the HT link is capable of 2Ghz and has always been so for socket 939. The information from the two cores is handled by a crossbar switch and system request interface on the cpu (this and the extra core are the only differences that exist between a single core and dual core) so there is never any issue of the mobo not being able to "cope" as it has no northbridge.
 
because there not intel who rob you blind and want to keep you on single core :p

Oh and becasue the second chip is hummping the 1st wait till tre core is out :eek: may just jump to quad or five core! :shrug:
 
doublejack said:
it's like siamese twins that have 1 body but 2 heads.
essentially, yes.
a DC at 2.7ghz isnt really "5.4ghz" the "chip" is still operating at 2.7ghz, because thats its host clock speed, but it has 2 processing centers, thus it simply has more capacity at 2.7ghz, to do certian tasks.
 
OC Detective said:
Remember the HT link is capable of 2Ghz and has always been so for socket 939. The information from the two cores is handled by a crossbar switch and system request interface on the cpu (this and the extra core are the only differences that exist between a single core and dual core) so there is never any issue of the mobo not being able to "cope" as it has no northbridge.
AMD has also designed the K8 platform with dual cores in mind from step one, that's why most motherboards support this feature.
 
Intel on the other hand practically did dualcore by just chopping off every other cpu on the wafer...

cheap, yet inferior.
 
GunnerMan said:
Because its much like a dual cpu environment just with 1 cpu 2 cores sharing all the same busses. So theoreticaly could you hit some kind of bottleneck? I think yes but in real world we don't hit that sence todays bus width and speeds outpase the data transfer of a cpu. A program needs to be multithreaded to be able to utilize a dual core.
Theoretically if I am running Adobe Photoshop and Doom 3 at the same time one core would handle Photoshop and one would handle Doom. But all data passes through the same busand same pins on the cpu. I am fairly uneducated on the way it works and when it works but this is what I think happenes. So making it compatible with the hardware was probably the least of AMD's worries as long as there is something there to assign each core to a process.


Photoshop is multihreaded, therefore it would slow down doom3 if you ran them together.
 
Back