Intel's "proper" quad cores

[Participant]

Ok this might be a slightly contraversial title but whereabouts are Intel's first native quad cores in terms of their roadmap and what will they be? Will they have an onboard memory controller? I'm talking about chips with four or more cores that aren't butchered C2Ds!

 

[FreeStream]

Penryn, 3rd or 4th quarter of this year IIRC

and no onboard mem controller.

 

[Participant]

Oh, I thought Penryn's were just turbo-charged C2Qs?

 

[FreeStream]

Oh, I thought Penryn's were just turbo-charged C2Qs?

Nope, proper quad core IIRC

 

[Azza]

Ok this might be a slightly contraversial title but whereabouts are Intel's first native quad cores in terms of their roadmap and what will they be? Will they have an onboard memory controller? I'm talking about chips with four or more cores that aren't butchered C2Ds!

First proper Quad core CPUs should be seen Q3/Q4 this year.


Nehalem will be the Intel core architecture with an onboard memory controller. It is going to be 45nm like the Penryn revision stuff buts its going to be a new marchitecture. That is coming some time in the second half of 2008

 

[Participant]

Great stuff. Wonder if "Nehalem" will be S775?

 

[Azza]

Great stuff. Wonder if "Nehalem" will be S775?

I think there is going to be a new socket, I could be wrong though.

 

[Crowze]

Surely it would have to be if the memory controller is integrated rather than on the northbridge.

 

[ergonomics]

Surely it would have to be if the memory controller is integrated rather than on the northbridge.

tru.dat

the socket would have to change for such a change in architecture

 

[Popov]

Hmmm intergrated memory controllers, sounds like bad news for overclockers to me.

 

[V|per]

Hmmm intergrated memory controllers, sounds like bad news for overclockers to me.

may i ask why?

 

[FirebarUK]

AMD's chips have had onboard memory controllers for yonks (A64 etc).

 

[Stelly]

may i ask why?

No idea where are lots of pros and cons... for example....

Intel puts a lot of cache on its chips, which cuts down the need for an integrated memory controller

A manufacturer can only process a finite amount of silicon on its most advanced lines each year: you can't rush the process

As for overclocking I wouldnt think that it would make that much of a difference

Stelly

 

[mrthingyx]

Intel puts a lot of cache on its chips, which cuts down the need for an integrated memory controller

Stelly

I'm not sure about this one: the cache is useful for stuff that is regularly used by the CPU, but regardless, stuff ALWAYS needs to go to memory. As such, any increase in speed over the standard memory-CPU interaction is a bonus.

However, as RAM latencies increase, the improvement demonstrated by an integrated memory controller will slowly reduce, just as AMD's chips took a performance hit of varying degrees when moving from Skt 939 to AM2.

Core2Duo features a rather nifty prediction-type thing for memory fetches, effectively nullifying any advantage AMD64's integrated controller brought to the table. So it will be interesting to see what Nehalem's version can bring especially with the ridiculous latencies of DDR3.

 

[Durzel]

Memory controllers onchip are bad news for overclocking, it's another thing that can go wrong when you are on the bleeding edge.

 

[Combat squirrel]

Memory controllers onchip are bad news for overclocking, it's another thing that can go wrong when you are on the bleeding edge.

Er yeah, my Opterons on die mem controller really ruins its overclock

 

[Participant]

Agreed, these things are very complicated and that's like saying 4 wheel drive is a bad idea in rally-driving cause it's another thing that can go wrong.